Stats

A news scan found the following text: QTR9090200 is a test of a parent trend with a target of 1.5 and a cost of 16 basis points. The test has two main parts. First, eight fixed exchange-traded fund sleeves were started with 1250 USD each, and cash was reinvested independently in each sleeve. The initial weights of the sleeves were equal, but they were allowed to drift over time without being rebalanced monthly. A new position was entered at the next session's open if the previous month's return was positive, with no short selling or leverage used. The parent trend was held until the last session of the month, with no stop or target, and was liquidated at the end of the month. Second, the test was trained on data from 2017 to 2020, with a holdout period from 2021 to 2025, and an extension from January to August 2026. A warm-up period was used in 2016, with no invented early history. Fees were 1 basis point plus 15 basis points for slippage on each side of a trade. The results of the test, which are simulated portfolio returns and not actual account returns, were as follows: in 2021, the return was -2.0080833581230917 percent, in 2022 it was -2.270405986327506 percent, in 2023 it was 0.41805183322796147 percent, in 2024 it was 2.3875767421257965 percent, and in 2025 it was 11.776261979038939 percent. The test used daily open, high, low, and close prices, and did not account for 15-minute venue exits, gap overshoot, or national best bid and offer prices. The test also used fractional sizing and did not include fixed business costs. The exchange-traded fund adjustment used in the test means that all returns are proxy data, and not excess returns from futures, inverse-volatility scale, or original long-short paper replication. The era used in the test was reused from prior research and is not untouched prospective evidence. The conclusion is that the parent trend is diagnostic only and cannot transfer its edge to different exits. The sources of the data used in the test are https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all and https://w4.stern.nyu.edu/facdir/lpederse/papers/TimeSeriesMomentum.pdf, and the test was run using QTR9090200 research/trend.py and archived results.

This is a test of a trading strategy called QTR9090200. The strategy has the following rules: it starts with eight fixed exchange-traded fund, or ETF, investments, each worth 1250 USD, and it reinvests any cash it receives. The strategy does not use leverage, or borrowing money to invest, and it does not sell stocks short, or bet that their prices will fall. It buys stocks when the previous month's return is positive and holds them until the end of the month. The strategy was tested using data from 2017 to 2026, with some years used for training and others used to see how the strategy would perform in real life. The strategy charges a fee of 1 basis point, or 0.01 percent, plus an additional 5 basis points for buying and selling stocks. The results of the test are as follows: in 2021, the strategy lost 0.51 percent; in 2022, it lost 1.25 percent; in 2023, it gained 1.61 percent; in 2024, it gained 4.48 percent; and in 2025, it gained 14.30 percent. These results are based on simulated trades, not actual investments, and they do not include any partial months. The data used for the test came from several sources, including a website and a research paper. The strategy is only for testing and cannot be used to make actual investment decisions.

The QTR9090200 train managed a trend with a target of 1.5 and a cost of 16 basis points. The strategy started with eight fixed ETF sleeves, each with an initial investment of 1250 USD, and reinvested cash independently without monthly rebalancing. The strategy entered a position at the next session's open if the prior 252-session close return was positive, without using shorts or leverage. A stop-loss was set at 1.5 times the prior 20-day mean high-low range, with a target fill and no reentry in the same month. The strategy was trained from 2017 to 2020, with a holdout period from 2021 to 2025, and an extension from January to August 2026. Fees were 1 basis point plus 15 basis points for slippage on each side. The results for the years 2017, 2018, 2019, and 2020 were -2.8149%, -3.6408%, -1.5477%, and 0.5002%, respectively. These results are simulated portfolio returns, not actual account returns. A news scan found the source https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all and https://w4.stern.nyu.edu/facdir/lpederse/papers/TimeSeriesMomentum.pdf, which are related to the QTR9090200 research and trend analysis.

The QTR9090200 train managed a trend with a target of 1.5 and a cost of 6.000000000000001 basis points. The strategy started with eight fixed exchange-traded fund sleeves, each with an initial investment of 1250 USD, and reinvested cash independently. The weights of the sleeves were equal at the start, but were allowed to drift over time without monthly rebalancing. A position was entered at the next session's open if the previous month's return was positive, with no short selling or leverage used. The strategy used a stop-loss at the prior 20-day mean high-low, with a target of 1.5 times the risk, and a breakeven point at the prior high plus 1 times the risk. The strategy also used an adverse stop priority and gap at open, with a conservative target fill. There was no re-entry in the same month, and the position was liquidated at the end of the month. The training period was from 2017 to 2020, with a chronological holdout period from 2021 to 2025, and an extension from January to August 2026. The warm-up period was in 2016, with no invented early history. The fees were 1 basis point plus 5 basis points for slippage on each side. The results were as follows: in 2017, the return was -1.4841720267347425, in 2018 it was -2.2140289894281184, in 2019 it was -0.0939076822628171, and in 2020 it was 2.293375215453519. These results are simulated portfolio returns, not actual account returns. The data used was from the sources https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all and https://w4.stern.nyu.edu/facdir/lpederse/papers/TimeSeriesMomentum.pdf, and the research was conducted using the QTR9090200 research/trend.py and archived results.

A news scan found the headline about QTR9090200 train parent trend 252 target 1.5 cost 16bp. The details are as follows: the fund has eight fixed exchange-traded fund, or ETF, sleeves that start with 1250 USD each and independently reinvest cash, with equal initial weights that can drift over time without monthly rebalancing. The fund enters a position at the next session's open if the prior completed month's 252-session close return is positive, without using shorts or leverage. It holds the position through the last session of the month without a managed stop or target, and does not reenter the same month, with liquidation at the end of the month. The training period is from 2017 to 2020, with a chronological holdout from 2021 to 2025, and an extension from January to August 2026, with a warmup in 2016 and no invented early history. The fees are 1 basis point, or bp, plus 15 bp for slippage on each side. The results for the years 2017 to 2020 are 4.8517, -6.9643, 4.8927, and -1.2946, respectively, which are simulated portfolio returns and not actual account returns. The data source is https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all and https://w4.stern.nyu.edu/facdir/lpederse/papers/TimeSeriesMomentum.pdf, with code and archived results in QTR9090200 research/trend.py.

This is a description of a trading strategy. The strategy has eight parts, called ETF sleeves, which start with $1250 each and reinvest any cash they make. They do not rebalance monthly, meaning their weights can change over time. The strategy buys when the previous month's return is positive and sells at the end of the month. It does not use leverage or short selling. A trading period, or trend, is considered long-only, meaning it only buys, not sells. The strategy was trained on data from 2017 to 2020, tested on data from 2021 to 2025, and further tested from January to August 2026. It includes fees of 1 basis point plus slippage of 5 basis points for each buy and sell. The results for each year are as follows: 2017 had a return of 6.355536569709552%, 2018 had a return of -5.562579921964328%, 2019 had a return of 6.404240580754461%, and 2020 had a return of 0.42316346016491035%. These results are simulated and do not reflect actual account performance. The data used is proxy data and does not include some costs. The strategy is for diagnostic purposes only and its success cannot be transferred to different trading exits. The data sources include https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all and https://w4.stern.nyu.edu/facdir/lpederse/papers/TimeSeriesMomentum.pdf.

A diagnostic test was executed on September 9 at 02:04 using the node /app/agent/dist/backtest.js program with the symbols SEI, LEU, and HUT, and a breakout strategy. The test looked back 20 periods over 730 days, with a stop loss of 3 percent and a target of 6 percent. The data came from https://runway.adityaperswal.workers.dev/data/stock-bars. The results for HUT were: 501 bars of data, a return of -38.57 percent, a buy and hold return of 899.20 percent, a maximum drawdown of 48.15 percent, a win rate of 23.21 percent, and an average trade return of -0.80 percent, based on a sample of 56 trades. This was a same-close IEX diagnostic test without considering the cost of holding, and the strategy was not validated. Separate tests, coded independently, were run from 1312 to 1322 to define different exits, but this does not claim to replicate the results.

A trading test was done on September 9 at 02:04 using a computer program called backtest.js. The test looked at the stocks SEI, LEU, and HUT using a strategy called breakout, which means buying when a stock's price breaks out of a certain range. The test looked back 20 days and considered 730 days of trading. It used a stop loss of 3 percent and a target profit of 6 percent. The data came from https://runway.adityaperswal.workers.dev/data/stock-bars. The results for LEU were: 501 trades, a return of 16.72 percent, a buy and hold return of 418.07 percent, a maximum loss of 26.54 percent, a win rate of 39.58 percent, and an average trade return of 0.40 percent, based on 48 samples. This test did not consider costs and did not validate the strategy. Another test, called next-open20/10, was done on different days, from 1312 to 1322, and found different results when the trades were closed, but this is not a replication of the same test.

A diagnostic test was executed on September 9 at 02:04 using the node /app/agent/dist/backtest.js program with the symbols SEI, LEU, and HUT, and a breakout strategy. The test looked back 20 days over a period of 730 days, with a stop loss of 3 percent and a target gain of 6 percent. The data came from https://runway.adityaperswal.workers.dev/data/stock-bars. The results for SEI showed 501 bars, a return of -16.316623579462195 percent, and a buy and hold return of 465.4288240495137 percent. The maximum drawdown was 24.67796041608561 percent, the win rate was 30 percent, and the average trade return was -0.37902855640591604 percent, based on a sample of 40 trades. This test did not include costs or a holdout period, and the strategy has not been validated. A separate test, coded independently, will run from September 20 to September 22 and will define different exit strategies, but this is not a claim of replication.

A trading strategy for LIN was tested using historical data from 2021 to September 8, 2026, with adjustments made for historical SIP. The strategy was based on a preregistered note that specified the conditions for entering and exiting a trade. The results showed negative returns for all parameters tested in the holdout period, with returns ranging from -4.056846414245119% to -10.004885943988047%. The mean confidence interval also contained zero, which means the results were not statistically significant. Due to these negative results and the small sample size, the strategy was not approved for deployment. Additionally, the lack of a stop or target limit for the rejected rule raises concerns about its soundness. The conclusion is that there will be no entry into a trade based on this strategy.

A trading strategy for OESX was tested using historical data from 2021 to September 8, 2026, with adjustments made for historical stock prices. The strategy was based on a preregistered note that considered the close price being higher than the prior 20-day high, with entry at the next open and exit 10 sessions later. The test used training data before 2026 and a holdout period starting in 2026. The results showed negative returns and high drawdowns, with a win rate of 30.30% and an average win percentage of 11.46%. The test also considered fees and slippage of 5 basis points and 20 basis points, respectively. The conclusion was that the strategy did not qualify for entry, due to negative returns and a mean confidence interval that contained zero. The small sample size, potential regime change, and current listing bias also limited the soundness of the results. Additionally, the lack of a stop or target proposal for the rejected rule was noted. A news scan found the results: a set of data including return, drawdown, win rate, and other metrics for different training and holdout periods.

A trading strategy for the stock ALM was tested. The strategy was based on two premises. First, the stock's historical data from 2021 to September 8, 2026, was used, with adjustments made for historical stock splits and dividends. The data was obtained from https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=ALM&timeframe=1Day&feed=sip&adjustment=all. Second, a preregistered note, number 1291, specified the rules for the strategy: if the stock's closing price was higher than the highest closing price of the previous 20 days, then the next day's opening price would be used to enter a trade, and the trade would be exited 10 sessions later. The strategy was trained on data before 2026 and tested on data from 2026. The results showed that the strategy had a return of 76.73% in training, but only 35.24% in the holdout period. The strategy also had a high drawdown of 29.35% in training and 34.44% in the holdout period. The win rate was 100% in training, but only 85.71% in the holdout period. The mean percentage return was 20.98% in training, but only 5.90% in the holdout period. The average win percentage was 20.98% in training, but only 11.67% in the holdout period. The average loss percentage was 0% in training, but 28.73% in the holdout period. The quarter Kelly criterion, a measure of risk-adjusted return, was 0 in training, but 0.1263 in the holdout period. The mean confidence interval percentage was between -7.82% and 16.67% in the holdout period. The simulation showed a 7.69% probability of loss and a median return of 11.78% in the holdout period. The 95th percentile drawdown was 9.14% in the holdout period. The indicators used in the strategy included the date, close price, prior 20-day high, 50-day simple moving average, average daily range, and realized volatility. The conclusion was that the strategy did not qualify for use because it had negative returns or a mean confidence interval that contained zero. The small sample size, potential regime change, and current listing bias also sounded a warning about the strategy's soundness. The positive Kelly criterion was not enough to approve the strategy for deployment. No stop or target was proposed for the rejected rule.

A trading strategy for LPTH was tested using historical data from 2021 to September 8, 2026. The strategy was based on a simple rule: buy when the close price is higher than the highest close price in the prior 20 days, and sell 10 sessions later. The test results showed negative returns and a high drawdown, which means the strategy lost a significant amount of money. The results were: a return of -31.45% to -49.83% in the training period, and -44.43% to -47.03% in the holdout period. The win rate was 38.46% in the training period and 33.33% in the holdout period. The strategy failed to meet the qualification criteria due to negative returns and a mean confidence interval that contained zero. The small sample size, potential regime change, and current listing bias also raise concerns about the soundness of the results. Therefore, no entry was made, and no stop or target was proposed for the rejected rule.

A trading rule for VIAV was tested. The rule was based on two premises. First, historical data from 2021 to September 8, 2026, was used, which was adjusted for historical SIP, or Security Information Processor, bars. The data came from https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=VIAV&timeframe=1Day&feed=sip&adjustment=all. Second, a preregistered note, number 1291, stated that if the close was higher than the prior 20-day high, the next open would be an entry point, and the exit would be 10 sessions later. The training data was from before 2026, and the holdout data, which is data not used in training, was from 2026, with a flat start. The test included fees of 5 basis points and slippage of 20 basis points on each side, with a total sensitivity of 10/50 basis points. The results were as follows: for a training set with a parameter of 0.001, the return was -21.56101728254379%, the drawdown was 43.87644573960039%, and the win rate was 54.54545454545454%. For a parameter of 0.0025, the return was -31.260719711185136%, the drawdown was 47.566238926226866%, and the win rate was 45.45454545454545%. For a parameter of 0.005, the return was -44.8355699210074%, the drawdown was 55.98419651480479%, and the win rate was 43.18181818181818%. In the holdout set, for a parameter of 0.001, the return was 39.57439740577789%, the drawdown was 31.49954356238619%, and the win rate was 75%. For a parameter of 0.0025, the return was 36.26448391220976%, the drawdown was 31.909316470071737%, and the win rate was 75%. For a parameter of 0.005, the return was 30.921419031791288%, the drawdown was 32.58683756305105%, and the win rate was 75%. The conclusion is that there will be no entry, as the returns were negative, or the mean confidence interval contained zero, which fails the test. No parameter was chosen from the holdout set. The small sample size, potential regime change, and current listing bias, as well as the missing 15-minute and overnight stop execution limit, affect the soundness of the results. A positive Kelly point is not enough to approve deployment. No stop or target was proposed for the rejected rule.

A news scan found the headline 'FCX frozen next-open breakout fails qualification'. Premises for this conclusion are: 1. Historical data for FCX from 2021 to September 8, 2026, was used, with bars adjusted for historical Systematic Internalizer Prices, or SIP, which is a type of price adjustment. The data source is https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=FCX&timeframe=1Day&feed=sip&adjustment=all. 2. A preregistered note, number 1291, stated that if the close is greater than the prior 20-day high, then enter at the next open, and exit 10 sessions later. This was trained on data before 2026 and tested on chronological 2026 data with a flat start. Vectorbt fees of 5 basis points and slippage of 20 basis points were applied to each side, with a total sensitivity of 10/50 basis points. 3. The results were: for a training set with a parameter of 0.001, the return was -16.445877910652634, the drawdown was 55.37675849852162, and the win rate was 47.61904761904761. For a parameter of 0.0025, the return was -26.33746169870869, the drawdown was 57.975418192935116, and the win rate was 47.61904761904761. For a parameter of 0.005, the return was -40.290445910765285, the drawdown was 63.87768145526771, and the win rate was 47.61904761904761. For the holdout set with a parameter of 0.001, the return was 27.88240266538614, the drawdown was 23.47485665424227, and the win rate was 57.14285714285714. For a parameter of 0.0025, the return was 25.22486932262869, the drawdown was 23.704088465794026, and the win rate was 57.14285714285714. For a parameter of 0.005, the return was 20.917765161436925, the drawdown was 24.08462011911663, and the win rate was 57.14285714285714. The conclusion is no entry, as negative returns or mean confidence interval containing zero fail, and no parameter was chosen from the holdout set. The small selected samples, regime change, current listing bias, and missing 15-minute and overnight stop execution limit soundness are concerns. A positive Kelly point is not approval for deployment. No stop or target was proposed for the rejected rule.

The trading strategy for CW failed to qualify. The strategy was based on two premises: first, using historical data from 2021 to September 8, 2026, with adjustments made for historical SIP, and second, a preregistered note that specified the conditions for entry and exit. The results showed negative returns or mean confidence interval containing zero, which failed the qualification. The small sample size, potential regime change, and current listing bias also raised concerns about the soundness of the strategy. Additionally, the lack of a 15-minute or overnight stop execution limit was a limitation. Since no parameter was chosen from the holdout, no entry was made. The positive Kelly point was not enough to approve deployment, and no stop or target was proposed for the rejected rule.

A trading strategy for HUT was tested. The strategy was based on two premises: 1. It used historical data from 2021 to September 8, 2026, which was adjusted for certain market factors, and 2. it was based on a preregistered note that specified the conditions for entering and exiting a trade. The results of the test were as follows: the strategy had a return of 250.58% in training, with a drawdown of 58.39% and a win rate of 56.41%, but in the holdout period, the return was 13.96% with a drawdown of 33.46% and a win rate of 50%. The strategy also had a quarter Kelly criterion of 0.0659 in training and 0.0539 in the holdout period. However, the conclusion is that there will be no entry into a trade because the returns were negative or the mean confidence interval contained zero, and no parameter was chosen from the holdout period. The small sample size, potential regime change, and current listing bias, as well as the lack of a 15-minute or overnight stop execution limit, also raise concerns about the soundness of the strategy. A positive Kelly criterion is not enough to approve deployment, and no stop or target is proposed for the rejected rule.

A trading rule for CCJ was tested. The rule was based on two premises: 1. Historical data from 2021 to September 8, 2026, was used, which was adjusted for certain market effects, and 2. a specific trading strategy was applied, where the stock's closing price had to be higher than its highest price in the last 20 days to enter a trade, and the trade would be closed 10 sessions later. The results showed negative returns in the training period, with a return of -33.31% for one set of parameters and -43.28% for another, and a drawdown of 67.90% and 71.23% respectively. In the holdout period, the returns were 12.09%, 10.76%, and 8.56% for different parameters, but the sample size was small. The trading rule also had a high drawdown and a low win rate. The conclusion is that the trading rule does not qualify for use because it had negative returns or mean confidence interval containing zero, and no parameter was chosen from the holdout period. Additionally, the small sample size, potential regime change, and missing execution limits raise concerns about the soundness of the results. A positive Kelly criterion is not enough to approve the rule for deployment, and no stop or target is proposed for the rejected rule.

A trading strategy for LEU was tested based on two premises. First, historical data from 2021 to September 8, 2026, was used, with adjustments made for historical SIP, sourced from https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LEU&timeframe=1Day&feed=sip&adjustment=all. Second, a preregistered note, number 1291, specified that the strategy would enter when the close was higher than the prior 20-day high, and exit 10 sessions later, with training done before 2026 and a holdout period starting from 2026. The strategy also accounted for fees and slippage. The results showed returns of 413.28, 349.81, and 260.98 for different training parameters, but negative returns of -32.42, -33.63, and -35.59 in the holdout period. The conclusion is that there will be no entry, as the negative returns and other factors, such as small sample sizes and potential biases, do not support the strategy's deployment.

A trading strategy for SEI was tested using historical data from 2021 to September 8, 2026. The strategy was based on a simple rule: if the closing price is higher than the highest closing price in the last 20 days, then enter a trade at the next opening price and exit 10 sessions later. The results showed that the strategy did not perform well, with negative returns in the holdout period. The mean return was -47.99657328093197% for a parameter of 0.001, -48.924262336320574% for 0.0025, and -50.43379499256859% for 0.005. The win rate was also low, at 16.666666666666664%. The quarter Kelly criterion, a measure of risk-adjusted return, was 0, indicating that the strategy did not provide a positive return. Due to the poor performance, no entry was recommended. The small sample size, potential regime change, and current listing bias also raised concerns about the soundness of the results. Additionally, the lack of a 15-minute or overnight stop execution limit and the absence of a proposed stop or target for the rejected rule were noted.

A preregistered note, number 1255, was written before retrieving prices. This note is about a trading strategy, which is a type of automated trading plan, called a breakout family, that uses the next open price to enter a trade. The strategy uses a fixed signal, which is a trigger to buy or sell, based on the average high and low prices over 20 sessions, and a stop loss, which is a price level at which to automatically sell, set at 1 times the average daily range, and a target price, which is a price level at which to automatically sell, set at 2 times the average daily range. The strategy also has a rule to exit the trade at the next open price after 5 completed sessions. This strategy only buys, does not sell short, and does not allow overlap or re-entry on the same day. If there is a large gap in price at the open, the strategy uses a dual touch stop, which is a stop loss that is triggered by two price levels. The strategy also accounts for fees, which are costs of trading, of 5 basis points, and slippage, which is the difference between the expected price and the actual price, of 10 basis points on each side of the trade. The strategy uses a system called vectorbt to generate explicit orders, which are instructions to buy or sell, at the open, touch, and close prices, and manual capital returns, which are the returns on investment, are reconciled exactly after correcting the vectorbt trade return denominator, which excludes the entry fee. The strategy was tested on 958 valid adjusted price bars from November 10, 2022, to September 8, 2026, with no missing dates for the ASPI index compared to the SPY index. The first SIP end at the current 15-minute window at 4:03 was corrected to 23:40 UTC and succeeded without changing the daily window. The strategy was trained through 2024 and resulted in a return of -51.7294% with 46 trades. The historical holdout period from 2025 to September 8, 2026, resulted in a return of -47.6615% with a drawdown, which is the peak-to-trough decline, of 61.0281% and 24 trades, with 5 wins and 19 losses. The average win was 16.4566% and the average loss was 7.1155%, with a mean return of -2.20465% and a quarter Kelly, which is a measure of risk, of 0. The double cost return was -51.2974%. Four training-only stop and target variants were tested, but all resulted in negative returns. The default strategy was retained and no further tuning was done. A command-line interface, or CLI, was required to execute a separate baseline, which resulted in a return of -28.2872% with 35 trades and a drawdown of 46.2699%. The CLI lacks next open, cost, and 15-minute controls, and is not eligible for trading. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars. There are limits to the strategy, including that the historical viewed data is not pristine prospective, cost assumptions are not empirical spreads, and OHLC touches are not actual 15-minute or overnight fills. The drawdown is marked at sampled prices, not all intraday lows, and all-adjusted corporate actions are not raw cash dividend models. No trade will be made and no further tuning will be done, as a richer narrative cannot fix a failed rule.

A preregistered note, number 1253, was written before data was retrieved. The test used nine sector ETFs, ranking them by their adjusted close returns over the prior 10 sessions at the weekly boundary. It bought the worst performer at the next open if its return was negative. Subtracting SPY, which is a type of stock market index fund, did not change the rankings. The test used a stop-loss, which is a rule to sell a security when it reaches a certain price, and a target price, which is a desired sale price. The stop-loss was set at 1 times the average high-low range, and the target price was set at 2 times the average high-low range. The test was trained on data from 2016 to 2020 and tested on data from 2021 to 2025, with a retrospective extension to 2026. Late trades that crossed the weekly boundary were omitted. The test used a full normalized capital of $10,000, with cash only, and fees of 1 basis point, which is a unit of measurement for interest rates, and slippage of 15 basis points. The results were verified by a vectorbt, which is a type of backtesting tool, and the final result was $6,092.56. The test used a stop priority on daily collisions, which is when two or more trades occur at the same time, and an adverse gap stop at the open, which is when the price gaps down at the open. The results showed a sample size of 222, with a p-value of 0.5495, a win rate of 28.85%, and a loss rate of 20.25%. The mean return was -0.20%. The test also showed yearly returns of -1.72% in 2021, -14.36% in 2022, -19.44% in 2023, -7.18% in 2024, and -3.19% in 2025. The exit counts were 112 for stop, 42 for time, 35 for breakeven, 26 for target, and 7 for target gap. The return and drawdown, which is the peak-to-trough decline in value, are based on full research capital and daily close marks, not account performance or worst intraday drawdown. No fixed monthly platform charges were deducted, which would have worsened the net economics. The sources used for this test include a working paper from the National Bureau of Economic Research and a stock bars feed from a website. The observations used were from 16049 to 16076, and the local research file was reversal.py.

A preregistered note, number 1253, was written before data was retrieved. The test involved nine sector ETFs, which were ranked based on their adjusted close returns over the prior 10 sessions at the weekly boundary. The ETF with the worst return was bought at the next open, but only if its own return was negative. The SPY subtraction, a way to compare the ETFs, did not change the cross-sectional ranks. The prior 20 session mean high-low range, a measure of price movement, was used to set the initial stop loss at 1R, the target at 2R, and the remaining stop entry at 1R, with a maximum next weekly open. The test was trained on data from 2016 to 2020, tested on data from 2021 to 2025, and retrospectively extended to 2026. Late boundary-crossing trades were omitted. The test used all 10 series, with 2685 aligned sessions from January 2016 to September 8, 2026, and no missing value fill, adjusted OHLC, and a survived ETF set. The test started with a full normalized capital of 10000, with cash only, and fees of 1bp and slippage of 5bp each side, including partials. A custom event cash and every equity point was verified by vectorbt from orders, with an exact final value of 9497.888005990975. The stop priority was set on daily collisions, with an adverse gap stop at open, target gap recognized, and partial/BE ambiguous days assigned adverse sequence. The test used continuous touch, which is optimistic versus 15-minute polling, even though the collision rule is conservative. The test did not exactly replicate the Nagel replication, which used 48 industry/individual stocks, long-short, 5 daily sleeves, conditional beta hedge, closing liquidity supplier, and VIX dependence. The ETF cash adaptation tests a different payoff. The premises of the test were: 1, recorded returns from a fixed rule; 2, base positive edge must survive costs and uncertainty; and 3, the whole family fails stress and CI. The conclusion is that this is research-only, with no allocation. The sample size was 222, with a p-value of 0.5495495495495496, a W of 1.4912659127355696%, an L of 1.8291627507384938%, and a mean of -0.004422674414909376%. The returns by year were: 2021, 6.896128739344354; 2022, -6.859962703773414; 2023, -11.855230711768904; 2024, 1.7638338800762154; and 2025, 6.350283913282184. The exit counts were: stop, 112; time, 42; breakeven, 35; target, 26; and target gap, 7. The trade sample counts were completed round trips, with partial legs aggregated. The return and drawdown are full research capital and daily close marks, not account performance or worst intraday DD. No fixed monthly platform charges were deducted, which would worsen the net economics. All historical windows were previously used by other fund studies, not pristine OOS, and no winner was chosen from a sweep. The sources used were https://www.nber.org/system/files/working_papers/w17653/w17653.pdf and https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all, with observations 16049-16076, and local research/reversal.py.

A preregistered note, number 1253, was written before data was retrieved. The test used nine sector ETFs, ranking them by their adjusted close returns over the prior 10 sessions at the weekly boundary. It bought the worst performing ETF at the next open if its return was negative. The SPY subtraction did not change the cross-sectional ranks. The prior 20 session mean high-low range was used to set the initial stop, target, and other parameters. The test was trained on data from 2016 to 2020, tested on data from 2021 to 2025, and retrospectively extended to 2026. Late boundary-crossing trades were omitted. The test used all 10 series with 2685 aligned sessions from January 2016 to September 8, 2026, with no missing value fill and adjusted OHLC. The full normalized capital was $10,000, with cash only, and fees of 1 basis point and slippage of 15 basis points each side, including partials. The custom event cash and every equity point were verified by vectorbt from orders, with an exact final value of $6230.33. The stop priority was on daily collisions, with adverse gap stop at open, target gap recognized, and partial or breakeven ambiguous days assigned adverse sequence. The test concluded that the premises were research-only and not for allocation, with a sample size of 222, a p-value of 0.5495, a win rate of 1.3058%, a loss rate of 2.0253%, and a mean return of -0.1947%. The years 2021 to 2025 had returns of -6.5788%, -11.4885%, -16.6819%, -9.8307%, and 0.2925%, respectively. The exit counts were 104 for stop, 45 for target, 35 for breakeven, 29 for time, and 9 for target gap. The return and drawdown were calculated using full research capital and daily close marks, not account performance or worst intraday drawdown. No fixed monthly platform charges were deducted, which would have worsened the net economics. The sources used were https://www.nber.org/system/files/working_papers/w17653/w17653.pdf and https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all, with observations 16049 to 16076, and local research using reversal.py.

A preregistered note, number 1253, was written before data was retrieved. The test used nine sector ETFs, ranking them by their adjusted close returns over the prior 10 sessions at the weekly boundary. It bought the worst performer at the next open if its return was negative. Subtracting SPY, which is a type of stock market index fund, did not change the rankings. The initial stop loss was set at 1 times the prior 20 session mean high-low range, the target at 1.5 times this range, and half of the position was closed at 1 times this range with the remaining position stopped at the next weekly open. The test was trained on data from 2016 to 2020, tested on data from 2021 to 2025, and retrospectively extended to 2026. Late boundary-crossing trades were omitted. All 10 series had 2685 aligned sessions from January 2016 to September 8, 2026, with no missing values filled, adjusted open, high, low, and close prices, and the ETF set survived. The test started with a normalized capital of 10000, used cash only, and included fees of 1 basis point and slippage of 5 basis points on each side, including partials. A basis point is a unit of measurement for interest rates and fees. The test results were verified by vectorbt from orders, with an exact final value of 9712.657588258156. The test prioritized stops on daily collisions, used an adverse gap stop at open, and recognized target gaps. The test was not an exact replication of the Nagel study, which used 48 industry and individual stocks, a long-short strategy, 5 daily sleeves, a conditional beta hedge, and closing liquidity supplier, and was dependent on the VIX, a measure of stock market volatility. The ETF cash adaptation tested a different payoff. The premises of the test were: 1, recorded returns from a fixed rule; 2, the base positive edge must survive costs and uncertainty; and 3, the whole family fails stress and confidence interval tests. The conclusion is that this is research only and not for allocation. The sample size was 222, with a p-value of 0.5495495495495496, a win rate of 1.5086271781915401%, and a loss rate of 1.8291627507384902%. The mean return was 0.0051182011960310084%. The returns by year were: 2021, 1.607574849725868; 2022, -3.732368419421772; 2023, -8.835527034948976; 2024, -1.1415502499449115; and 2025, 10.177340209378416. The exit counts were: stop, 104; target, 45; breakeven, 35; time, 29; and target gap, 9. The return and drawdown are for the full research capital and daily close marks, not account performance or worst intraday drawdown. No fixed monthly platform charges were deducted, which would have worsened the net economics. All historical windows were previously used by other fund studies, not pristine out-of-sample, and no winner was chosen from a sweep. The sources used were https://www.nber.org/system/files/working_papers/w17653/w17653.pdf and https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all, with observations 16049-16076, and the local research file reversal.py.