Investment Management Market Test / BLOG
New thoughts heading into January 2023: Markets are shifting
Jan 4, 2023: Jeffrey Cohen, Investment Advisor Representative
US Advanced Computing Infrastructure, Inc.
Good morning. The market shifted. Our CQNS Long model seems to lose money every day. The high-BETA stocks are declining explosively. The hedge, shorting the S&P 500, is not generating significant returns. The index is where it was a month ago.
The 'interesting' stocks that make up our model, those with non-normal return characteristics and that 'zig and zag' to reduce price volatility are declining. The market has shifted. This is causing us to think deeply on how we optimize our portfolio. This is causing us to think about each type of non-normal return statistic to see if the market is looking for a different path forward.
Of course, there is fundamental analysis. We are working through some 'low price' stocks this week. We are creating simple valuations and looking at sustainability. Unfortunately, every single 'cheap stock' is cheap for a reason. They may not make it through the upcoming recession. Earnings are down and transformation is required to survive if we don't see a 'magic wand' of macro-economic growth increase sales and profits. Interest costs are rising, along with (surprisingly to us) SG&A costs, and these eat away gross operating profits.
There is also industry analysis. Semiconductors was a good, aggressive, high-BETA bet due to the chip shortage in 2021 and has been bleeding investors in 2022. Energy rose since 2020, but had a very down day yesterday (Jan 3, 2023) on recession fears. Even coal joined oil and gas to fall. Gold is rising, along with silver, but that is a monetary bet not an economic or productive industry bet. High Tech is really a bet on long-maturity returns, and the rising interest rate and inflation environment eats up those returns before they even happen. Put another way, why buy a biotech stock that can hit it big in 10 years, when today's money will be relatively worthless in 10 years?
Our CQNS shorts have declined significantly. There were significant profits in riding down the CQNS shorts in Q4 2022. However, now we are seeing some explosively positive days in those stocks, so patience is required. It looks like for every 3 shorts, one will rise in a given day. This is purely anecdotal, but appears to be a thing where investors are squeezing the occasional stock higher, or a few investors are covering large short positions, taking profits, and this just continues the anomalous behavior of high volatility and low expected returns.
As we look forward, the CQNS short model looks like a safe bet to profit in 2023. However, we need to tweak or recalibrate our CQNS long model to capture the correct metrics and create alpha. This is our job for January 2023. We are up to the challenge.
Good luck to all. GLTA
US Advanced Computing Infrastructure, Inc.
Good morning. The market shifted. Our CQNS Long model seems to lose money every day. The high-BETA stocks are declining explosively. The hedge, shorting the S&P 500, is not generating significant returns. The index is where it was a month ago.
The 'interesting' stocks that make up our model, those with non-normal return characteristics and that 'zig and zag' to reduce price volatility are declining. The market has shifted. This is causing us to think deeply on how we optimize our portfolio. This is causing us to think about each type of non-normal return statistic to see if the market is looking for a different path forward.
Of course, there is fundamental analysis. We are working through some 'low price' stocks this week. We are creating simple valuations and looking at sustainability. Unfortunately, every single 'cheap stock' is cheap for a reason. They may not make it through the upcoming recession. Earnings are down and transformation is required to survive if we don't see a 'magic wand' of macro-economic growth increase sales and profits. Interest costs are rising, along with (surprisingly to us) SG&A costs, and these eat away gross operating profits.
There is also industry analysis. Semiconductors was a good, aggressive, high-BETA bet due to the chip shortage in 2021 and has been bleeding investors in 2022. Energy rose since 2020, but had a very down day yesterday (Jan 3, 2023) on recession fears. Even coal joined oil and gas to fall. Gold is rising, along with silver, but that is a monetary bet not an economic or productive industry bet. High Tech is really a bet on long-maturity returns, and the rising interest rate and inflation environment eats up those returns before they even happen. Put another way, why buy a biotech stock that can hit it big in 10 years, when today's money will be relatively worthless in 10 years?
Our CQNS shorts have declined significantly. There were significant profits in riding down the CQNS shorts in Q4 2022. However, now we are seeing some explosively positive days in those stocks, so patience is required. It looks like for every 3 shorts, one will rise in a given day. This is purely anecdotal, but appears to be a thing where investors are squeezing the occasional stock higher, or a few investors are covering large short positions, taking profits, and this just continues the anomalous behavior of high volatility and low expected returns.
As we look forward, the CQNS short model looks like a safe bet to profit in 2023. However, we need to tweak or recalibrate our CQNS long model to capture the correct metrics and create alpha. This is our job for January 2023. We are up to the challenge.
Good luck to all. GLTA
Thinking about comparables to our model. How did we do against the market?
Was thinking about how to describe our results over the 52 trading days. The question we asked ourselves was how we performed against the broader market. We will look at the SPY and IWM as comparables. One is large caps, and one is small caps, US listed stocks.
SPY opens on August 31:$399.92
SPY closes on November 11: $398.51
Total SPY return -$1.41 / $399.92 = 0.35% return (1/3 of one percent)
Min: $356.56
Max: $410.97
Avg: $378.90
IWM on August 31: $184.94
IWM on November 11: $186.90
Total IWM return +$1.96 / $184.94 = 1.06% return (about one percent)
Min: $164.17
Max: $189.67
Avg: $176.06
Had you invested your money into SPY and IWM over the trading period, you would have broken even, roughly speaking. Had you invested in short-term bonds, SPY and IWM equally, you might have made 0.5%.
However, our trading model generated 20% returns on a notional $50,000 investment over the same period. During some points, our fund holdings were as high as almost $70,000, which implies some leverage. It is interesting that our model delivered these returns during a flat market. Our CQNS Short stocks delivered the most gains, despite a Russell 2000 rise during the period.
This shows one type of value in a smart volatility model. It can generate returns in a flat market.
Contact Jeffrey Cohen (see the Contact page for phone number and email address) for more information, and please read our brochure for more details on fees, expenses and how we do business.
Past performance is not a guarantee of future results. You can lose money in this investment (not FDIC insured). Please be aware of the unique risks of investing based on volatility and expected returns before investing.
SPY opens on August 31:$399.92
SPY closes on November 11: $398.51
Total SPY return -$1.41 / $399.92 = 0.35% return (1/3 of one percent)
Min: $356.56
Max: $410.97
Avg: $378.90
IWM on August 31: $184.94
IWM on November 11: $186.90
Total IWM return +$1.96 / $184.94 = 1.06% return (about one percent)
Min: $164.17
Max: $189.67
Avg: $176.06
Had you invested your money into SPY and IWM over the trading period, you would have broken even, roughly speaking. Had you invested in short-term bonds, SPY and IWM equally, you might have made 0.5%.
However, our trading model generated 20% returns on a notional $50,000 investment over the same period. During some points, our fund holdings were as high as almost $70,000, which implies some leverage. It is interesting that our model delivered these returns during a flat market. Our CQNS Short stocks delivered the most gains, despite a Russell 2000 rise during the period.
This shows one type of value in a smart volatility model. It can generate returns in a flat market.
Contact Jeffrey Cohen (see the Contact page for phone number and email address) for more information, and please read our brochure for more details on fees, expenses and how we do business.
Past performance is not a guarantee of future results. You can lose money in this investment (not FDIC insured). Please be aware of the unique risks of investing based on volatility and expected returns before investing.
Final results of trading our model for 52 days in a flat market
The results are conclusive, and positive. The Chicago Quantum Net Score / Smart Volatility model portfolio (paper traded each trading day) made money.
Dates:
We started our model trading on August 31, 2022 and traded the model daily through November 11, 2022 (at market open) when we liquidated all positions. On September 13, we initiated trading in CQNS short positions, and ended November 11, 2022.
Amounts invested on average:
CQNS Long: Average investment $12,752
SPY Short (BETA Weighted): Average investment $23,742
CQNS Short: $12,442
Methodology
We ran our CQNS Long and Short model every day after market close. We analyzed our results and planned our trades for the next day pre-market. We would use pre-market values most days (~75%) and closing prices ~25%.
We had two objectives each day in pre-market:
1. Ensure we had holdings that were consistent with the best 3 portfolios selected by the model. We could be off one or two positions (say within a few ticks), but the holdings had to represent the best portfolios selected.
2. Our positions were roughly equivalent in size. All holdings were in whole shares. We allowed for odd lots, but preferred round lots in our shorts (which tended to be lower cost shares). After a while we set a 4% +/- window and optimized that way. If a stock was too expensive to allow for adjusting, we let that one drift more (e.g., DDS trades around $300/share, and our average position size was $900. If it rose to $1,000, we could not adjust). This meant we sold winners and bought losers, consistently.
We traded within the first 5 minutes of stock trading each day, and sometimes faster (as quickly as we could manually entering trades sequentially). In production, we could enter trades at market open, but worry this could distort execution due to overly wide spreads at the open. We could also wait to 09:40am EST daily to enjoy lower spreads.
We had significantly more than $50,000 invested on multiple days. The most we had invested was $68,149, which included $50,000 in short positions. We had intended to only run a SPY hedge, but eventually decided to run the CQNS Short model in addition. It puts more money at risk, but it creates a new means of unlocking value. In our case, all three models were profitable.
For a $50,000 investment, this would mean that the investment would exceed the value of the investment and would have a margin balance. We could also reduce the size of all three models to ensure the total investment fits within the investment at all times.
So, when analyzing our results you should consider whether you would want a long/short model where we hedge longs with shorts, or whether you want a long/SPY or short/SPY hedge, or whether you want to run all three, the Long/Hedge and the Short model combined.
The Results
The results are calculated by looking at the basis of each model position each day and summing the daily investments. This is the sum of the amount paid, minus the amount received, each day. We do not take into account dividends, interest charges, or trading fees. We use a zero-commission broker, but there are sometimes SEC / exchange fees which are nominal.
CQNS Long Run: $819 profit
CQNS Short: $6,801 profit
BETA-Weighted SPY hedge (100% short on SPY multiplied by the BETA of the long portfolio): $2,383 profit
The total profit is $10,003 on a notional investment of $50,000.
This is a 20% profit in 3 months.
We calculated our results by tracking the change in basis to our positions daily. At the end of the trading period, our basis was negative in all three portfolio tranches. Caveats: We did not account for dividends, trading fees, borrowing fees, or interest expenses. We only accounted for our trades, at the real-time bid or ask, or in between the spread (~15% of the time), that we saw in our trading application at the moment of each trade. We do not see a bias introduced in that method. Prices were volatile and fluctuated widely during these morning trading sessions.
Other things to consider:
Total value invested varied over time.
Some days it was as low as $25,000 and other days it was as high as $68,000. This is due to the level of Risk in the model. Some days, the model wanted very few holdings. These are risk-on days, and we traded them with less money invested (same investment per stock, on fewer stocks), and we then lowered the amount of SPY hedge and CQNS shorts to match. Other days we had more invested because the model was Risk Off. On those days we invested in many more stocks (sometimes over 20 long positions), and we increased our hedges accordingly.
It is difficult to calculate the daily net asset value of the fund, and to know how or when to benchmark the value of the holdings. We found that for most of the trading days the pre-market moves were material to the results. If we rebalanced the portfolio based on prior night's closing prices, they could be materially different positions than when looking at pre-market prices even 15 minutes before market opening. We could look at share prices while the market is open, but we would run into the same problem, the prices have already moved. We found that by trading on pre-market prices, we had the most accurate positions heading into the trading day. We also found that the prices of many stocks had already moved into their new levels by market open. We did price stock changes during the day from our initial checkpoint (whether pre-market or prior night's close) against closing prices that day to see if the model made or lost money. However, the real profit was decided based on buys and sells, not on mark-to-market levels.
There were 2 or 3 day streaks in the results on the CQNS long portfolio. The markets would favor our portfolio, then it would fall out of favor. It was pointless to time these moves, as they appear somewhat random, and the gains were concentrated in a few days. The same thing with the CQNS Short positions. There were days when the short portfolio lost money. Never very much, but it has losing streaks. However, the gains could be material on any given day.
After we stopped trading our model, there were a few stocks that fell dramatically which were on our list. Chances are we would have benefitted from those moves, and profits on CQNS short portfolio would have continued.
Lessons Learned
High Turnover: We had significant portfolio turnover. Some is required such as initiating and terminating full positions at the beginning and end of the investments. Others were to remove and add stocks based on model results. In a few cases, the same stock was purchased and sold repeatedly. We tried to 'wait out' the model and required a 2 or 3 day sell or buy signal to make the trade, and that helped lower turnover. However, the models are intended to be updated each day. We did not find daily rebalancing of positions to be material. Buying or selling up to 5% of a few stocks in a day was not material.
Average daily turnover
CQNS Longs: 23%
SPY Hedge: 9%
CQNS Shorts: 15%
High Spreads: In some cases, the stocks we were trading had spreads of 5% at market open. We could see a $5.00 bid-ask spread at the open (in the first 90 seconds of trading). However, the spreads normally tightened if we let a few minutes pass. Most of the stocks picked by the model had high liquidity and low spreads, with many stocks trading with a $0.01 spread. We attempted to reduce trading in high-spread stocks (like Dillards Department Stores $DDS) but we just had to make those trades. The few stocks with high spreads were included in the model, and the transaction costs were tolerated and included in the results.
Pre-market movements. For many of these stocks, the moves in pre-market were the moves of the day. The stock might have moved 3% or 5% in pre-market, only to close at around the pre-market price. In this case, we were stuck possibly buying the model's recommendations after the day (during pre-market) had played out. This happened more in the September timeframe than in November, which could be due to market conditions. We still believe we should trade at or around market open (the first 90 seconds).
Accounting complexity. In this cycle of trading, we kept track of all trades on two separate sets of spreadsheets. One was set for the first 11 days (through and including September 15) and the next one was for 42 days including September 15. We changed the format of the spreadsheet and tracking as we learned more. In a few cases, our trades were not recorded properly in the spreadsheet and we might miss a position (these errors typically cost about $250 in accuracy apiece). We went back and cleaned up the spreadsheets, and fixed the accounting, but realize that the best answer is to fully complete fund accounting immediately after trades were made. So, the morning routine would be to plan the trades, make the trades, and account for the trades before getting up from the traders workstation. We also have a better understanding of the information we need around value at risk, total position sizes, and ongoing basis of positions that we will enable with the next cycle of investments.
Buying low and selling high: The model had us maintaining roughly equivalent positions in our stocks. This meant that if a stock went up 10%, we sold ~6% of the shares the next day. A stock that fell had us buying more. We do not know if this is the best strategy, but it allows the math to work. We ended up building up positions as all stocks rose, because all stocks had to be reset to the higher level. We believe that happened with the CQNS longs. We sized up a few hundred dollars as stocks rose.
Higher-capitalization stocks were over-weight in our model. It seems like many of our stock positions were stocks with over $10B or even $100B in market capitalization. We think this might be what brought those stocks into our model in the first place, they had a better risk-return tradeoff. So, even when we let our model pick stocks (currently) with $100mm market capitalizations, none of those small cap stocks make the grade and get included in the best portfolios. This model tends to favor larger stocks.
Dividend stocks. This model tends to favor stocks that pay a dividend. Not the highest dividends, like the LPs or REITS, but dividends of around the market (1% to 3%). We did not account for dividends, but over time they should boost reported gains. Stocks in companies that lose money tend to not have significant dividends.
We have learned a significant amount in this process, and plan to run another cycle shortly, this time with real money. If you are interested in investing, and can accommodate the risk (e.g., high net worth individual, corporation or portfolio manager), we welcome your inquiry.
We hope that was helpful. This period of time convinced us that the model has value and can generate profitable trades over weeks or months of consistent use.
Dates:
We started our model trading on August 31, 2022 and traded the model daily through November 11, 2022 (at market open) when we liquidated all positions. On September 13, we initiated trading in CQNS short positions, and ended November 11, 2022.
Amounts invested on average:
CQNS Long: Average investment $12,752
SPY Short (BETA Weighted): Average investment $23,742
CQNS Short: $12,442
Methodology
We ran our CQNS Long and Short model every day after market close. We analyzed our results and planned our trades for the next day pre-market. We would use pre-market values most days (~75%) and closing prices ~25%.
We had two objectives each day in pre-market:
1. Ensure we had holdings that were consistent with the best 3 portfolios selected by the model. We could be off one or two positions (say within a few ticks), but the holdings had to represent the best portfolios selected.
2. Our positions were roughly equivalent in size. All holdings were in whole shares. We allowed for odd lots, but preferred round lots in our shorts (which tended to be lower cost shares). After a while we set a 4% +/- window and optimized that way. If a stock was too expensive to allow for adjusting, we let that one drift more (e.g., DDS trades around $300/share, and our average position size was $900. If it rose to $1,000, we could not adjust). This meant we sold winners and bought losers, consistently.
We traded within the first 5 minutes of stock trading each day, and sometimes faster (as quickly as we could manually entering trades sequentially). In production, we could enter trades at market open, but worry this could distort execution due to overly wide spreads at the open. We could also wait to 09:40am EST daily to enjoy lower spreads.
We had significantly more than $50,000 invested on multiple days. The most we had invested was $68,149, which included $50,000 in short positions. We had intended to only run a SPY hedge, but eventually decided to run the CQNS Short model in addition. It puts more money at risk, but it creates a new means of unlocking value. In our case, all three models were profitable.
For a $50,000 investment, this would mean that the investment would exceed the value of the investment and would have a margin balance. We could also reduce the size of all three models to ensure the total investment fits within the investment at all times.
So, when analyzing our results you should consider whether you would want a long/short model where we hedge longs with shorts, or whether you want a long/SPY or short/SPY hedge, or whether you want to run all three, the Long/Hedge and the Short model combined.
The Results
The results are calculated by looking at the basis of each model position each day and summing the daily investments. This is the sum of the amount paid, minus the amount received, each day. We do not take into account dividends, interest charges, or trading fees. We use a zero-commission broker, but there are sometimes SEC / exchange fees which are nominal.
CQNS Long Run: $819 profit
CQNS Short: $6,801 profit
BETA-Weighted SPY hedge (100% short on SPY multiplied by the BETA of the long portfolio): $2,383 profit
The total profit is $10,003 on a notional investment of $50,000.
This is a 20% profit in 3 months.
We calculated our results by tracking the change in basis to our positions daily. At the end of the trading period, our basis was negative in all three portfolio tranches. Caveats: We did not account for dividends, trading fees, borrowing fees, or interest expenses. We only accounted for our trades, at the real-time bid or ask, or in between the spread (~15% of the time), that we saw in our trading application at the moment of each trade. We do not see a bias introduced in that method. Prices were volatile and fluctuated widely during these morning trading sessions.
Other things to consider:
Total value invested varied over time.
Some days it was as low as $25,000 and other days it was as high as $68,000. This is due to the level of Risk in the model. Some days, the model wanted very few holdings. These are risk-on days, and we traded them with less money invested (same investment per stock, on fewer stocks), and we then lowered the amount of SPY hedge and CQNS shorts to match. Other days we had more invested because the model was Risk Off. On those days we invested in many more stocks (sometimes over 20 long positions), and we increased our hedges accordingly.
It is difficult to calculate the daily net asset value of the fund, and to know how or when to benchmark the value of the holdings. We found that for most of the trading days the pre-market moves were material to the results. If we rebalanced the portfolio based on prior night's closing prices, they could be materially different positions than when looking at pre-market prices even 15 minutes before market opening. We could look at share prices while the market is open, but we would run into the same problem, the prices have already moved. We found that by trading on pre-market prices, we had the most accurate positions heading into the trading day. We also found that the prices of many stocks had already moved into their new levels by market open. We did price stock changes during the day from our initial checkpoint (whether pre-market or prior night's close) against closing prices that day to see if the model made or lost money. However, the real profit was decided based on buys and sells, not on mark-to-market levels.
There were 2 or 3 day streaks in the results on the CQNS long portfolio. The markets would favor our portfolio, then it would fall out of favor. It was pointless to time these moves, as they appear somewhat random, and the gains were concentrated in a few days. The same thing with the CQNS Short positions. There were days when the short portfolio lost money. Never very much, but it has losing streaks. However, the gains could be material on any given day.
After we stopped trading our model, there were a few stocks that fell dramatically which were on our list. Chances are we would have benefitted from those moves, and profits on CQNS short portfolio would have continued.
Lessons Learned
High Turnover: We had significant portfolio turnover. Some is required such as initiating and terminating full positions at the beginning and end of the investments. Others were to remove and add stocks based on model results. In a few cases, the same stock was purchased and sold repeatedly. We tried to 'wait out' the model and required a 2 or 3 day sell or buy signal to make the trade, and that helped lower turnover. However, the models are intended to be updated each day. We did not find daily rebalancing of positions to be material. Buying or selling up to 5% of a few stocks in a day was not material.
Average daily turnover
CQNS Longs: 23%
SPY Hedge: 9%
CQNS Shorts: 15%
High Spreads: In some cases, the stocks we were trading had spreads of 5% at market open. We could see a $5.00 bid-ask spread at the open (in the first 90 seconds of trading). However, the spreads normally tightened if we let a few minutes pass. Most of the stocks picked by the model had high liquidity and low spreads, with many stocks trading with a $0.01 spread. We attempted to reduce trading in high-spread stocks (like Dillards Department Stores $DDS) but we just had to make those trades. The few stocks with high spreads were included in the model, and the transaction costs were tolerated and included in the results.
Pre-market movements. For many of these stocks, the moves in pre-market were the moves of the day. The stock might have moved 3% or 5% in pre-market, only to close at around the pre-market price. In this case, we were stuck possibly buying the model's recommendations after the day (during pre-market) had played out. This happened more in the September timeframe than in November, which could be due to market conditions. We still believe we should trade at or around market open (the first 90 seconds).
Accounting complexity. In this cycle of trading, we kept track of all trades on two separate sets of spreadsheets. One was set for the first 11 days (through and including September 15) and the next one was for 42 days including September 15. We changed the format of the spreadsheet and tracking as we learned more. In a few cases, our trades were not recorded properly in the spreadsheet and we might miss a position (these errors typically cost about $250 in accuracy apiece). We went back and cleaned up the spreadsheets, and fixed the accounting, but realize that the best answer is to fully complete fund accounting immediately after trades were made. So, the morning routine would be to plan the trades, make the trades, and account for the trades before getting up from the traders workstation. We also have a better understanding of the information we need around value at risk, total position sizes, and ongoing basis of positions that we will enable with the next cycle of investments.
Buying low and selling high: The model had us maintaining roughly equivalent positions in our stocks. This meant that if a stock went up 10%, we sold ~6% of the shares the next day. A stock that fell had us buying more. We do not know if this is the best strategy, but it allows the math to work. We ended up building up positions as all stocks rose, because all stocks had to be reset to the higher level. We believe that happened with the CQNS longs. We sized up a few hundred dollars as stocks rose.
Higher-capitalization stocks were over-weight in our model. It seems like many of our stock positions were stocks with over $10B or even $100B in market capitalization. We think this might be what brought those stocks into our model in the first place, they had a better risk-return tradeoff. So, even when we let our model pick stocks (currently) with $100mm market capitalizations, none of those small cap stocks make the grade and get included in the best portfolios. This model tends to favor larger stocks.
Dividend stocks. This model tends to favor stocks that pay a dividend. Not the highest dividends, like the LPs or REITS, but dividends of around the market (1% to 3%). We did not account for dividends, but over time they should boost reported gains. Stocks in companies that lose money tend to not have significant dividends.
We have learned a significant amount in this process, and plan to run another cycle shortly, this time with real money. If you are interested in investing, and can accommodate the risk (e.g., high net worth individual, corporation or portfolio manager), we welcome your inquiry.
We hope that was helpful. This period of time convinced us that the model has value and can generate profitable trades over weeks or months of consistent use.
Nov 11: How did we do in a very aggressive post-election, post-CPI market rise?
Good morning.
We made some directional trades based on our 'feelings' and 'intuitive sense' of how the market was performing and not based on the model. It is odd to see the market move so dramatically, and we adjusted our model portfolio to be over-hedged. In short, we sold the long positions that the model said to sell, but we did not buy the positions it said to buy. So, we ended up selling 12 of 19 long positions. They were up, but 9 of the 12 positions closed on the day higher than our sell price. So, we left money on the table.
We also only reduced our hedges by 20% instead of the almost 60% reduction in longs. Those hedges rose, which ate into profits on Thursday. Let's see how we ended up.
If we assume we held the long positions until the end of the day with the extreme hedges, the model made money.
Notional profit vs. weighted SPY: $690 on the day
Notional profit vs. CQNS Shorts: $560 on the day
Once we account for the sells, and only calculate the remaining 7 stocks? (this is a sense of how we do today if yesterday repeats exactly the same):
Notional loss vs. weighted SPY: $69
Notional loss vs. CQNS Shorts: $195
So, what we learn is that when the market is rising, it is profitable to let it rise and stay invested. The risk of those positions is increasing, and we see that in the model run from last night. However, all things being equal, we would have increased investor profitability if we would have kept our long/short balanced yesterday. Good lesson learned about the cost of portfolio manager intuition on a model's stock market returns.
On the other hand, if the market falls today, we are over-hedged.
Our intention is to rebalance this morning with the market's picks. We ran the models last night (SPY, QQQ and Short) and we see smaller portfolios of higher BETA stocks, and our portfolio today is well positioned against that model portfolio. Not many buys and sells required today to get there.
Closing price variance of different stock populations:
SPY Variance: up 1 tick.
QQQ Variance: up 2 ticks.
Profitable stocks: up 1 tick.
Unprofitable stocks: up 3 ticks.
Net result of this very simplistic analysis is that unprofitable stocks should fall 'the most' to justify holding more volatility, QQQ should fall a little less, and SPY/Profitable company stocks should fall the least.
KTA: In this market, please invest in stocks of profitable companies. They have the least pressure to fall at this moment.
We made some directional trades based on our 'feelings' and 'intuitive sense' of how the market was performing and not based on the model. It is odd to see the market move so dramatically, and we adjusted our model portfolio to be over-hedged. In short, we sold the long positions that the model said to sell, but we did not buy the positions it said to buy. So, we ended up selling 12 of 19 long positions. They were up, but 9 of the 12 positions closed on the day higher than our sell price. So, we left money on the table.
We also only reduced our hedges by 20% instead of the almost 60% reduction in longs. Those hedges rose, which ate into profits on Thursday. Let's see how we ended up.
If we assume we held the long positions until the end of the day with the extreme hedges, the model made money.
Notional profit vs. weighted SPY: $690 on the day
Notional profit vs. CQNS Shorts: $560 on the day
Once we account for the sells, and only calculate the remaining 7 stocks? (this is a sense of how we do today if yesterday repeats exactly the same):
Notional loss vs. weighted SPY: $69
Notional loss vs. CQNS Shorts: $195
So, what we learn is that when the market is rising, it is profitable to let it rise and stay invested. The risk of those positions is increasing, and we see that in the model run from last night. However, all things being equal, we would have increased investor profitability if we would have kept our long/short balanced yesterday. Good lesson learned about the cost of portfolio manager intuition on a model's stock market returns.
On the other hand, if the market falls today, we are over-hedged.
Our intention is to rebalance this morning with the market's picks. We ran the models last night (SPY, QQQ and Short) and we see smaller portfolios of higher BETA stocks, and our portfolio today is well positioned against that model portfolio. Not many buys and sells required today to get there.
Closing price variance of different stock populations:
SPY Variance: up 1 tick.
QQQ Variance: up 2 ticks.
Profitable stocks: up 1 tick.
Unprofitable stocks: up 3 ticks.
Net result of this very simplistic analysis is that unprofitable stocks should fall 'the most' to justify holding more volatility, QQQ should fall a little less, and SPY/Profitable company stocks should fall the least.
KTA: In this market, please invest in stocks of profitable companies. They have the least pressure to fall at this moment.
Nov 9: The day after US elections
Update:
It ended up being a very good day for our CQNS Down 'short' hedged managed account model portfolio.
The portfolio earned more than 1% yesterday during a volatile day.
How?
Our 19 equally weighted long positions fell 3% due to their higher than 1.0 BETA value.
Our 3 equally weighted short positions fell 6.9% due to a day where 'stonks' fell aggressively as #crypto fell.
Our BETA Weighted SPY short position fell 2.1%, but due to the BETA weight, it fell covered 93% of the long move.
So, if you chose the BETA Weighted SPY hedge (the most efficient hedge), you roughly broke even yesterday.
However, if you chose the CQNS Short / CQNS Long portfolio (the greatest use of the model), the hedge paid 226%.
We will be accounting for the different models shortly as we 'reset' our model portfolio back to a starting balance of $50,000. We expect to find that the BETA Weighted SPY portfolio has less volatility and less gain than the CQNS Short/Long portfolio. Let's find out together and move forward.
On the other hand, we have learned a great deal by running a QQQ-based model this week. We again (for Thursday, Nov 10, 2022) have a 66% stock / 33% bond portfolio coming out of our seeded monte carlo solver (an interim solution). We also have smaller solutions as we lowered the risk free rate of return to 3.75% from 4.25%. The short end of the yield curve, and the availability and size of bank money market interest rates, has fallen this week, necessitating the change.
GLTA
It ended up being a very good day for our CQNS Down 'short' hedged managed account model portfolio.
The portfolio earned more than 1% yesterday during a volatile day.
How?
Our 19 equally weighted long positions fell 3% due to their higher than 1.0 BETA value.
Our 3 equally weighted short positions fell 6.9% due to a day where 'stonks' fell aggressively as #crypto fell.
Our BETA Weighted SPY short position fell 2.1%, but due to the BETA weight, it fell covered 93% of the long move.
So, if you chose the BETA Weighted SPY hedge (the most efficient hedge), you roughly broke even yesterday.
However, if you chose the CQNS Short / CQNS Long portfolio (the greatest use of the model), the hedge paid 226%.
We will be accounting for the different models shortly as we 'reset' our model portfolio back to a starting balance of $50,000. We expect to find that the BETA Weighted SPY portfolio has less volatility and less gain than the CQNS Short/Long portfolio. Let's find out together and move forward.
On the other hand, we have learned a great deal by running a QQQ-based model this week. We again (for Thursday, Nov 10, 2022) have a 66% stock / 33% bond portfolio coming out of our seeded monte carlo solver (an interim solution). We also have smaller solutions as we lowered the risk free rate of return to 3.75% from 4.25%. The short end of the yield curve, and the availability and size of bank money market interest rates, has fallen this week, necessitating the change.
GLTA
Good morning. Analyzing our results this morning from last night's run.
Somebody broke the stock market today. Volumes are too low. Crypto crashing could be the cause, but not sure.
The level 2 book from NASDAQ that we purchase is showing very thin bids and asks on the stocks we are looking at right now.
This is a 'wait and see' day in the markets.
We are not trading our model portfolio this morning. Letting or positions from yesterday ride to avoid significant spreads.
A few things we see in our model run. This is based on data from market close yesterday and does not reflect today's markets.
1. Our QQQ Seeded Monte Carlo analysis (seeded with individual CQNS top picks) selected a three-stock portfolio which has 1/4 the alpha of the 'best' or most optimized portfolio found. It is to hold 1/3 QQQ, SHY, and SPY. This is a surprisingly competent portfolio that most people can buy, hold and forget about. It suggests a 2/3 investment spread across the two primary US equity indices, and 1/3 in short-term bonds (but not cash). The old literature suggests stocks and cash, but this is more like the 60/40 stock/bond portfolios you see today.
The same solver in the SPY run selected QQQ, SPY, AMD, INTU and STT (so, 40% passive equities, 40% growth / tech stocks, and 20% conservative financial services. A different solution, with only ~35% of the alpha as the best optimized CQNS portfolio found.
2. There is moderate overlap between the QQQ and SPY based run.
Best QQQ-based US equity portfolio found:
-0.000092 ['AAPL', 'AMD', 'AMZN', 'APPS', 'BX', 'FTNT', 'INTU', 'MELI', 'META', 'MXL', 'NVDA', 'RKT', 'SYNA', 'TSLA'] 14
Slightly larger version:
-0.000092 ['AAPL', 'AMD', 'AMZN', 'APPS', 'BX', 'EWCZ', 'FTNT', 'INTU', 'MELI', 'META', 'MSFT', 'MXL', 'NVDA', 'RKT', 'SYNA', 'TSLA'] 16
Best SPY-based US equity portfolio found:
-0.000177 ['APPS', 'BX', 'COIN', 'GT', 'MELI', 'NVDA', 'SITM', 'SYNA', 'TGI', 'UEC', 'VRT'] 11
Slightly larger version:
-0.000177 ['AMD', 'APPS', 'BX', 'COIN', 'GT', 'MELI', 'NVDA', 'RKT', 'SITM', 'SYNA', 'TGI', 'UEC', 'VRT'] 13
Overlap: APPS BX MELI NVDA SYNA
SPY ONLY: AAPL AMD AMZN FTNT INTU META MXL RKT TSLA
QQQ ONLY: COIN GT SITM TGI UEC VRT
3. We now have our model incorporating a number of ETFs. This helps us to better understand BETA, or market performance correlations, against non-stock assets (like VIX, commodities or interest rates).
QQQ-based:
Here are the stocks with BETA values outside the data validation range.
SQQQ ProShares UltraPro Short QQQ -3x Shares -2.946
TLT iShares 20+ Year Treasury Bond ETF -0.008
TZA Direxion Daily Small Cap Bear 3X Shares -2.372
UUP Invesco DB US Dollar Index Bullish Fund -0.108
UVXY ProShares Ultra VIX Short-Term Futures ETF 2x Shares -2.741
VIXY ProShares VIX Short-Term Futures ETF -1.858
SPY-based:
Here are the stocks with BETA values outside the data validation range.
SQQQ ProShares UltraPro Short QQQ -3x Shares -3.760
TLT iShares 20+ Year Treasury Bond ETF -0.015
TZA Direxion Daily Small Cap Bear 3X Shares -3.225
UUP Invesco DB US Dollar Index Bullish Fund -0.166
UVXY ProShares Ultra VIX Short-Term Futures ETF 2x Shares -3.759
VIXY ProShares VIX Short-Term Futures ETF -2.549
4. There is overlap in the top 10 stocks picked individually in our model.
Top 10 stocks in QQQ-based portfolios:
Here are 50 Stocks with the best, or lowest, CQNS scores.
0 QQQ Invesco QQQ Trust Series 1 -0.00006
1 SPY S&P 500 ETF TRUST ETF -0.00004
2 SHY iShares 1-3 Year Treasury Bond ETF -0.00003
3 IWM iShares Russell 2000 ETF 0.00001
4 PNM PNM Resources Inc 0.00003
5 MSFT Microsoft Corporation 0.00003
6 APH Amphenol Corp. 0.00004
7 AAPL Apple Inc 0.00004
8 EQC Equity Commonwealth 0.00005
9 AAAU Goldman Sachs Physical Gold ETF 0.00005
Top 10 stocks in SPY-based portfolios:
Here are 50 Stocks with the best, or lowest, CQNS scores.
0 SPY S&P 500 ETF TRUST ETF -0.00003
1 QQQ Invesco QQQ Trust Series 1 -0.00002
2 SHY iShares 1-3 Year Treasury Bond ETF -0.00001
3 IWM iShares Russell 2000 ETF 0.00002
4 APH Amphenol Corp. 0.00004
5 PNM PNM Resources Inc 0.00005
6 AME Ametek Inc 0.00006
7 BLK Blackrock Inc. 0.00007
8 AAAU Goldman Sachs Physical Gold ETF 0.00007
9 MSFT Microsoft Corporation 0.00007
There is a significant difference in the variance between the indices and stock populations:
SPY Variance: 0.00021436
QQQ Variance: 0.00037756
All Stock Variance (profitable) = 0.0002455
All Stock Variance (unprofitable) = -0.0007194
Somebody broke the stock market today. Volumes are too low. Crypto crashing could be the cause, but not sure.
The level 2 book from NASDAQ that we purchase is showing very thin bids and asks on the stocks we are looking at right now.
This is a 'wait and see' day in the markets.
We are not trading our model portfolio this morning. Letting or positions from yesterday ride to avoid significant spreads.
A few things we see in our model run. This is based on data from market close yesterday and does not reflect today's markets.
1. Our QQQ Seeded Monte Carlo analysis (seeded with individual CQNS top picks) selected a three-stock portfolio which has 1/4 the alpha of the 'best' or most optimized portfolio found. It is to hold 1/3 QQQ, SHY, and SPY. This is a surprisingly competent portfolio that most people can buy, hold and forget about. It suggests a 2/3 investment spread across the two primary US equity indices, and 1/3 in short-term bonds (but not cash). The old literature suggests stocks and cash, but this is more like the 60/40 stock/bond portfolios you see today.
The same solver in the SPY run selected QQQ, SPY, AMD, INTU and STT (so, 40% passive equities, 40% growth / tech stocks, and 20% conservative financial services. A different solution, with only ~35% of the alpha as the best optimized CQNS portfolio found.
2. There is moderate overlap between the QQQ and SPY based run.
Best QQQ-based US equity portfolio found:
-0.000092 ['AAPL', 'AMD', 'AMZN', 'APPS', 'BX', 'FTNT', 'INTU', 'MELI', 'META', 'MXL', 'NVDA', 'RKT', 'SYNA', 'TSLA'] 14
Slightly larger version:
-0.000092 ['AAPL', 'AMD', 'AMZN', 'APPS', 'BX', 'EWCZ', 'FTNT', 'INTU', 'MELI', 'META', 'MSFT', 'MXL', 'NVDA', 'RKT', 'SYNA', 'TSLA'] 16
Best SPY-based US equity portfolio found:
-0.000177 ['APPS', 'BX', 'COIN', 'GT', 'MELI', 'NVDA', 'SITM', 'SYNA', 'TGI', 'UEC', 'VRT'] 11
Slightly larger version:
-0.000177 ['AMD', 'APPS', 'BX', 'COIN', 'GT', 'MELI', 'NVDA', 'RKT', 'SITM', 'SYNA', 'TGI', 'UEC', 'VRT'] 13
Overlap: APPS BX MELI NVDA SYNA
SPY ONLY: AAPL AMD AMZN FTNT INTU META MXL RKT TSLA
QQQ ONLY: COIN GT SITM TGI UEC VRT
3. We now have our model incorporating a number of ETFs. This helps us to better understand BETA, or market performance correlations, against non-stock assets (like VIX, commodities or interest rates).
QQQ-based:
Here are the stocks with BETA values outside the data validation range.
SQQQ ProShares UltraPro Short QQQ -3x Shares -2.946
TLT iShares 20+ Year Treasury Bond ETF -0.008
TZA Direxion Daily Small Cap Bear 3X Shares -2.372
UUP Invesco DB US Dollar Index Bullish Fund -0.108
UVXY ProShares Ultra VIX Short-Term Futures ETF 2x Shares -2.741
VIXY ProShares VIX Short-Term Futures ETF -1.858
SPY-based:
Here are the stocks with BETA values outside the data validation range.
SQQQ ProShares UltraPro Short QQQ -3x Shares -3.760
TLT iShares 20+ Year Treasury Bond ETF -0.015
TZA Direxion Daily Small Cap Bear 3X Shares -3.225
UUP Invesco DB US Dollar Index Bullish Fund -0.166
UVXY ProShares Ultra VIX Short-Term Futures ETF 2x Shares -3.759
VIXY ProShares VIX Short-Term Futures ETF -2.549
4. There is overlap in the top 10 stocks picked individually in our model.
Top 10 stocks in QQQ-based portfolios:
Here are 50 Stocks with the best, or lowest, CQNS scores.
0 QQQ Invesco QQQ Trust Series 1 -0.00006
1 SPY S&P 500 ETF TRUST ETF -0.00004
2 SHY iShares 1-3 Year Treasury Bond ETF -0.00003
3 IWM iShares Russell 2000 ETF 0.00001
4 PNM PNM Resources Inc 0.00003
5 MSFT Microsoft Corporation 0.00003
6 APH Amphenol Corp. 0.00004
7 AAPL Apple Inc 0.00004
8 EQC Equity Commonwealth 0.00005
9 AAAU Goldman Sachs Physical Gold ETF 0.00005
Top 10 stocks in SPY-based portfolios:
Here are 50 Stocks with the best, or lowest, CQNS scores.
0 SPY S&P 500 ETF TRUST ETF -0.00003
1 QQQ Invesco QQQ Trust Series 1 -0.00002
2 SHY iShares 1-3 Year Treasury Bond ETF -0.00001
3 IWM iShares Russell 2000 ETF 0.00002
4 APH Amphenol Corp. 0.00004
5 PNM PNM Resources Inc 0.00005
6 AME Ametek Inc 0.00006
7 BLK Blackrock Inc. 0.00007
8 AAAU Goldman Sachs Physical Gold ETF 0.00007
9 MSFT Microsoft Corporation 0.00007
There is a significant difference in the variance between the indices and stock populations:
SPY Variance: 0.00021436
QQQ Variance: 0.00037756
All Stock Variance (profitable) = 0.0002455
All Stock Variance (unprofitable) = -0.0007194
Nov 8: Gained the insight we needed on QQQ and SPY, and made additional adjustments
Will update everyone on the QQQ learnings.
We reduced the risk-free rate of interest to 3.75% as short term interest rates have fallen and banks are paying a max of 3.0% on deposits. We continue to monitor interest rates.
We may do a first accounting after a month of paper trading our model portfolio. Will report that here too.
Finally, we are working on our brochure.
We reduced the risk-free rate of interest to 3.75% as short term interest rates have fallen and banks are paying a max of 3.0% on deposits. We continue to monitor interest rates.
We may do a first accounting after a month of paper trading our model portfolio. Will report that here too.
Finally, we are working on our brochure.
Nov 6: Should we pivot away from the SPY and towards the QQQ for our baseline, or model standard index?
One thought is that the QQQ is down the most of any of the indices we follow, or approximately 31% including dividends. Also, the QQQ contains most of our stock picks, and all of our largest picks. Therefore, if the market is trading around the QQQ on individual stocks, maybe we should reset our stock picks to that index.
In the past when all indices were rising, the choice of index did not matter much at all. The model picked mostly the same stocks regardless of whether we used SPY or QQQ. We will take a look today and see if there are differences. (some model coding required, as we lost this functionality during numerous upgrades to the code).
Update: We ran it successfully and now have a QQQ run. It has a different solution for the 'best' optimized portfolio.
Technically, we ran the solvers twice, with the second run using the answers from the first run (which should help it).
We did not change anything else, only the use of QQQ vs. SPY as the basis of the model run (BETA calculations, market returns & market variance).
So, what happened?
When comparing the two solutions:
After running the model both ways and seeing the difference in the runs, we now have to decide what to do with those differences in Model Portfolio.
It appears the QQQ-based portfolio has lower risk, just based on the QQQ performance over the past year.
Stay tuned for more updates.
In the past when all indices were rising, the choice of index did not matter much at all. The model picked mostly the same stocks regardless of whether we used SPY or QQQ. We will take a look today and see if there are differences. (some model coding required, as we lost this functionality during numerous upgrades to the code).
Update: We ran it successfully and now have a QQQ run. It has a different solution for the 'best' optimized portfolio.
Technically, we ran the solvers twice, with the second run using the answers from the first run (which should help it).
We did not change anything else, only the use of QQQ vs. SPY as the basis of the model run (BETA calculations, market returns & market variance).
So, what happened?
- The model portfolio (long) based on the QQQ is significantly larger than for SPY (contains more stocks).
- There are 4 ticks more edge, going from 51 to 55 ticks.
- There are a few exceptionally low variance stocks in the long portfolio that we don't ever see in the SPY portfolio.
- Interestingly enough, it has us holding GOOG and GOOGL, effectively double weighting Alphabet. Not sure we would do that, but interesting all the same.
- It contains the NASDAQ Company stock: NDAQ.
- Both the SPY and QQQ portfolios can scale up in size significantly with little change in 'alpha' or the CQNS score. 30% more stocks in the SPY long portfolio and ~50% more stocks for QQQ.We can stay within one tick (10-6) of the best portfolio and add 50% more stocks.
When comparing the two solutions:
- SPY & QQQ agree: 12 stocks
- SPY include but not QQQ: 10 stocks
- QQQ include, but not SPY: 17 stocks
After running the model both ways and seeing the difference in the runs, we now have to decide what to do with those differences in Model Portfolio.
It appears the QQQ-based portfolio has lower risk, just based on the QQQ performance over the past year.
Stay tuned for more updates.
Nov 6: Friday's action causes more churn in our CQNS long picks
Two thoughts as we process our CQNS Long portfolio:
1. If we market capitalization weight our CQNS Long holdings, only 5 of them would have 3% or larger position sizes. The rest (about 20) would be very small or even insignificant holdings (less than 1%). The challenge is that we calculate the model using equal weighting, so the model requires equal weighting. However, the smaller capitalization stocks move the most. Food for thought.
2. After processing the run after Friday, November 4, 2022 market close, the model suggests we sell 6 positions (one is QRTEA which now has negative market capitalization), and we buy 5 positions. This is a significant turnover in our model, and requires a decision on whether we make the full set of changes (as we did earlier this week), or phase them in. One reasonable portfolio management compromise is to incorporate the one change which was due to a flip in net income, then incorporate 2 buys and sell out of the 5 buys and sells. We could buy and sell either the highest or lowest market capitalizations? We could also do it by lowest bid/ask spread, which would reduce the transaction costs of the moves.
1. If we market capitalization weight our CQNS Long holdings, only 5 of them would have 3% or larger position sizes. The rest (about 20) would be very small or even insignificant holdings (less than 1%). The challenge is that we calculate the model using equal weighting, so the model requires equal weighting. However, the smaller capitalization stocks move the most. Food for thought.
2. After processing the run after Friday, November 4, 2022 market close, the model suggests we sell 6 positions (one is QRTEA which now has negative market capitalization), and we buy 5 positions. This is a significant turnover in our model, and requires a decision on whether we make the full set of changes (as we did earlier this week), or phase them in. One reasonable portfolio management compromise is to incorporate the one change which was due to a flip in net income, then incorporate 2 buys and sell out of the 5 buys and sells. We could buy and sell either the highest or lowest market capitalizations? We could also do it by lowest bid/ask spread, which would reduce the transaction costs of the moves.
Nov 6: Earnings matter to the model portfolio
We are in a position to have a larger number of longs. On Friday, earnings came out for one of our longs. The company lost money. They took a non-cash charge. However, that stock fell 21.2% on the news (a true bloodbath) and is now forced to be removed from our CQNS Long portfolio. That portfolio only holds profitable companies.
This raises a question on what we do to account for companies that are about to release earnings? Should we remove those 'news' 'story' 'earnings' stocks to avoid the shocks, or do we keep them in and let them run through the results? In the case of $QRTEA, we kept the stock in, and they fell 22.1% on the day, and now the next day we need to sell the entire position.
This raises a question on what we do to account for companies that are about to release earnings? Should we remove those 'news' 'story' 'earnings' stocks to avoid the shocks, or do we keep them in and let them run through the results? In the case of $QRTEA, we kept the stock in, and they fell 22.1% on the day, and now the next day we need to sell the entire position.
Oct 31: At the end of the day...one hedge broke even and the other lost 0.6% on our investments.
Another day of tight action. Our stocks fell and our hedges fell either the exact same amount, or slightly less.
14 stocks vs. 3 hedge stocks.
14 stocks vs. 3 hedge stocks.
Oct 31: Friday's bear market rally pushed up our hedges
Good morning. It is Monday, October 31, and we are heading into FOMC week. The US equity, commodity and bond futures are all lower. Friday, it seems all the risk assets were higher.
Today, we are taking no action. The portfolio ended the day on Friday almost perfectly balanced. The model lost a little money (0.9% loss exactly with both hedges, which itself is a coincidence).
The model suggests offloading AMZN this morning due to the increased volatility. However, if we are to keep it, it wants us to 'dilute' the addition by adding 5 other stocks, and taking our stock count from 14 to 19. In addition, there are other NASDAQ stocks (tech stocks) to add as well. We decided to run with the portfolio as is (2nd best portfolio, plus AMZN).
On the shorts, there is a stock that we think is a MEME stock, and is showing up in our model CQNS Down run. We have ignored it for the past week. It is $BORR and it is coming in as the #2 stock to avoid in this market.
On Friday, the model looked like it did well. Of the 14 stocks, 13 were higher on the day. What's interesting is that we started the day using the pre-market data which in itself was higher than the prior evening's closing price. Therefore, all of the movement before the market opened was incorporated in our smoothing or ensuring the position sizes were equal across all longs and shorts (+/- 4%). This week, we will employ our 'smoothing' and portfolio balancing is based on prior evening close prices. We will test the results this week.
As a reminder, this is a paper traded, hedged fund that we are looking to offer to the investing public (high net worth individuals, funds, institutional investors) once we establish and harden the processes and evaluate fine tuning adjustments.
Good luck in the markets today. GLTA.
Today, we are taking no action. The portfolio ended the day on Friday almost perfectly balanced. The model lost a little money (0.9% loss exactly with both hedges, which itself is a coincidence).
The model suggests offloading AMZN this morning due to the increased volatility. However, if we are to keep it, it wants us to 'dilute' the addition by adding 5 other stocks, and taking our stock count from 14 to 19. In addition, there are other NASDAQ stocks (tech stocks) to add as well. We decided to run with the portfolio as is (2nd best portfolio, plus AMZN).
On the shorts, there is a stock that we think is a MEME stock, and is showing up in our model CQNS Down run. We have ignored it for the past week. It is $BORR and it is coming in as the #2 stock to avoid in this market.
On Friday, the model looked like it did well. Of the 14 stocks, 13 were higher on the day. What's interesting is that we started the day using the pre-market data which in itself was higher than the prior evening's closing price. Therefore, all of the movement before the market opened was incorporated in our smoothing or ensuring the position sizes were equal across all longs and shorts (+/- 4%). This week, we will employ our 'smoothing' and portfolio balancing is based on prior evening close prices. We will test the results this week.
As a reminder, this is a paper traded, hedged fund that we are looking to offer to the investing public (high net worth individuals, funds, institutional investors) once we establish and harden the processes and evaluate fine tuning adjustments.
Good luck in the markets today. GLTA.
Oct 23: Another week in a bear market
Hope everyone is having a great weekend. In Chicagoland, we are having spectacular weather for fall, and it is only after spending many hours in the sun that we are thinking about our week ahead (Sunday at around 4:30pm).
For the coming week, our model is picking most of the same stocks. For Monday, there are a few CQNS Down stocks that we could swap, but likely will hold 'as-is' as the differences in edges is minor. We could swap them out if the expected benefit outweighs the transaction costs. There is also a question of 'news stocks' or 'earnings stocks.' One of the newly recommended stock released earnings late last week and was very volatile. Not sure we want to add that into our model until we review the news (or we just decide to 'trust the process' and swap another position for this new 'news' stock.
A few other things we notice in this weekend's analysis:
The price volatility of stocks is up again for the year ending October 21 (market close). This should reduce the price levels of stocks and increase their expected returns. However, this week the market was erratic and so this just adds risk for investors.
There is less 'edge' or 'alpha' in our model this week. It almost feels like the S&P 500 Index ETF blew out some of the correlations due to targeted bets on that index. To that end, all BETA values of all of our 12 long investments are lower today than they were in the middle of last week. The correlation between stocks and the index is slightly lower. This is interesting.
This week was a pretty good week for the model. We made money. We didn't have much of a draw down, even on the worst days.
We boosted our holdings about 2x (by adding 5 or 6 positions to our current 12 long positions). Each one is approximately the same amount (+/- 4% each morning - rebalanced), and the entire portfolio is hedged two ways. We are maintain two hedges, and tracking results, for future investors to have a choice on how to hedge their investments.
One hedge is to BETA weight the S&P 500 ETF SPY, and to hold that. We recalculate our long BETA each morning as well, and on Friday it was 2.33.
Another hedge is to pick three stocks from our CQNS Down run that have excessive volatility compared to their expected future returns. We hold those three stocks evenly, and short them.
We are assuming a 9.5% expected annual return to market risk assets (meaning return to index funds), and a 4% risk-free rate of return. This is despite a very down market. We assume that new money coming into the market expects a positive return over the next year, including dividends. Otherwise, there would be no 'new money' flowing into equities.
For the coming week, our model is picking most of the same stocks. For Monday, there are a few CQNS Down stocks that we could swap, but likely will hold 'as-is' as the differences in edges is minor. We could swap them out if the expected benefit outweighs the transaction costs. There is also a question of 'news stocks' or 'earnings stocks.' One of the newly recommended stock released earnings late last week and was very volatile. Not sure we want to add that into our model until we review the news (or we just decide to 'trust the process' and swap another position for this new 'news' stock.
A few other things we notice in this weekend's analysis:
The price volatility of stocks is up again for the year ending October 21 (market close). This should reduce the price levels of stocks and increase their expected returns. However, this week the market was erratic and so this just adds risk for investors.
There is less 'edge' or 'alpha' in our model this week. It almost feels like the S&P 500 Index ETF blew out some of the correlations due to targeted bets on that index. To that end, all BETA values of all of our 12 long investments are lower today than they were in the middle of last week. The correlation between stocks and the index is slightly lower. This is interesting.
This week was a pretty good week for the model. We made money. We didn't have much of a draw down, even on the worst days.
We boosted our holdings about 2x (by adding 5 or 6 positions to our current 12 long positions). Each one is approximately the same amount (+/- 4% each morning - rebalanced), and the entire portfolio is hedged two ways. We are maintain two hedges, and tracking results, for future investors to have a choice on how to hedge their investments.
One hedge is to BETA weight the S&P 500 ETF SPY, and to hold that. We recalculate our long BETA each morning as well, and on Friday it was 2.33.
Another hedge is to pick three stocks from our CQNS Down run that have excessive volatility compared to their expected future returns. We hold those three stocks evenly, and short them.
We are assuming a 9.5% expected annual return to market risk assets (meaning return to index funds), and a 4% risk-free rate of return. This is despite a very down market. We assume that new money coming into the market expects a positive return over the next year, including dividends. Otherwise, there would be no 'new money' flowing into equities.
Oct 18: Maintain the high-risk positioning (and retain cash cushion)
Good morning. We ran our model yesterday with more stocks. We lowered market capitalization thresholds to $500M and see a few 'new faces' in our stock picks. We spent the morning considering whether to reduce risk, by increasing our number of long positions and increasing our short positions.
Let's discuss why to hold as-is:
1. The new stocks have tiny market capitalizations. This likely adds risk. The two largest market cap stocks would be 93% of the total invested amount if this were an equity capitalization weighted model. The new positions would be 22% of the long position, but only 0.6% to market cap weighting.
2. The new stocks lower the 'edge' of the model by 6%
3. We would add money into both our longs and shorts, reducing our cash cushion as liquidity is falling.
4. Pre-market action is aggressively positive, after an aggressively positive pre-market yesterday.
Let's discuss why to change by adding 2 long stocks:
5. We would increase diversification and only reduce the model's edge by 6%.
6. We reduce the BETA weighting of the longs
7. We diversify industries with these new picks.
However, we decided not to add the two long positions. We will let it ride.
Question: Do we market capitalization weight the portfolio?
Pro: market capitalization weighted indices perform better (e.g., S&P 500 Index)
Pro: we can add this code into the model, although it adds complexity and relies on another fundamental data point being accurate.
Con: The model today looks at optimizing an evenly-weighted portfolio. It is a change in strategy for an unknown payoff.
Con: We may have an edge by over-weighting small-cap stocks.
We did hear back from our custodian that our brochure needs to be updated with fund information, and we are updating it now.
Good luck in the markets today. GLTA
Let's discuss why to hold as-is:
1. The new stocks have tiny market capitalizations. This likely adds risk. The two largest market cap stocks would be 93% of the total invested amount if this were an equity capitalization weighted model. The new positions would be 22% of the long position, but only 0.6% to market cap weighting.
2. The new stocks lower the 'edge' of the model by 6%
3. We would add money into both our longs and shorts, reducing our cash cushion as liquidity is falling.
4. Pre-market action is aggressively positive, after an aggressively positive pre-market yesterday.
Let's discuss why to change by adding 2 long stocks:
5. We would increase diversification and only reduce the model's edge by 6%.
6. We reduce the BETA weighting of the longs
7. We diversify industries with these new picks.
However, we decided not to add the two long positions. We will let it ride.
Question: Do we market capitalization weight the portfolio?
Pro: market capitalization weighted indices perform better (e.g., S&P 500 Index)
Pro: we can add this code into the model, although it adds complexity and relies on another fundamental data point being accurate.
Con: The model today looks at optimizing an evenly-weighted portfolio. It is a change in strategy for an unknown payoff.
Con: We may have an edge by over-weighting small-cap stocks.
We did hear back from our custodian that our brochure needs to be updated with fund information, and we are updating it now.
Good luck in the markets today. GLTA
Oct 17: More of the same
Good morning. Today's pre-market action and the CQNS model runs over the weekend have us mostly holding set. We buy a tiny amount of one long, and sell a tiny amount of two shorts. That's it today.
A few things at play.
1. Market fell and rose and fell last week, but the stocks seem to be following the market. Not much divergence, so the model picks are almost identical.
2. Market is up in pre-market to almost exactly match Friday morning's relative portfolio positions.
So, we wait.
GLTA
A few things at play.
1. Market fell and rose and fell last week, but the stocks seem to be following the market. Not much divergence, so the model picks are almost identical.
2. Market is up in pre-market to almost exactly match Friday morning's relative portfolio positions.
So, we wait.
GLTA
Oct 13: Are we having fun yet?
Simple post today.
1. We believe that when markets fall, we have to re-evaluate our model inputs.
Interest rates continue to rise, and we have raised our risk-free interest rate to 3.75%.
Equity valuations continue to fall (indices down SPY -17.4%, IWM -23.5% and QQQ -26.5% over the past year, excluding dividends). We raised our expectation of future earnings, including dividends, to 9.50%.
2. This rise, coupled with specific stock trading history, has reduced the number of long positions we are taking. They are evenly sized where they were before, so overall our long positions are smaller than last week. The higher the expected return, the greater the edge of taking a RISK ON approach, and so we are taking it.
3. The reduction in long stock positions (down ~50% this past week) means we have to unwind our two short hedge positions. We carry two hedge positions as a test. They are creating the profits in our fund.
One is to short the SPY BETA weighted (so we take a straight hedge (long value / SPY) and multiply the number of shares by the BETA of our long portfolio. It is over 2.0, so we are holding over 200% of short SPY.
Two is a set of three CQNS DOWN Run stocks that we short. We have been buying back shares in those positions (making each one smaller to match our smaller long position. The math is (long value / 3 / DOWN share price). So, each DOWN short equals 1/3 of our entire long position. We concentrate our short positions to avoid diversification of this hedge.
4. We made our trades at market open, as we always do. Earlier this week we bought ~2% intra-day due to select stocks falling, and realize this was likely counter-productive. We will not be doing that again. The positions are equal after morning trading, then the market does it's work.
Our long positions are: APPS COIN MELI NVDA SI SITM VRT
Our short positions are: AMC EVBG DM
Our model called to replace DM as it fell below $750M in equity market capitalization, but we are waiting to save on transaction costs.
1. We believe that when markets fall, we have to re-evaluate our model inputs.
Interest rates continue to rise, and we have raised our risk-free interest rate to 3.75%.
Equity valuations continue to fall (indices down SPY -17.4%, IWM -23.5% and QQQ -26.5% over the past year, excluding dividends). We raised our expectation of future earnings, including dividends, to 9.50%.
2. This rise, coupled with specific stock trading history, has reduced the number of long positions we are taking. They are evenly sized where they were before, so overall our long positions are smaller than last week. The higher the expected return, the greater the edge of taking a RISK ON approach, and so we are taking it.
3. The reduction in long stock positions (down ~50% this past week) means we have to unwind our two short hedge positions. We carry two hedge positions as a test. They are creating the profits in our fund.
One is to short the SPY BETA weighted (so we take a straight hedge (long value / SPY) and multiply the number of shares by the BETA of our long portfolio. It is over 2.0, so we are holding over 200% of short SPY.
Two is a set of three CQNS DOWN Run stocks that we short. We have been buying back shares in those positions (making each one smaller to match our smaller long position. The math is (long value / 3 / DOWN share price). So, each DOWN short equals 1/3 of our entire long position. We concentrate our short positions to avoid diversification of this hedge.
4. We made our trades at market open, as we always do. Earlier this week we bought ~2% intra-day due to select stocks falling, and realize this was likely counter-productive. We will not be doing that again. The positions are equal after morning trading, then the market does it's work.
Our long positions are: APPS COIN MELI NVDA SI SITM VRT
Our short positions are: AMC EVBG DM
Our model called to replace DM as it fell below $750M in equity market capitalization, but we are waiting to save on transaction costs.
Last night while we were doing fund accounting work (executing our end-of-day processes to value our positions), we were also preparing to run our CQNS models to get ready for tomorrow.
We have a solver that used to work well and recently has been faltering. We had removed it from our job stream, and felt nostalgically bad about it. Like a horse set out to pasture. A war hero in a nursing home. A dog without a bone.
This is a monte carlo solver (pure random chance) that he had seeded with 'winning' or 'allstar' stocks. This was a hypothesis we had in our first academic article pre-print in arXiv, where we talked about allstars and dogstars. Well, we have come to find out that in today's market, taking a portfolio of individual top performers does not do very well. They are too highly correlated. Birds of a feather flocking together, and all that.
We coded last night, and will test today, a new model. We will implement Project BETAMAX. Stay tuned for news on this most exciting event. We are hoping that by pre-seeding and focusing our monte carlo solver on portfolio sizes that we believe are relevant and useful, we can quickly identify (or find) a very good stock portfolio solution.
As we go to print, the market this morning has shifted. Our CQNS DOWN Run hedges are all up (unprofitable for us), and our longs are up as well. The SPY is down, so that hedge is still profitable. Stay tuned for more updates.
This looks like a short squeeze.
We have a solver that used to work well and recently has been faltering. We had removed it from our job stream, and felt nostalgically bad about it. Like a horse set out to pasture. A war hero in a nursing home. A dog without a bone.
This is a monte carlo solver (pure random chance) that he had seeded with 'winning' or 'allstar' stocks. This was a hypothesis we had in our first academic article pre-print in arXiv, where we talked about allstars and dogstars. Well, we have come to find out that in today's market, taking a portfolio of individual top performers does not do very well. They are too highly correlated. Birds of a feather flocking together, and all that.
We coded last night, and will test today, a new model. We will implement Project BETAMAX. Stay tuned for news on this most exciting event. We are hoping that by pre-seeding and focusing our monte carlo solver on portfolio sizes that we believe are relevant and useful, we can quickly identify (or find) a very good stock portfolio solution.
As we go to print, the market this morning has shifted. Our CQNS DOWN Run hedges are all up (unprofitable for us), and our longs are up as well. The SPY is down, so that hedge is still profitable. Stay tuned for more updates.
This looks like a short squeeze.
Oct 12: More risk as the prices of stocks fall further
We increased our market risk expected return by 0.5% (over the next year) and the number of stocks fell by 5 in the 2nd best portfolio. It would have fallen by 6 had we selected the best one (but we are reducing transaction costs / volumes). Our expected return is now 9.5% over the next year (including dividends), and our risk-free rate is 3.75%.
This means we are cutting back our long investments by 5/12 or about 40%, with matching reductions in our hedges. We are not necessarily raising cash because we are buying back our shorts, but we are deleveraging into inflation CPI reporting on Thursday.
We are also swapping out one of our CQNS Down shorts (so, 2 the same as yesterday and one new position).
As an aside, we also increased our stock pool by lowering the minimum market capitalization threshold for our run to $750M from $1B. This slightly increased our count of stock tickers. It did introduce a new short into our CQNS Down Hedge that had fallen below $1B recently.
Finally, we do have one individual stock that now has a better risk-return trade-off than the S&P 500 Index ETF SPY, and that is Nvidia NVDA. It has a slightly negative CQNS score in our UP run, making it the top stock to hold (if you can only hold one).
This means we are cutting back our long investments by 5/12 or about 40%, with matching reductions in our hedges. We are not necessarily raising cash because we are buying back our shorts, but we are deleveraging into inflation CPI reporting on Thursday.
We are also swapping out one of our CQNS Down shorts (so, 2 the same as yesterday and one new position).
As an aside, we also increased our stock pool by lowering the minimum market capitalization threshold for our run to $750M from $1B. This slightly increased our count of stock tickers. It did introduce a new short into our CQNS Down Hedge that had fallen below $1B recently.
Finally, we do have one individual stock that now has a better risk-return trade-off than the S&P 500 Index ETF SPY, and that is Nvidia NVDA. It has a slightly negative CQNS score in our UP run, making it the top stock to hold (if you can only hold one).
Oct 11: Same long & short portfolios for almost a week!
We notice something strange in our model. As the overall value of stocks is declining, we are making small adjustments 'up' in expected returns. We are at a 9% return on equity forward looking. We also have raised our riskfree return slightly to account for higher interest rates, and are now using 3.75%. With (9%, 3.75%) selected, we are seeing a 10-stock long portfolio that has held fairly consistent. We have held a few stocks that show up in the top 5 portfolios to reduce trading costs (as they do not impact the overall 'alpha' or CQNS score significantly).
We also notice that the same ~20 stocks are showing up as the highest risk stocks with the least amount of relative expected return. The top 3 individual 'down' stocks have remained the same for almost two weeks!
So, our model is only being slightly tuned each morning with the buys and sells being almost insignificant (just to keep positions balanced, and hedges as close to 100% as possible each morning at market open).
We also notice that the same ~20 stocks are showing up as the highest risk stocks with the least amount of relative expected return. The top 3 individual 'down' stocks have remained the same for almost two weeks!
So, our model is only being slightly tuned each morning with the buys and sells being almost insignificant (just to keep positions balanced, and hedges as close to 100% as possible each morning at market open).
October 10: More risk...positive returns to our hedged model
Update: 0916:40
Using pre-market pricing, we will be adding to five of our long positions which keeps our holding more 'equal' to each other. We have to bring up some of the positions that fell the most. These are not large buys. Each buy is less than 10% of the value of the individual position.
We are keeping our SPY hedge (BETA Weighted) the same, as the SPY is up in pre-market while many individual stock names are lower.
We are adding to all three of our CQNS Down run, or short hedge positions. We will be shorting those three names (a little more for each) at the open.
So our strategy today in the market is to buy longs that are down in price and to short shorts that are down in price.
Using pre-market pricing, we will be adding to five of our long positions which keeps our holding more 'equal' to each other. We have to bring up some of the positions that fell the most. These are not large buys. Each buy is less than 10% of the value of the individual position.
We are keeping our SPY hedge (BETA Weighted) the same, as the SPY is up in pre-market while many individual stock names are lower.
We are adding to all three of our CQNS Down run, or short hedge positions. We will be shorting those three names (a little more for each) at the open.
So our strategy today in the market is to buy longs that are down in price and to short shorts that are down in price.
Good morning. Monday, October 10, 2022. We continue to trade our new Chicago Quantum Net Score / Smart Volatility model fund. We are paper-trading this new fund, working through the best (of 3) hedging strategies. We are looking at when and how often to trade. Lots of questions...lots of processes to work through and document.
However, the model continues to generate small and consistent profits on a hedged basis. What we mean is that when the market rises, and when the market falls, this fund generates profit most of the time. It is not based on the overall direction of the market.
We did notice that in very aggressive market conditions, the hedges may not act as expected. The fund's hedges work best in 'choppy' weather, but not necessarily during a hurricane or tsunami on a daily basis.
One change over the weekend.
We increased our risk-free rate of return due to increasing risk-free interest rates, and we also increased the expected return to new money invested in risk assets. That slight (1/4 percentage) net increase in risk did not change the results of our model.
Why increase the risk premium of the market? When stock prices fall significantly, it implies a higher expected return in the future, if all else is held constant. We see a sustained and consistent period of US equities weakness and see the declines in the level of the S&P 500 Index. We expect that new money put into the market is expecting an even greater return. We are not increasing that return by very much (it is up 1.5% after a market decline of ~10%), or a 15% recovery of recent losses over the next year. However, the lower the market goes, without structural harm to our economy, the greater the expectation of a market recovery in prices (by 10%, 20% or even 40% of those losses in the first year).
We are not making a statement on timing. We don't know when the market would recover, but we assert that it will and that new money invested in the stock market during a dramatic decline is expecting higher future returns then when the markets are calm and steadily rising.
A few notes:
1. No stocks have negative BETA anymore.
2. Volatility (stock adjusted closing prices) for all stocks in both the UP and DOWN runs is increasing.
3. No stocks have a negative CQNS Score, meaning no individual stocks have a better risk-return trade-off than the whole market. This is not a 'stock picker' market. Stocks do achieve a negative CQNS UP rating or score if held together. There are stocks that zig and zag together and improve their risk-return profile when held in portfolios. This is what our model and fund capitalize on.
4. Fewer stocks (more accurately, fewer tickers) traded on Friday in US markets.
5. Last Friday, profitable company stocks traded 99.5% of last year's average trading volume, and unprofitable company stocks traded 102.1%.
Gotta go! Futures were red (or down) this morning for US equities.
We will probably have to add to our longs, or cut back on our hedges this morning.
US fixed income markets (bond markets) are closed today to celebrate the National holiday.
Good luck to all
GLTA!
However, the model continues to generate small and consistent profits on a hedged basis. What we mean is that when the market rises, and when the market falls, this fund generates profit most of the time. It is not based on the overall direction of the market.
We did notice that in very aggressive market conditions, the hedges may not act as expected. The fund's hedges work best in 'choppy' weather, but not necessarily during a hurricane or tsunami on a daily basis.
One change over the weekend.
We increased our risk-free rate of return due to increasing risk-free interest rates, and we also increased the expected return to new money invested in risk assets. That slight (1/4 percentage) net increase in risk did not change the results of our model.
Why increase the risk premium of the market? When stock prices fall significantly, it implies a higher expected return in the future, if all else is held constant. We see a sustained and consistent period of US equities weakness and see the declines in the level of the S&P 500 Index. We expect that new money put into the market is expecting an even greater return. We are not increasing that return by very much (it is up 1.5% after a market decline of ~10%), or a 15% recovery of recent losses over the next year. However, the lower the market goes, without structural harm to our economy, the greater the expectation of a market recovery in prices (by 10%, 20% or even 40% of those losses in the first year).
We are not making a statement on timing. We don't know when the market would recover, but we assert that it will and that new money invested in the stock market during a dramatic decline is expecting higher future returns then when the markets are calm and steadily rising.
A few notes:
1. No stocks have negative BETA anymore.
2. Volatility (stock adjusted closing prices) for all stocks in both the UP and DOWN runs is increasing.
3. No stocks have a negative CQNS Score, meaning no individual stocks have a better risk-return trade-off than the whole market. This is not a 'stock picker' market. Stocks do achieve a negative CQNS UP rating or score if held together. There are stocks that zig and zag together and improve their risk-return profile when held in portfolios. This is what our model and fund capitalize on.
4. Fewer stocks (more accurately, fewer tickers) traded on Friday in US markets.
5. Last Friday, profitable company stocks traded 99.5% of last year's average trading volume, and unprofitable company stocks traded 102.1%.
Gotta go! Futures were red (or down) this morning for US equities.
We will probably have to add to our longs, or cut back on our hedges this morning.
US fixed income markets (bond markets) are closed today to celebrate the National holiday.
Good luck to all
GLTA!
October 6: Selling off longs, cutting back on shorts, raising cash
Good morning, Thursday October 6. The model has been running and producing consistent results.
The model now runs faster, which makes it easier for us to review the data in the evening. Thanks goes to our bespoke (custom) simulated annealing and genetic algorithms. You guys get a raise and promotion once python scripts become sentient. Long live the AI revolution.
For us mortals, our model and Mrs. Market is making our fund circle the wagons a bit today.
1. Our BETA of our long positions is now 2.27. It was 1.82 about 3 weeks ago. More risk in the longs. Not our doing, these are many of the same stocks...just moving more closely with the SPY. Not sure this is a good thing, but we trade the market not our feelings.
2. Selling one stock from our longs.
3. Selling tiny amounts of 3 longs to balance the portfolio (+/- 4%) to have equally sized positions.
4. Buying back ~ 6% of our CQNS Short / Down stock hedges
5. Buying back ~8% of our SPY beta weighted hedges.
This will raise a little cash from our longs, but 'put back' or spend more cash to cover our hedges. Model is moving slightly to cash as our longs become more risky / higher BETA.
Good luck out there.
The model now runs faster, which makes it easier for us to review the data in the evening. Thanks goes to our bespoke (custom) simulated annealing and genetic algorithms. You guys get a raise and promotion once python scripts become sentient. Long live the AI revolution.
For us mortals, our model and Mrs. Market is making our fund circle the wagons a bit today.
1. Our BETA of our long positions is now 2.27. It was 1.82 about 3 weeks ago. More risk in the longs. Not our doing, these are many of the same stocks...just moving more closely with the SPY. Not sure this is a good thing, but we trade the market not our feelings.
2. Selling one stock from our longs.
3. Selling tiny amounts of 3 longs to balance the portfolio (+/- 4%) to have equally sized positions.
4. Buying back ~ 6% of our CQNS Short / Down stock hedges
5. Buying back ~8% of our SPY beta weighted hedges.
This will raise a little cash from our longs, but 'put back' or spend more cash to cover our hedges. Model is moving slightly to cash as our longs become more risky / higher BETA.
Good luck out there.
October 5: Digesting big gains in the market on October 4
Good morning. Happy Yom Kippur. Hope those who celebrate have a safe and 'easy' fast.
The market has had two very bullish and UP days this week. We saw individual stocks in our model climb 10% in a single day.
However, our model's profit was based on the hedging strategy chosen. If you chose to hedge with the CQNS Down Run stocks, you lost money on the day. If you hedged with the BETA Weighted SPY, you made money. This is because the CQNS Down Run stocks, which are the most risky, non-biotech stocks in the stock market, rose faster than the longs. They fall faster, which leads to profit on down days, but they also rise faster.
This morning, pre-market, the markets are set to open lower. The SPY is trading down 1% pre-market, and there is not a dry eye in the theatre (all stocks UP and Down are either trading lower pre-market, or untraded / unchanged.
Well, process-wise we update all position values based on current pre-market trading values. To that end, we have a fair, current valuation of all positions. We set those long positions to be roughly equivalent to each other (+/- 4%).
So, what is our game-plan today? How does the model account for a large gain, followed by losses in pre-market?
Well, the model chose roughly the same portfolio. There is one stock that we 'should' sell from our longs, but we will wait to see if the recommendation repeats. We just bought that stock yesterday and it is in the 12th best portfolio. The shorts are unchanged.
1. We are selling small amounts of stock in 54% of our long positions.
2. We are covering, or buying back ~3% of our hedge in the beta weighted SPY
3. We are covering, or buying back a small amount of all of our CQNS Down Run hedges.
On a technical note:
1. We had a programming / logic break-through yesterday once the market closed. We tuned our genetic algorithm and simulated annealer solvers to run faster, and in the case of the simulated annealer, to run significantly better. We have been running these solvers for 2 years, and this will save us important minutes in running our job. Last night was the fastest run for us, ever, and we picked the same portfolio. We may pop champagne (or at least a bubbly Rose') tonight after sundown.
2. The BETA weighting is very important in our model. We realize that it dampens down performance, and limits upside. However, it significantly limits the downside of the model as well. Yesterday, we made money because our long stock positions beat the BETA Weighted SPY, and so we were profitable.
Good luck to all in the markets today.
The market has had two very bullish and UP days this week. We saw individual stocks in our model climb 10% in a single day.
However, our model's profit was based on the hedging strategy chosen. If you chose to hedge with the CQNS Down Run stocks, you lost money on the day. If you hedged with the BETA Weighted SPY, you made money. This is because the CQNS Down Run stocks, which are the most risky, non-biotech stocks in the stock market, rose faster than the longs. They fall faster, which leads to profit on down days, but they also rise faster.
This morning, pre-market, the markets are set to open lower. The SPY is trading down 1% pre-market, and there is not a dry eye in the theatre (all stocks UP and Down are either trading lower pre-market, or untraded / unchanged.
Well, process-wise we update all position values based on current pre-market trading values. To that end, we have a fair, current valuation of all positions. We set those long positions to be roughly equivalent to each other (+/- 4%).
So, what is our game-plan today? How does the model account for a large gain, followed by losses in pre-market?
Well, the model chose roughly the same portfolio. There is one stock that we 'should' sell from our longs, but we will wait to see if the recommendation repeats. We just bought that stock yesterday and it is in the 12th best portfolio. The shorts are unchanged.
1. We are selling small amounts of stock in 54% of our long positions.
2. We are covering, or buying back ~3% of our hedge in the beta weighted SPY
3. We are covering, or buying back a small amount of all of our CQNS Down Run hedges.
On a technical note:
1. We had a programming / logic break-through yesterday once the market closed. We tuned our genetic algorithm and simulated annealer solvers to run faster, and in the case of the simulated annealer, to run significantly better. We have been running these solvers for 2 years, and this will save us important minutes in running our job. Last night was the fastest run for us, ever, and we picked the same portfolio. We may pop champagne (or at least a bubbly Rose') tonight after sundown.
2. The BETA weighting is very important in our model. We realize that it dampens down performance, and limits upside. However, it significantly limits the downside of the model as well. Yesterday, we made money because our long stock positions beat the BETA Weighted SPY, and so we were profitable.
Good luck to all in the markets today.
October 3: Full Steam Ahead
We did well on Friday. Friday was a scary day (for example with our personal portfolio) where interest rates moved violently, stocks were down, and rumors abound.
Our long stocks (equal sized positions) changed $1.00 (not a typo) from the last pre-market price update we made. We keep checking pre-market pricing to ensure we don't have to change our buy or sell sizes, and to ensure accurate dollar cost averaging (DCA). Our shorts or hedges against the SPY and our CQNS Down Run stocks were profitable, so overall the model made money.
Monday, here is our plan if pre-market does not have significant changes from Friday's close:
1. Add a long position (equal size)
2. DCA down a long position
3. Increase hedges to reflect drop on Friday and cover the new position (net DCA).
We fine-tune in pre-market & execute at open.
For some additional color:
1. This is a notional, paper-traded $50,000 fund that buys equally sized long positions, and hedges them two ways (depending on investor tastes). They are either hedged based on our CQNS DOWN Run stocks (*we hold max 3 stocks short, equal sizes), or we hedge against the SPY or S&P 500 Index ETF BETA weighted. So, if the overall portfolio has an average BETA weighting, against the SPY, is 2.0, then we short twice the value of of our long positions in the SPY.
2. We run our model every evening after the market closes. It gives us a good feel on stock volatility (Friday's read is unchanged). It gives us the CQNS UP portfolios (about 100 portfolios in priority order). The best portfolios typically have the fewest stocks, and we usually tick down the list slightly to incorporate positions we have. This reduces slightly our edge or ALPHA but it also decreases transaction costs. This week, we are swapping out two CQNS Down Run stocks (buying one back to cover, and selling short a new one). We are also adding one long position to complete the 6th best portfolio, and to minimize trading activity.
3. We try to keep our positions the same size. Of course, this isn't possible due to some stock prices being higher than the +/- or degree of precision, and that we can only buy or sell whole numbers of shares. So, today, we sell one share of our best performing stock on Friday to keep position sizes in line.
This model keeps generating value for us during what has been a challenging two week period. The hedges carry most days, and the longs have also profited a few days.
Overall, we have about $12,000 in longs, $12,000 in shorts, and our hedge against the SPY is approximately $24,000. When we go into production, a $50,000 fund investment will look like a $20,000 long, $20,000 short CQNS down, or a $15,000 long and $30,000 short SPY. These have different risk characteristics.
One big question for us today:
Should we keep our expected return on new money into the market (inclusive of dividends) at 8.5% for the upcoming year?
The model is most sensitive to this assumption.
Higher expected returns would mean fewer positions held, or great concentration into a few high-BETA names.
Lower expected returns makes us look like a trust department of a stodgy asset manager with widow and orphan money, with significantly more stocks, more diversification, and little concentration risk.
At the extreme, an expected return of 90% (we ran this as a test), suggests we hold one stock, and fully hedge it.
We will keep our expected return at 8.5% for today, but are considering whether to raise the expected return as the market continues to fall further (say to 3,500). An investor would need to be compensated more for new money investments, which coincides with lower stock prices. We would reflect those increased expectations (inverse to price / value).
GLTA
The SPY has fallen to under 3,600 on Friday.
Our long stocks (equal sized positions) changed $1.00 (not a typo) from the last pre-market price update we made. We keep checking pre-market pricing to ensure we don't have to change our buy or sell sizes, and to ensure accurate dollar cost averaging (DCA). Our shorts or hedges against the SPY and our CQNS Down Run stocks were profitable, so overall the model made money.
Monday, here is our plan if pre-market does not have significant changes from Friday's close:
1. Add a long position (equal size)
2. DCA down a long position
3. Increase hedges to reflect drop on Friday and cover the new position (net DCA).
We fine-tune in pre-market & execute at open.
For some additional color:
1. This is a notional, paper-traded $50,000 fund that buys equally sized long positions, and hedges them two ways (depending on investor tastes). They are either hedged based on our CQNS DOWN Run stocks (*we hold max 3 stocks short, equal sizes), or we hedge against the SPY or S&P 500 Index ETF BETA weighted. So, if the overall portfolio has an average BETA weighting, against the SPY, is 2.0, then we short twice the value of of our long positions in the SPY.
2. We run our model every evening after the market closes. It gives us a good feel on stock volatility (Friday's read is unchanged). It gives us the CQNS UP portfolios (about 100 portfolios in priority order). The best portfolios typically have the fewest stocks, and we usually tick down the list slightly to incorporate positions we have. This reduces slightly our edge or ALPHA but it also decreases transaction costs. This week, we are swapping out two CQNS Down Run stocks (buying one back to cover, and selling short a new one). We are also adding one long position to complete the 6th best portfolio, and to minimize trading activity.
3. We try to keep our positions the same size. Of course, this isn't possible due to some stock prices being higher than the +/- or degree of precision, and that we can only buy or sell whole numbers of shares. So, today, we sell one share of our best performing stock on Friday to keep position sizes in line.
This model keeps generating value for us during what has been a challenging two week period. The hedges carry most days, and the longs have also profited a few days.
Overall, we have about $12,000 in longs, $12,000 in shorts, and our hedge against the SPY is approximately $24,000. When we go into production, a $50,000 fund investment will look like a $20,000 long, $20,000 short CQNS down, or a $15,000 long and $30,000 short SPY. These have different risk characteristics.
One big question for us today:
Should we keep our expected return on new money into the market (inclusive of dividends) at 8.5% for the upcoming year?
The model is most sensitive to this assumption.
Higher expected returns would mean fewer positions held, or great concentration into a few high-BETA names.
Lower expected returns makes us look like a trust department of a stodgy asset manager with widow and orphan money, with significantly more stocks, more diversification, and little concentration risk.
At the extreme, an expected return of 90% (we ran this as a test), suggests we hold one stock, and fully hedge it.
We will keep our expected return at 8.5% for today, but are considering whether to raise the expected return as the market continues to fall further (say to 3,500). An investor would need to be compensated more for new money investments, which coincides with lower stock prices. We would reflect those increased expectations (inverse to price / value).
GLTA
The SPY has fallen to under 3,600 on Friday.
Sept 29: Few changes today (RED pre-market)
Good morning.
How did we do yesterday? It was a very GREEN day with stocks rising. Our long investments were up 4.8%, and our short hedges were either down 1.8% or 3.9%. Put into dollars, our longs were up $501 and our hedges were either down $435 or $401. So, we made money yesterday, but not as much as on other 'flat' days. There is less alpha when all stocks are rising.
So, where are we today? Few changes to report.
1. We are DCA'ing one stock higher, to about 10% above all the other stocks. It had fallen about 10% below.
2. We are adding a tiny amount to our SPY hedges (so we are shorting a bit more SPY).
3. We are adding slightly to our CQNS Down Run hedges.
We made two compromises today that we want to share with you in order to save on transaction costs.
1. We picked the 7th best portfolio because it exactly matches what we are holding today. The difference between that and the best portfolio is to sell 'CG' and 'SYNA' and we would pick up ~8 ticks (or 8 x 10-5) in edge. If we sold those two stocks, our longs would be the smallest since we started the fund, and then we start to plan for increasing the size of all positions to account for fewer stocks. That is a big step for next week (when average position size would jump from ~$900 to ~$1,800).
2. The 3rd and 4th CQNS DOWN run stock picks are within 90 ticks of each other 9 x 10-4, and are adjacent in the list. The third stocks is 'SNAP' and the fourth is 'PTON'. We currently hold a short position on PTON. We are choosing to give up the 90 ticks and stay with our short position in PTON. These two stocks tend to flip positions, so we will save the buy and sell and hold the short in PTON.
A final note. It is interesting to see how our model worked in the first 'UP' day in 10. The longs did great, earning 4.8%. This is a significant gain in one day. However, our hedges kept pace and were also up 1.8% (not BETA weighted) and 3.9%. Net-Net, our model gained $8 in NAV and went down $208 in cash balance yesterday. We expect after today to see greater gains hit the balance sheet.
One final note. Once we are out of our bear market rally / bear market, we can consider changing our aggressive and protective hedging strategy. For now, it is driving the performance of the fund.
Thank you for following our journey. Feedback welcome.
Jeff
How did we do yesterday? It was a very GREEN day with stocks rising. Our long investments were up 4.8%, and our short hedges were either down 1.8% or 3.9%. Put into dollars, our longs were up $501 and our hedges were either down $435 or $401. So, we made money yesterday, but not as much as on other 'flat' days. There is less alpha when all stocks are rising.
So, where are we today? Few changes to report.
1. We are DCA'ing one stock higher, to about 10% above all the other stocks. It had fallen about 10% below.
2. We are adding a tiny amount to our SPY hedges (so we are shorting a bit more SPY).
3. We are adding slightly to our CQNS Down Run hedges.
We made two compromises today that we want to share with you in order to save on transaction costs.
1. We picked the 7th best portfolio because it exactly matches what we are holding today. The difference between that and the best portfolio is to sell 'CG' and 'SYNA' and we would pick up ~8 ticks (or 8 x 10-5) in edge. If we sold those two stocks, our longs would be the smallest since we started the fund, and then we start to plan for increasing the size of all positions to account for fewer stocks. That is a big step for next week (when average position size would jump from ~$900 to ~$1,800).
2. The 3rd and 4th CQNS DOWN run stock picks are within 90 ticks of each other 9 x 10-4, and are adjacent in the list. The third stocks is 'SNAP' and the fourth is 'PTON'. We currently hold a short position on PTON. We are choosing to give up the 90 ticks and stay with our short position in PTON. These two stocks tend to flip positions, so we will save the buy and sell and hold the short in PTON.
A final note. It is interesting to see how our model worked in the first 'UP' day in 10. The longs did great, earning 4.8%. This is a significant gain in one day. However, our hedges kept pace and were also up 1.8% (not BETA weighted) and 3.9%. Net-Net, our model gained $8 in NAV and went down $208 in cash balance yesterday. We expect after today to see greater gains hit the balance sheet.
One final note. Once we are out of our bear market rally / bear market, we can consider changing our aggressive and protective hedging strategy. For now, it is driving the performance of the fund.
Thank you for following our journey. Feedback welcome.
Jeff
Sept 28: Risk on for a 3rd day in a row
"It isn't easy being green" Kermit the Frog
On Monday we adjusted our model to reflect a higher expected return in the market, along with a higher risk-free rate of return. They both seem to go hand-in-hand, but can be very stressful to trade. It is unnerving to buy a risky set of stocks, even when they are balanced and held together in a way that cuts the risk significantly. The market is going down...this is a bear market...and yet here we are holding a smaller set of stocks that have higher average BETA scores.
The risk-free rate went up 1/4 point and the expected return from risky assets went up 1 1/4 point. These are small moves, and caused a big change in our Chicago Quantum Net Score / Smart Volatility long positions. It did not significantly change our short positions (the CQNS Down Hedge) but it did have an effect.
So, what happened? Pretty much everything is up today (except one out of three of our CQNS Down stocks).
It looks like we may make a few dollars of profit today given where we are mid-morning.
Why the quote by Kermit?
1. It is very stressful to buy stocks that should move quickly higher or lower when markets are moving lower. It takes commitment and fortitude.
2. Market volatility is flattening out, albeit at a very high level. About two weeks ago when the selling really accelerated we noticed that stock prices are not moving as much.
3. Some stocks are seeing greater spreads when we go to trade them, and the liquidity in the books (the number of shares bid and asked) is lower. This means that moves can be more dramatic, either up or down.
4. Risk free interest rates in the US (US Treasuries) are falling today. We look at the calendar and see it is September 28, and only this week left in Q3. However, interest rates have been rising and this is negative for risky stocks. Another reason to worry, what if interest rates keep rising? If they fall, equities should rise.
5. There is concern that as equity valuations and bond valuations fall that something will break in our financial system, and this will cause a 'flash crash.' Our model trades at the open, and so would be left somewhat defenseless in the face of an intra-day collapse. Of course we watch the markets, but our model is a 'swing trade' model and not a 'day trade' model. We expect the hedges to cover intra-day movements but have little confidence in this protection.
6. When the full set of #fintwit posters we follow is screaming to sell or short, we are moving in the opposite direction and buying risk.
Now, there are some mitigating factors when we move to a riskier footing.
1. We are net sellers of long positions as we get more risky. Our model picks fewer stocks, and our position size stays the same. So, for example, today we raised cash by selling a full long position (AMZN) and bought a few small dollar cost average (DCA) positions to keep position sizes equivalent. We are also buying back / covering our hedges in an equivalent amount. This means our $50,000 fund has fewer dollars invested and more in cash.
2. As we increase our BETA, we increase our SPY BETA Weighted hedge proportionally. This largely eliminates the possibility of a large drawdown in our fund, but does limit upside to true 'over-performance' or alpha.
FINAL NOTES
We have been fighting to keep transaction costs low. We do this by looking at stocks with lower transaction costs to trade, and may hold certain stocks (buys or sells) until we see the model recommend them repeatedly, or we see the model abandon those stocks (not just remove them from the top 1 or 2 portfolios). This helps us reduce trading costs. We also increased the minimum equity floor of our stocks (to $1B) and increase the minimum number of shares traded every day for the past year (to 50k). This helps us to avoid becoming trapped in a low liquidity stock.
We continue to make improvements into our quantitative model. We have reduced the time it takes to run the model, while last night coaxing out a better solution. There is a bit of the nerd in this endeavor, and I realize that my wife and daughter don't really care about simulated annealer settings, or how we jump across the hilbert space to find potentially better stock portfolios. I find it fascinating. For example, we added in a way to allow our genetic algorithms to 'die out' due to stagnation in the portfolio 'gene pool' and not waste time with unnecessary generations of breeding.
In regards to our simulated annealer, we did 'goof up' a bit last night. We are changing the way we allow the model to pick a different solution that might not be better right then and there, but is close, almost as good, and can 'jump' to another solution. It is requiring a completely different way to mathematically calculate when and whether to jump to a sub-par solution to keep things fresh and moving. You see, otherwise, if we only jump when the solution is better, we can get stuck in a local minima. This happens more often than not, and means the solver ends up providing only 'pretty good' portfolios after costing us time (time is our cost function). If we can use our simulated annealer to jump strategically, we can potentially find really different solutions that can be better than those found by other solvers. This diversity in solvers is so important when picking stocks. It is like diversity in the human population...it makes things better.
However, we tightened up the jumping categories so much last night that the simulated annealer only jumped when temperatures were really hot (like 9 orders of magnitude hotter) and not very often. Back to the drawing board today.
Good luck in the markets today.
Jeffrey Cohen, Investment advisor representative
US Advanced Computing Infrastructure, Inc.
Chicago Quantum (SM)
On Monday we adjusted our model to reflect a higher expected return in the market, along with a higher risk-free rate of return. They both seem to go hand-in-hand, but can be very stressful to trade. It is unnerving to buy a risky set of stocks, even when they are balanced and held together in a way that cuts the risk significantly. The market is going down...this is a bear market...and yet here we are holding a smaller set of stocks that have higher average BETA scores.
The risk-free rate went up 1/4 point and the expected return from risky assets went up 1 1/4 point. These are small moves, and caused a big change in our Chicago Quantum Net Score / Smart Volatility long positions. It did not significantly change our short positions (the CQNS Down Hedge) but it did have an effect.
So, what happened? Pretty much everything is up today (except one out of three of our CQNS Down stocks).
It looks like we may make a few dollars of profit today given where we are mid-morning.
Why the quote by Kermit?
1. It is very stressful to buy stocks that should move quickly higher or lower when markets are moving lower. It takes commitment and fortitude.
2. Market volatility is flattening out, albeit at a very high level. About two weeks ago when the selling really accelerated we noticed that stock prices are not moving as much.
3. Some stocks are seeing greater spreads when we go to trade them, and the liquidity in the books (the number of shares bid and asked) is lower. This means that moves can be more dramatic, either up or down.
4. Risk free interest rates in the US (US Treasuries) are falling today. We look at the calendar and see it is September 28, and only this week left in Q3. However, interest rates have been rising and this is negative for risky stocks. Another reason to worry, what if interest rates keep rising? If they fall, equities should rise.
5. There is concern that as equity valuations and bond valuations fall that something will break in our financial system, and this will cause a 'flash crash.' Our model trades at the open, and so would be left somewhat defenseless in the face of an intra-day collapse. Of course we watch the markets, but our model is a 'swing trade' model and not a 'day trade' model. We expect the hedges to cover intra-day movements but have little confidence in this protection.
6. When the full set of #fintwit posters we follow is screaming to sell or short, we are moving in the opposite direction and buying risk.
Now, there are some mitigating factors when we move to a riskier footing.
1. We are net sellers of long positions as we get more risky. Our model picks fewer stocks, and our position size stays the same. So, for example, today we raised cash by selling a full long position (AMZN) and bought a few small dollar cost average (DCA) positions to keep position sizes equivalent. We are also buying back / covering our hedges in an equivalent amount. This means our $50,000 fund has fewer dollars invested and more in cash.
2. As we increase our BETA, we increase our SPY BETA Weighted hedge proportionally. This largely eliminates the possibility of a large drawdown in our fund, but does limit upside to true 'over-performance' or alpha.
FINAL NOTES
We have been fighting to keep transaction costs low. We do this by looking at stocks with lower transaction costs to trade, and may hold certain stocks (buys or sells) until we see the model recommend them repeatedly, or we see the model abandon those stocks (not just remove them from the top 1 or 2 portfolios). This helps us reduce trading costs. We also increased the minimum equity floor of our stocks (to $1B) and increase the minimum number of shares traded every day for the past year (to 50k). This helps us to avoid becoming trapped in a low liquidity stock.
We continue to make improvements into our quantitative model. We have reduced the time it takes to run the model, while last night coaxing out a better solution. There is a bit of the nerd in this endeavor, and I realize that my wife and daughter don't really care about simulated annealer settings, or how we jump across the hilbert space to find potentially better stock portfolios. I find it fascinating. For example, we added in a way to allow our genetic algorithms to 'die out' due to stagnation in the portfolio 'gene pool' and not waste time with unnecessary generations of breeding.
In regards to our simulated annealer, we did 'goof up' a bit last night. We are changing the way we allow the model to pick a different solution that might not be better right then and there, but is close, almost as good, and can 'jump' to another solution. It is requiring a completely different way to mathematically calculate when and whether to jump to a sub-par solution to keep things fresh and moving. You see, otherwise, if we only jump when the solution is better, we can get stuck in a local minima. This happens more often than not, and means the solver ends up providing only 'pretty good' portfolios after costing us time (time is our cost function). If we can use our simulated annealer to jump strategically, we can potentially find really different solutions that can be better than those found by other solvers. This diversity in solvers is so important when picking stocks. It is like diversity in the human population...it makes things better.
However, we tightened up the jumping categories so much last night that the simulated annealer only jumped when temperatures were really hot (like 9 orders of magnitude hotter) and not very often. Back to the drawing board today.
Good luck in the markets today.
Jeffrey Cohen, Investment advisor representative
US Advanced Computing Infrastructure, Inc.
Chicago Quantum (SM)
September 26: Risk on
On Sept 26 (after a weekend), we saw our model change gears significantly.
The model proposed to hold fewer stocks and to buy back a portion of our hedges (both the CQNS DOWN Run hedges and the short SPY positions). The BETA Weighted SPY hedge was reduced proportionately less than the long portfolio as the BETA weighting of the portfolio increased with the fewer, newer stocks.
We made the trades, and are harvesting the benefits less than an hour into the trading day. Let's see if it lasts.
Net-Net for today:
1. We sold off conservative long positions and bought fewer, higher risk positions.
2. We bought back a percentage of our CQNS Down Run positions to match the total invested in the long positions.
3. We reduced our SPY hedge to a point where it is BETA weighted (higher BETA) against a smaller long portfolio.
This consumed a significant amount of our cash balance, as the buybacks / covering of shorts was significant, and was larger than the money raised by reducing the overall size of our long positions. So, we have less cash and theoretically, less leverage.
Good luck today.
Jeffrey Cohen, Investment Advisor Representative
US Advanced Computing Infrastructure, Inc.
The model proposed to hold fewer stocks and to buy back a portion of our hedges (both the CQNS DOWN Run hedges and the short SPY positions). The BETA Weighted SPY hedge was reduced proportionately less than the long portfolio as the BETA weighting of the portfolio increased with the fewer, newer stocks.
We made the trades, and are harvesting the benefits less than an hour into the trading day. Let's see if it lasts.
Net-Net for today:
1. We sold off conservative long positions and bought fewer, higher risk positions.
2. We bought back a percentage of our CQNS Down Run positions to match the total invested in the long positions.
3. We reduced our SPY hedge to a point where it is BETA weighted (higher BETA) against a smaller long portfolio.
This consumed a significant amount of our cash balance, as the buybacks / covering of shorts was significant, and was larger than the money raised by reducing the overall size of our long positions. So, we have less cash and theoretically, less leverage.
Good luck today.
Jeffrey Cohen, Investment Advisor Representative
US Advanced Computing Infrastructure, Inc.
Sept 23: Model recommendations
By Jeffrey Cohen, Investment Advisor Representative and President
US Advanced Computing Infrastructure, Inc.
Today in the markets the main indices fell and just about every stock we track fell. This was a very down day.
US Advanced Computing Infrastructure, Inc.
Today in the markets the main indices fell and just about every stock we track fell. This was a very down day.
finviz.com image of today's performance of all the S&P 500 stocks (RED means decline)
We posted a tweet on what our model recommended today 6 minutes before the market opened.
Let's see how we held up in such a down market.
By the end of the morning trades, we found we were about even with cash, we spent about what we took in with additional shorts. We have full BETA-weighted SPY hedges and/or CQNS Down Run (1:1) short hedges in place against our ~20 long stocks, equally weighted (each around $850 apiece).
Let's see how we held up in such a down market.
By the end of the morning trades, we found we were about even with cash, we spent about what we took in with additional shorts. We have full BETA-weighted SPY hedges and/or CQNS Down Run (1:1) short hedges in place against our ~20 long stocks, equally weighted (each around $850 apiece).
Yahoo Finance summary view of stocks in our managed account (paper-traded). Top three are shorts, SPY is a short hedge, and the rest are longs. This is as of Friday, Sept 23, at 16:07 ET.
What we see is that our CQNS Down run stocks (the top 3) were two down and one up. So, this hedge was profitable. The SPY was down so that hedge was profitable. Since we BETA weight the hedge, and BETA is > 1.0, we hedged more than the 1.6% to offset losses in our long picks.
The long picks were decidedly negative, but interestingly enough the medium return was about 2% (at a very quick glance), so it probably was offset by the SPY hedge, and we probably made money hedging against the CQNS down run. All in all, a good performance during a terrible day in the market.
In a few minutes, we will finish calculating our NAV and cash position for our paper-traded fund.
This weekend we will execute a few work steps:
- make our run faster (eliminate a few bells and whistles that support fundamental analysis)
- look at raising our expected returns in the market. As prices drop, expected returns rise.
- look at raising our risk-free rate of return for investors. As interest rates rise, investors can safely earn more outside the market for risky assets.
The long picks were decidedly negative, but interestingly enough the medium return was about 2% (at a very quick glance), so it probably was offset by the SPY hedge, and we probably made money hedging against the CQNS down run. All in all, a good performance during a terrible day in the market.
In a few minutes, we will finish calculating our NAV and cash position for our paper-traded fund.
This weekend we will execute a few work steps:
- make our run faster (eliminate a few bells and whistles that support fundamental analysis)
- look at raising our expected returns in the market. As prices drop, expected returns rise.
- look at raising our risk-free rate of return for investors. As interest rates rise, investors can safely earn more outside the market for risky assets.
Sept 21 FOMC day: Today our model earned its stripes.
It creates a disciplined approach to managing money and risk.
It drove up NAV in our fund by ~25% over the past week.
It has consistently performed depending on the hedging strategy chosen.
We did a video about the model (a deep dive) today on Youtube, and will put the video link below. Check it out.
We will do another video tomorrow and show/explain how the fund did today.
If only our personal account would do this well. We have been blood RED all week while the model consistently either made money, or had modest losses which were all recovered today.
GLTA
It drove up NAV in our fund by ~25% over the past week.
It has consistently performed depending on the hedging strategy chosen.
We did a video about the model (a deep dive) today on Youtube, and will put the video link below. Check it out.
We will do another video tomorrow and show/explain how the fund did today.
If only our personal account would do this well. We have been blood RED all week while the model consistently either made money, or had modest losses which were all recovered today.
GLTA
Sept 16: We made more progress.
Yesterday we reviewed two weeks of trades and positions from our fund.
A few things things stand out.
1. A hedge based on the BETA weight of the stocks we hold during each trading day does better than a 1:1 hedge, and intuitively this makes sense during a time when the market is falling. There are multiple down days where the hedge is what 'paid the bills.'
2. We see that stocks move significantly when the market is not open. Therefore, calculations of Net Asset Value (NAV) or performance must account for this. We cannot do 'daily' performance and expect the numbers to add up.
3. Our model is very sensitive to investor expectations of future returns and risk-free rates of return. Our model has been surprisingly consistent in the stocks that it picks when those expectations are consistent from day to day. There are changes each day, but maybe only 5% or 10% of the positions have any 'adjustments' required each morning.
4. We trade in the morning at market open. We could still decide to trade during the day if we want to try to gain edge from trading. At this point, all trades are made in the first few minutes of the day.
5. We are calculating our Net Asset Value based on a notional $50,000 investment. That money is used to buy and hold the long positions, and to offset the short 'hedge' positions. There is a cash balance to cover changes each day, such as adding stocks (which uses cash) or buying back hedges (covering uses cash). We have not accounted for interest rates on our savings, but that could add a percent or two on cash. Let's see what our 'future' custodial partner, Charles Schwab, offers in their RIA in a box capability.
6. This will be an everyday job. If we decide to go 'discretionary' and take in investor money, we will be hiring, finding office space, and making this a larger operation. Today, we are a one-man shop and I realize that setting up and managing a fund, along with marketing and customer operations, is significant work that will justify our future fees.
Good luck to all in the markets today.
Jeffrey Cohen, Investment Advisor Representative and President,
US Advanced Computing Infrastructure, Inc.
A few things things stand out.
1. A hedge based on the BETA weight of the stocks we hold during each trading day does better than a 1:1 hedge, and intuitively this makes sense during a time when the market is falling. There are multiple down days where the hedge is what 'paid the bills.'
2. We see that stocks move significantly when the market is not open. Therefore, calculations of Net Asset Value (NAV) or performance must account for this. We cannot do 'daily' performance and expect the numbers to add up.
3. Our model is very sensitive to investor expectations of future returns and risk-free rates of return. Our model has been surprisingly consistent in the stocks that it picks when those expectations are consistent from day to day. There are changes each day, but maybe only 5% or 10% of the positions have any 'adjustments' required each morning.
4. We trade in the morning at market open. We could still decide to trade during the day if we want to try to gain edge from trading. At this point, all trades are made in the first few minutes of the day.
5. We are calculating our Net Asset Value based on a notional $50,000 investment. That money is used to buy and hold the long positions, and to offset the short 'hedge' positions. There is a cash balance to cover changes each day, such as adding stocks (which uses cash) or buying back hedges (covering uses cash). We have not accounted for interest rates on our savings, but that could add a percent or two on cash. Let's see what our 'future' custodial partner, Charles Schwab, offers in their RIA in a box capability.
6. This will be an everyday job. If we decide to go 'discretionary' and take in investor money, we will be hiring, finding office space, and making this a larger operation. Today, we are a one-man shop and I realize that setting up and managing a fund, along with marketing and customer operations, is significant work that will justify our future fees.
Good luck to all in the markets today.
Jeffrey Cohen, Investment Advisor Representative and President,
US Advanced Computing Infrastructure, Inc.
Sept 16: It is a red day in the futures market today, and our model just increased the number of stocks to hold from 19 to 22. That means we put more money to work in the market, and increase our shorts (on the S&P) to offset it, at the open. I wonder how many other money managers have the same mandate.
We will take a look at the new stocks and decide whether to add them, or to wait a day...
We will take a look at the new stocks and decide whether to add them, or to wait a day...
Sept 15: Hedging strategies are just as important as what to hold long
Sept 15, 2022
We have been running three different hedging strategies. The first is a straight hedge, dollar for dollar, against the S&P 500 index. This makes sense on one level, where the stocks rise in a rising market faster than the S&P 500 due to their BETA (if it is greater than 1.0) and they rise more slowly than the S&P 500 if the BETA is less than 1.0.
We have a second hedging strategy where we BETA weight the S&P 500 hedge. This means that we calculate the number of shares required to hedge 1:1, then we multiply that amount of shares by the BETA of the long stock portfolio. Therefore, if we have a BETA of 1.5, we hold 50% more hedge.
The third method is to short a few key stocks from the CQNS DOWN run. This would be a 1:1 weighting of dollars of x number of UP run stocks, and a maximum of 3 DOWN run stocks. On the first day of this approach, all three CQNS DOWN stocks went up, so our hedge was badly performing.
We will keep you informed. The model portfolio seems to be working.
However, our personal portfolio is not working well for us. We may move our investable capital to this model in the near future. The cobbler needs shoes.
We have been running three different hedging strategies. The first is a straight hedge, dollar for dollar, against the S&P 500 index. This makes sense on one level, where the stocks rise in a rising market faster than the S&P 500 due to their BETA (if it is greater than 1.0) and they rise more slowly than the S&P 500 if the BETA is less than 1.0.
We have a second hedging strategy where we BETA weight the S&P 500 hedge. This means that we calculate the number of shares required to hedge 1:1, then we multiply that amount of shares by the BETA of the long stock portfolio. Therefore, if we have a BETA of 1.5, we hold 50% more hedge.
The third method is to short a few key stocks from the CQNS DOWN run. This would be a 1:1 weighting of dollars of x number of UP run stocks, and a maximum of 3 DOWN run stocks. On the first day of this approach, all three CQNS DOWN stocks went up, so our hedge was badly performing.
We will keep you informed. The model portfolio seems to be working.
However, our personal portfolio is not working well for us. We may move our investable capital to this model in the near future. The cobbler needs shoes.
Sept 14: A choice after a very down day yesterday (US equities fell ~4%)
Good morning. On Sept 13, 2022, we had a significant fall in the values of US growth stocks. Overnight, Asia fell as well. This is based on fears that continuing inflation is expected to cause long-maturity interest rates to rise. This reduces the Net Present Value of growth company expected cash flows, and makes their stocks and fixed income decline in fundamental value.
The most interesting thing is that the moves happened very quickly, as if a few very large traders decided to sell off the market at precisely the same time. So, another way to describe the fall is to say a few money managers / trading houses decided to go RED on the market and sell it down, just for fun, profit, kicks, momentum, or because they had bought a great deal of puts (or were short a significant amount of money in US equities). We have seen this before, and it does not surprise us. It is interesting that it was done 3-days before single stock options expiration on Friday, Sept 16, 2022. The timing makes those options (calls worth less, puts worth more) change in value, which creates massive trading opportunities for risk-takers.
So, where does this leave us and our paper-trading model based on the Chicago Quantum Net Score UP model?
1. We had a very sizeable hedge on our portfolio, equal to the S&P 500 ETF x BETA weight of the portfolio as of the market close the night before. So, if the market fell 4%, we earned ~7%. Then, if we look at the stocks we were holding, most of them were down. However, if the average decline was less than 7%, we actually made money. Big smile...
2. We did try something (hypothetically) yesterday before the fall. We took three of our worst performers in the CQNS DOWN run and paper-traded a hedge based on those. We avoid pharmaceutical stocks because those are trading and moving based on 'insider' and non-correlated news, and are rightfully unpredictable. Those three stocks did almost all their movements pre-market. So, by the time we would be ready to buy or sell them (*at the market open at 0930ET), the moves already happened. Now, we are holding those positions for day 2, and need to adjust those positions slightly at the open. Do they behave better as hedges than the SPY? Let's find out today. Net-Net: we learned that first day hedges using stocks (and potentially indices) do not work, as much of the change can happen pre-market. The hedges are more effective on an ongoing basis to capture over-night and pre-market action.
3. We have a choice on whether to buy up our down positions, or to sell down our up positions this morning. The market fell yesterday and a few stocks fell enough to require additional purchases to keep all equities at the 'roughly' equivalent value at market open. If we buy more, we are using our cash, and will require additional hedging (more shorts of the SPY) as well. These are small adjustments, but the bigger question is important.
" Do you dollar cost average down to ensure equivalent valuations of all stocks, or do you sell strength to match decliners?"
Both seem like odd things to do. We would be buying the worst performing stocks yesterday (that were selected by our model both days), or we would be selling the best performing stocks yesterday. We have to do something, as the stocks need an equal valuation.
The other part of this decision is that if we buy more, we hedge more (short more SPY at the open) and if we sell down, we hedge less (and buy back a little SPY).
In both cases, we end up fully hedged, but in one case we raise a little cash (after a down day we sell) and in another case we spend a little cash and reduce reserves.
We will let you know what we decide.
Good luck to all.
The most interesting thing is that the moves happened very quickly, as if a few very large traders decided to sell off the market at precisely the same time. So, another way to describe the fall is to say a few money managers / trading houses decided to go RED on the market and sell it down, just for fun, profit, kicks, momentum, or because they had bought a great deal of puts (or were short a significant amount of money in US equities). We have seen this before, and it does not surprise us. It is interesting that it was done 3-days before single stock options expiration on Friday, Sept 16, 2022. The timing makes those options (calls worth less, puts worth more) change in value, which creates massive trading opportunities for risk-takers.
So, where does this leave us and our paper-trading model based on the Chicago Quantum Net Score UP model?
1. We had a very sizeable hedge on our portfolio, equal to the S&P 500 ETF x BETA weight of the portfolio as of the market close the night before. So, if the market fell 4%, we earned ~7%. Then, if we look at the stocks we were holding, most of them were down. However, if the average decline was less than 7%, we actually made money. Big smile...
2. We did try something (hypothetically) yesterday before the fall. We took three of our worst performers in the CQNS DOWN run and paper-traded a hedge based on those. We avoid pharmaceutical stocks because those are trading and moving based on 'insider' and non-correlated news, and are rightfully unpredictable. Those three stocks did almost all their movements pre-market. So, by the time we would be ready to buy or sell them (*at the market open at 0930ET), the moves already happened. Now, we are holding those positions for day 2, and need to adjust those positions slightly at the open. Do they behave better as hedges than the SPY? Let's find out today. Net-Net: we learned that first day hedges using stocks (and potentially indices) do not work, as much of the change can happen pre-market. The hedges are more effective on an ongoing basis to capture over-night and pre-market action.
3. We have a choice on whether to buy up our down positions, or to sell down our up positions this morning. The market fell yesterday and a few stocks fell enough to require additional purchases to keep all equities at the 'roughly' equivalent value at market open. If we buy more, we are using our cash, and will require additional hedging (more shorts of the SPY) as well. These are small adjustments, but the bigger question is important.
" Do you dollar cost average down to ensure equivalent valuations of all stocks, or do you sell strength to match decliners?"
Both seem like odd things to do. We would be buying the worst performing stocks yesterday (that were selected by our model both days), or we would be selling the best performing stocks yesterday. We have to do something, as the stocks need an equal valuation.
The other part of this decision is that if we buy more, we hedge more (short more SPY at the open) and if we sell down, we hedge less (and buy back a little SPY).
In both cases, we end up fully hedged, but in one case we raise a little cash (after a down day we sell) and in another case we spend a little cash and reduce reserves.
We will let you know what we decide.
Good luck to all.
Sept 12: A calm day for the model
On Friday, Sept 9, we executed many transactions in the first 4 minutes of trading (market open + 4 minutes), and we were in a very good position to take advantage of the day's advances. Of course, we are hedged, so we only 'bank' the difference between the stocks increasing and the hedge of the SPY.
Heading into Monday, the model picked almost exactly the same portfolio. All systems go. Here is the interesting point, we will be adding 1 stock (which adds ~ $1,000 to our long position as each stock is invested the same amount), and shorting 5 shares of SPY, which equals $2,000 in short exposure. It would not surprise me if the rise in the market on Friday causes a slight tremor at the open while new and larger hedges are laid. However, our model was only down slightly in its BETA weighting, from ~1.9 to ~1.8.
So, Monday will be a very transaction efficient day.
We did tweak our model again to increase the stringency of the data validation steps. We now have a minimum market capitalization, a 'higher' floor on trading volumes, and seem to be tightening things up nicely to ensure that we can enter and exit positions efficiently.
One insight from Friday morning: We have a few stocks that have spreads while the market is open of $1.00 or higher, and more that have spreads of $0.10 to $1.00. If we can stick to stocks with tighter spreads (say $0.01 or $0.02), then we have effectively made our model transaction-cost free. We are not there yet, but keep learning.
The other insight is that we are making our trades automatically at market open. Not right at open, but starting about 15 seconds into the trading day and finishing within the first 10 minutes of trading. Friday, we were done with our 'paper trades' in our brokerage account within the first 4.5 minutes of the trading day. We are not trying to 'penny' the stocks, and enter a market order.
It looks like Friday was a very good day for our model. Let's see what Monday has in store.
GLTA.
Heading into Monday, the model picked almost exactly the same portfolio. All systems go. Here is the interesting point, we will be adding 1 stock (which adds ~ $1,000 to our long position as each stock is invested the same amount), and shorting 5 shares of SPY, which equals $2,000 in short exposure. It would not surprise me if the rise in the market on Friday causes a slight tremor at the open while new and larger hedges are laid. However, our model was only down slightly in its BETA weighting, from ~1.9 to ~1.8.
So, Monday will be a very transaction efficient day.
We did tweak our model again to increase the stringency of the data validation steps. We now have a minimum market capitalization, a 'higher' floor on trading volumes, and seem to be tightening things up nicely to ensure that we can enter and exit positions efficiently.
One insight from Friday morning: We have a few stocks that have spreads while the market is open of $1.00 or higher, and more that have spreads of $0.10 to $1.00. If we can stick to stocks with tighter spreads (say $0.01 or $0.02), then we have effectively made our model transaction-cost free. We are not there yet, but keep learning.
The other insight is that we are making our trades automatically at market open. Not right at open, but starting about 15 seconds into the trading day and finishing within the first 10 minutes of trading. Friday, we were done with our 'paper trades' in our brokerage account within the first 4.5 minutes of the trading day. We are not trying to 'penny' the stocks, and enter a market order.
It looks like Friday was a very good day for our model. Let's see what Monday has in store.
GLTA.
Sept 9: Day 7 we tweaked again to fix the issue of higher transaction costs
Good morning to our faithful readers. Thank you for being there. Once we are live with our fund, things will be different and the 'sampo' from Finnish folklore will be operational.
Last night we were thinking about the fact that DDS (Dillards Department Store) fell out then back into our model. It had a $6.00 spread when we went to trade it, and a few minutes later (when we did trade it) the spread was down to ~$3.00. This seems high in absolute terms, and made us think about spreads and stocks and trading.
We need to stay in stocks that have liquidity and can be traded.
A few small changes (raising the return to newly invested risk assets by 0.25%) and increasing trading volumes slightly, along with another day in the market, has turned over 17 stocks, and we hold 11 stock for another day. This seems like alot of churn in our portfolio, and is not sustainable for the longer-term. It is a good thing we are starting with paper-trading.
What I am thinking is that when we slightly increase the risk characteristics of the 'buyer' or 'investor' in the fund by 0.25%, which does not seem like alot, it caused a move into riskier assets.
The BETA of the overall portfolio is increased from 1.473 to 1.834 because the model added in SI and SYNA, which are very high BETA stocks, and removes some of the lower BETA names. We will be hedging out that extra BETA exposure.
Gotta go, markets open in 10 minutes and have to make the trades in real-time.
Good luck to all.
Jeffrey Cohen, Investment Advisor Representative
US Advanced Computing Infrastructure, Inc.
Chicago Quantum (SM)
Last night we were thinking about the fact that DDS (Dillards Department Store) fell out then back into our model. It had a $6.00 spread when we went to trade it, and a few minutes later (when we did trade it) the spread was down to ~$3.00. This seems high in absolute terms, and made us think about spreads and stocks and trading.
We need to stay in stocks that have liquidity and can be traded.
A few small changes (raising the return to newly invested risk assets by 0.25%) and increasing trading volumes slightly, along with another day in the market, has turned over 17 stocks, and we hold 11 stock for another day. This seems like alot of churn in our portfolio, and is not sustainable for the longer-term. It is a good thing we are starting with paper-trading.
What I am thinking is that when we slightly increase the risk characteristics of the 'buyer' or 'investor' in the fund by 0.25%, which does not seem like alot, it caused a move into riskier assets.
The BETA of the overall portfolio is increased from 1.473 to 1.834 because the model added in SI and SYNA, which are very high BETA stocks, and removes some of the lower BETA names. We will be hedging out that extra BETA exposure.
Gotta go, markets open in 10 minutes and have to make the trades in real-time.
Good luck to all.
Jeffrey Cohen, Investment Advisor Representative
US Advanced Computing Infrastructure, Inc.
Chicago Quantum (SM)
Update 15 minutes into the trading day. We made our 17 trades this morning. We did these manually, and realize these all need to be entered before market open, to trade at the open. It is inefficient to make these trades manually, and we can make errors, or try to penny or nickel a trade and possibly lose the early-morning movement up or down. So, lesson learned, enter trades early pre-market for market open and let them execute at market. We run a risk of up or down movements, and of large spreads, but manually trading them won't make it systematically better.
Our stocks had spreads that ranged from one cent (APPL) to $3.54 (DDS). 8: (1c to 10c), 7: (11c to 100c), 2: (103,354c).
Those larger spreads, and the small spreads that are large percentages, are very expensive trades for us. When we work through the model options in the morning and decide which 'best' portfolio to select, we should take into account spreads of individual stocks. The problem is that the higher BETA stocks (which our model likes better when expected returns rise) have higher spreads. They also have more expensive options.
This is another insight. Higher BETA stocks have higher spreads, and cost more to enter a position.
In general, we saw what felt like shallower top of book positions today. Fewer bids and asks, fewer shares offered and bid, and spreads seemed higher than they would be if the stocks were actively trading.
That cannot be good for the markets on this Summer Friday. Prices are rising, but on lower volume in individual names. Well, let's see how the market performs today.
Our stocks had spreads that ranged from one cent (APPL) to $3.54 (DDS). 8: (1c to 10c), 7: (11c to 100c), 2: (103,354c).
Those larger spreads, and the small spreads that are large percentages, are very expensive trades for us. When we work through the model options in the morning and decide which 'best' portfolio to select, we should take into account spreads of individual stocks. The problem is that the higher BETA stocks (which our model likes better when expected returns rise) have higher spreads. They also have more expensive options.
This is another insight. Higher BETA stocks have higher spreads, and cost more to enter a position.
In general, we saw what felt like shallower top of book positions today. Fewer bids and asks, fewer shares offered and bid, and spreads seemed higher than they would be if the stocks were actively trading.
That cannot be good for the markets on this Summer Friday. Prices are rising, but on lower volume in individual names. Well, let's see how the market performs today.
Update: 42 minutes into the trading day.
We are down to 17 stocks vs. the SPY (hedged short).
All stocks chosen by the model are up, as is the SPY. The SPY is up 1.16%, and the BETA of the stocks chosen are 1.835. The question is whether the 17 stocks will run faster than their BETA score today while the markets look strong.
We know these stocks will outperform a straight non-BETA adjusted hedge (largely due to their >1.0 BETA on an up-day), but will they outperform SPY x BETA of the portfolio? This becomes the big question of our hedging strategy. To beta hedge or not?
We are down to 17 stocks vs. the SPY (hedged short).
All stocks chosen by the model are up, as is the SPY. The SPY is up 1.16%, and the BETA of the stocks chosen are 1.835. The question is whether the 17 stocks will run faster than their BETA score today while the markets look strong.
We know these stocks will outperform a straight non-BETA adjusted hedge (largely due to their >1.0 BETA on an up-day), but will they outperform SPY x BETA of the portfolio? This becomes the big question of our hedging strategy. To beta hedge or not?
Sept 8: Day 6 Update
Good morning all. Thank you for reading and following our blog. We are learning so much about paper trading and setting up a fund based on the Chicago Quantum Net Score (CQNS) UP run. Think I need a bigger hat this morning.
1. The model switched gears on us last night. It noticed something in the market on our first 'UP' or 'green' day in a while. It picked different stocks. We have 9 sells this morning, 6 buys, and we need to cover 47 of 159 short shares of SPY, or 30% of our hedge. So, an up day causes us to sell some higher BETA stocks (model locks in profits?), buy some big names with lower BETA values (ready to run?), and reduce our short position / hedge.
2. There are 'too many?' transactions this morning. We have 16 transactions to accomplish before the market gets underway. This must be what the 'buy on market open' trade type is for. This could impact our cost efficiency. The question for us is whether we should be watching the tape to see if stocks are rising before we sell them to lock in a few extra pennies. Also, depending on how the SPY / S&P 500 Index performs this morning, that is a very large transaction for us (equal to the change in hedge).
Open question for us is the order of the trades, timing: whether to enter them for market open (and take our chances), and whether we add value by trying to trade better than the market (let stocks run before selling, let stocks fall before buying). It seems that this would be a dangerous exercise, and could eliminate the edge of the model. It might work well on some days, but then work badly on others. The model does not take into account transaction costs and execution quality.
Going back to it, we have a very active morning ahead of us. 9 sells, 6 buys, and covering 30% of our short hedge. If other fund managers are doing this, we are in for a wild, and bullish open.
We still have not figured out how we will value / track our performance. For now, we are keeping track of buys and sells, and need to put this together.
As always, collaborations are welcome.
Good luck to all today in the markets.
Jeffrey Cohen, Investment Advisor Representative
US Advanced Computing Infrastructure, Inc.
As of September 8, 2022, we are paper trading this model.
We are a non-discretionary investment advisor registered in Illinois.
note: we did not tighten data validation, but we did scrub the code to make it run faster. It could be that our 'efficiency' and cleaner code caused the model to pick different stocks, despite coming up with an equivalent CQNS score and number of stocks.
1. The model switched gears on us last night. It noticed something in the market on our first 'UP' or 'green' day in a while. It picked different stocks. We have 9 sells this morning, 6 buys, and we need to cover 47 of 159 short shares of SPY, or 30% of our hedge. So, an up day causes us to sell some higher BETA stocks (model locks in profits?), buy some big names with lower BETA values (ready to run?), and reduce our short position / hedge.
2. There are 'too many?' transactions this morning. We have 16 transactions to accomplish before the market gets underway. This must be what the 'buy on market open' trade type is for. This could impact our cost efficiency. The question for us is whether we should be watching the tape to see if stocks are rising before we sell them to lock in a few extra pennies. Also, depending on how the SPY / S&P 500 Index performs this morning, that is a very large transaction for us (equal to the change in hedge).
Open question for us is the order of the trades, timing: whether to enter them for market open (and take our chances), and whether we add value by trying to trade better than the market (let stocks run before selling, let stocks fall before buying). It seems that this would be a dangerous exercise, and could eliminate the edge of the model. It might work well on some days, but then work badly on others. The model does not take into account transaction costs and execution quality.
Going back to it, we have a very active morning ahead of us. 9 sells, 6 buys, and covering 30% of our short hedge. If other fund managers are doing this, we are in for a wild, and bullish open.
We still have not figured out how we will value / track our performance. For now, we are keeping track of buys and sells, and need to put this together.
As always, collaborations are welcome.
Good luck to all today in the markets.
Jeffrey Cohen, Investment Advisor Representative
US Advanced Computing Infrastructure, Inc.
As of September 8, 2022, we are paper trading this model.
We are a non-discretionary investment advisor registered in Illinois.
note: we did not tighten data validation, but we did scrub the code to make it run faster. It could be that our 'efficiency' and cleaner code caused the model to pick different stocks, despite coming up with an equivalent CQNS score and number of stocks.
Update before executing trades (0920ET). Most of the buys and sells are lower this morning in pre-market. We seem to have caught a tailwind on our purchases and covering our hedge.
Sept 7: Day 5 Update
Good morning. It is 0817 ET and we are working through the daily changes to our fund's holdings. We are still in 'paper trading' mode and learning what it takes to run the Chicago Quantum Net Score effectively, and how to convert that model output into a workable managed account / hedge fund.
Today there are four buys and two sells. There are a total of 25 stocks required to give us the best edge over the S&P 500 Index. That edge is small mathematically, but we expect will still generate a profit.
This fund is hedged against the S&P 500 Index by using short positions on the S&P 500 Index ETF. Therefore, this model should generate a profit regardless of market direction.
What we learned today:
- Continuing to make data validation 'tightening' moves will make it easier to trade, but it does cause transaction costs. We are adding back constraints on stocks, and making previous constraints more 'logical.' For example, previously we had allowed stocks in that had a data error for net income (e.g., does not file a 10-Q as a small bank, or just a data error from our MDS provider). Now, we exclude stocks with net income data errors unless their Cash Flow From Operations is positive, then we keep them.
- Just about every stock is down today from when we bought it. This is a relentlessly lower market that is impacting most large, familiar stocks. Even the blue chips are down a buck or two from yesterday. At the money ATM call options are likely losing value while puts may not be gaining enough from relatively small moves. This could be part of the process, draining call premium.
- Our model portfolio has a lower BETA value today, down to 1.6x vs. the S&P 500 Index. It was almost 1.7x yesterday at the open.
- The model is picking more familiar names, and those names are falling (only two stocks are up since we bought them) in this declining market.
- Our model is moving more conservative due to the core assumptions on the risk free rate of return (higher), the expected return of the market (lower) and correlations / variance changes. We notice that the market has higher volatility and variance of stock price movements, and this seems to lower current stock prices while raising future expected returns.
- When we manage money in a managed account / hedge fund, we have to trade every day despite the sense that the market is only moving in one direction. There is no choice or 'YOLO' movements. We follow the hedged model and collect the alpha.
- It goes without saying, but here it is. The model fund survives and thrives because we are hedging out SPY risk. The gains on the shorts of the market offset the losses on ~90% of the holdings of the model.
- We need to finalize on our data validation steps and the 'tightness' of our validation. Each time we adjust the model assumptions or parameters, it changes the stocks we are to hold. This causes transaction costs from trading.
- We ran the model 2x longer last night to find an even larger edge. The edge is smaller on an absolute basis than yesterday, but we have more confidence we found the best portfolio we would (and we found many that were almost as good). The model picked mostly the same stocks (buy 4 and sell 3 out of 25 stocks). Of that, two stocks were removed due to tighter data validation, which demonstrates the 'cost' of tightening data quality.
The question for us is when to buy and when to sell this morning. Yesterday, we did our adjustments in the first 10 minutes of the day when the market was trading higher, and we likely overpaid for buys and sold higher for sells / shorts. We could have waited until the market fell (mid-day), but then the day's moves are already over. We believe we should buy and sell at the open.
Finally, we are going to work on how we measure success. We have long and short positions, but not sure how we can code whether we made money each day. Working on this today.
We may also run a quantum computing run against our best portfolios to see how we are doing, and if the quantum computer would do something different. We use the D-Wave Systems quantum annealing system. We use Intrinio market data (powered by Intrinio).
Good luck to all!
Today there are four buys and two sells. There are a total of 25 stocks required to give us the best edge over the S&P 500 Index. That edge is small mathematically, but we expect will still generate a profit.
This fund is hedged against the S&P 500 Index by using short positions on the S&P 500 Index ETF. Therefore, this model should generate a profit regardless of market direction.
What we learned today:
- Continuing to make data validation 'tightening' moves will make it easier to trade, but it does cause transaction costs. We are adding back constraints on stocks, and making previous constraints more 'logical.' For example, previously we had allowed stocks in that had a data error for net income (e.g., does not file a 10-Q as a small bank, or just a data error from our MDS provider). Now, we exclude stocks with net income data errors unless their Cash Flow From Operations is positive, then we keep them.
- Just about every stock is down today from when we bought it. This is a relentlessly lower market that is impacting most large, familiar stocks. Even the blue chips are down a buck or two from yesterday. At the money ATM call options are likely losing value while puts may not be gaining enough from relatively small moves. This could be part of the process, draining call premium.
- Our model portfolio has a lower BETA value today, down to 1.6x vs. the S&P 500 Index. It was almost 1.7x yesterday at the open.
- The model is picking more familiar names, and those names are falling (only two stocks are up since we bought them) in this declining market.
- Our model is moving more conservative due to the core assumptions on the risk free rate of return (higher), the expected return of the market (lower) and correlations / variance changes. We notice that the market has higher volatility and variance of stock price movements, and this seems to lower current stock prices while raising future expected returns.
- When we manage money in a managed account / hedge fund, we have to trade every day despite the sense that the market is only moving in one direction. There is no choice or 'YOLO' movements. We follow the hedged model and collect the alpha.
- It goes without saying, but here it is. The model fund survives and thrives because we are hedging out SPY risk. The gains on the shorts of the market offset the losses on ~90% of the holdings of the model.
- We need to finalize on our data validation steps and the 'tightness' of our validation. Each time we adjust the model assumptions or parameters, it changes the stocks we are to hold. This causes transaction costs from trading.
- We ran the model 2x longer last night to find an even larger edge. The edge is smaller on an absolute basis than yesterday, but we have more confidence we found the best portfolio we would (and we found many that were almost as good). The model picked mostly the same stocks (buy 4 and sell 3 out of 25 stocks). Of that, two stocks were removed due to tighter data validation, which demonstrates the 'cost' of tightening data quality.
The question for us is when to buy and when to sell this morning. Yesterday, we did our adjustments in the first 10 minutes of the day when the market was trading higher, and we likely overpaid for buys and sold higher for sells / shorts. We could have waited until the market fell (mid-day), but then the day's moves are already over. We believe we should buy and sell at the open.
Finally, we are going to work on how we measure success. We have long and short positions, but not sure how we can code whether we made money each day. Working on this today.
We may also run a quantum computing run against our best portfolios to see how we are doing, and if the quantum computer would do something different. We use the D-Wave Systems quantum annealing system. We use Intrinio market data (powered by Intrinio).
Good luck to all!
Sept 6 Update of our paper-traded fund / managed account
We have started to optimize our model to make it easier to know how we must update our fund each day. Trades to buy, to sell and updated to our hedging position. Learning a great deal by doing this each morning.
Are we making money with this hedged Chicago Quantum Net Score model fund? Not sure yet...time will tell. It is interesting to be investing in a long / hedged fund during a committed market bear leg lower.
Are we making money with this hedged Chicago Quantum Net Score model fund? Not sure yet...time will tell. It is interesting to be investing in a long / hedged fund during a committed market bear leg lower.
August 31 Market Test
We will test our best CQNS UP Run portfolio in what are considered extremely bearish market conditions. Let's see how we do as we build our model portfolio with proper hedging strategies and the use of volatility bets (options).
The start is the optimized CQNS UP portfolio.
We are building our scalable investment portfolio and learning a great deal on what to do and how to do it.
We have made two videos to share our progress. We will let you know as things change and progress is made.
This is a full time job from very early in the morning (say 7am ET through 11am ET) and most likely through the trading day. I am learning that each day, and the tricks of timing, trading, transaction costs, and how a small change to our quantitative model can result in significant changes in holdings.
The start is the optimized CQNS UP portfolio.
We are building our scalable investment portfolio and learning a great deal on what to do and how to do it.
We have made two videos to share our progress. We will let you know as things change and progress is made.
This is a full time job from very early in the morning (say 7am ET through 11am ET) and most likely through the trading day. I am learning that each day, and the tricks of timing, trading, transaction costs, and how a small change to our quantitative model can result in significant changes in holdings.
As a note, this is still an 'academic' exercise for clients. We are currently a non-discretionary investment advisor, which means we can sell the data and analysis, but not handle client funds in a managed investment account. Once we perfect our fund, then we have a few steps to complete, and approvals to gain, before we can offer this to the public. Also, it will probably only be available to accredited investors, high-net-worth individuals, portfolio managers, wholesale traders, and financial industry professionals who can understand and handle the risks.
We will let everyone know when we have worked out the kinks in the fund, and we would be ready to move forward. It is still a question of if, and not when. Let's see if we can consistently beat the S&P 500 during down and up markets.
Good luck to all.
We will let everyone know when we have worked out the kinks in the fund, and we would be ready to move forward. It is still a question of if, and not when. Let's see if we can consistently beat the S&P 500 during down and up markets.
Good luck to all.
Update: Sep. 2 (23 stocks from 5)
Our model suggested we broaden our stock to include in our Chicago Quantum Net Score portfolio to 23 stocks, up from 5 stocks earlier this week. We changed two input assumptions that primarily changed the sensitivity of the model.
Of the 23 stocks we are holding today, 21 are trading higher. This is amazing progress and will drive portfolio performance today. The SPY is up 1.19% today, so the question is whether the 23 stocks will over-perform that gain (which we hedge away during this bear market).
Of the 23 stocks we are holding today, 21 are trading higher. This is amazing progress and will drive portfolio performance today. The SPY is up 1.19% today, so the question is whether the 23 stocks will over-perform that gain (which we hedge away during this bear market).
