Stats

A news scan found the headline 'QTM90800 SPY delay later: frozen turn-of-month test'. This refers to a test of an automated trading strategy. The test, called Protocol829, was stopped before it could be reviewed again. It looked at 59 groups of market years from 2021 to 2025. These years were already checked for other rules, so they are not completely independent. The strategy buys at the opening of the last monthly trading session and sells at the opening of the fourth trading session of the next month. This timing is different from the original strategy, which was based on closing prices. There are no stop-loss or target prices. The test used a long-only, cash-only strategy, with a 5 basis point fee and 5 basis point slippage. The cash balance was updated independently and matched the final equity within 1e-5 USD on an initial cash balance of 10,000 USD. The test only counted complete trades, not partial ones. The drawdown included both open and close prices, as well as some synthetic marks, but not all price movements within a trading session. The test used historical adjusted ETF prices, which are approximate. A 10 basis point audit and 2/20 basis point sensitivity test were saved but not selected. The test did not verify 15-minute or extended-hours trading, whole-share profit management, or exact-time execution. Recent positive results cannot override the training data or later gates, and there is no current calendar entry. The sources for this test are a research paper from Purdue University and a dataset from Aditya Perswal. All inputs, code, and results were archived in 291 parts and verified. This is an adaptation of an ETF strategy, not an exact replication of the CRSP strategy, and it has not been approved for deployment.

A news scan found the text describing the QTM90800 SPY delay train test, which is a frozen turn-of-month test. The test has three main points: first, the Protocol829 was frozen before a return inspection, and from 2016 to 2020, 58 completed cohorts were trained after a warm-up period, with cross-split turns purged. The strategy is to buy at the open of the last monthly session and sell at the open of the fourth next-month session. The timing adaptation is different from the original close-to-close paper and does not include a stop or target. Second, the test uses a long-only cash-only vectorbt order ledger with a 5 basis point fee and 5 basis point slippage per fill. The cash compounding is independent and reconciles the final equity within 1e-5 USD on an initial cash amount of 10,000 USD, with nonnegative cash. The sample count includes full entry cohorts, not partial fills. The drawdown includes open and close, as well as synthetic threshold marks, but not all intrabar excursions. Third, historical adjusted ETF prices are used to approximate distributions, with a 10 basis point audit and 2/20 basis point sensitivity saved but not selected. The test does not verify 15-minute or extended-hours fills, whole-share profit management, or exact time execution. Recent positive samples cannot override the training or later gates, and there is no current calendar entry. The sources of the information are https://business.purdue.edu/faculty/mcconnell/publications/Equity-Returns-at-the-Turn-of-the-Month.pdf and https://runway.adityaperswal.workers.dev/data/stock-bars. All inputs, code, and results are archived in 291 parts and have exact readback verified. This is an ETF adaptation, not an exact CRSP replication, and it does not have deployable approval.

A news scan found the headline 'QTM90800 SPY parent recent: frozen turn-of-month test'. This refers to a test of the SPY fund, which is an exchange-traded fund, or ETF, that tracks the S&P 500 stock market index. The test has three parts. First, the test looks at the returns of the SPY fund from 2026 through September 4, using what is called a turn-of-month strategy, which means buying at the end of one month and selling at the beginning of the next month. The test uses a method called close-to-close, which means it looks at the prices at the close of each day, and it does not use stop-loss or target prices, which are prices at which to automatically sell a security. Second, the test uses a system called vectorbt, which is a way of automatically placing trades, with a fee of 5 basis points, or 0.05%, and slippage, which is the difference between the expected price and the actual price, also of 5 basis points. The test also looks at the drawdown, which is the decrease in value of the fund, and it includes both the decrease from the opening price to the closing price, and a synthetic threshold, which is a minimum amount of decrease that is considered significant. Third, the test uses historical prices that have been adjusted for things like stock splits, and it does not include prices from extended hours or from very small time periods, such as 15 minutes. The test also does not verify that the trades were executed at exactly the right time, or that the profits were managed in a specific way. The results of the test are based on past data and do not guarantee future performance. The test is based on research published by Purdue University and uses data from a website called runway.adityaperswal.workers.dev. All of the inputs and code used for the test are archived and have been verified to be accurate.

A news scan found the headline 'QTM90800 SPY parent later: frozen turn-of-month test'. This is a description of a test of an automated trading strategy. The test was done in two parts: first, a set of rules called Protocol829 was applied to historical data from 2021 to 2025. This protocol involves buying at the close of the second-to-last trading session of the month and selling at the close of the third trading session of the next month. The test looked at the daily returns at the turn of the month, using auction prices as a substitute for actual prices, without setting a stop loss or target price. The test used a total of 59 sets of data, which had been previously used to test other rules. The second part of the test involved using a system called vectorbt to place orders, with a fee of 5 basis points and a slippage of 5 basis points. The test started with $10,000 in cash and made sure that the cash balance never went below zero. The test also looked at the drawdown, which is the difference between the highest and lowest values of the cash balance, including both open and close prices, but not all the small changes within each trading session. The test used historical prices of exchange-traded funds, or ETFs, which are similar to stocks but can be traded like funds. The test did not verify if the trades could be executed exactly at the specified times or if the profits could be managed exactly. The test results cannot be used to make current trades. The sources of the data used in the test are https://business.purdue.edu/faculty/mcconnell/publications/Equity-Returns-at-the-Turn-of-the-Month.pdf and https://runway.adityaperswal.workers.dev/data/stock-bars. All the inputs, code, and results of the test have been archived in 291 parts and verified to be accurate.

A news scan found the text of a trading fund manager's notes. The manager froze a protocol called Protocol829 before inspecting its performance. From 2016 to 2020, the protocol completed 58 training sessions after a warm-up period. The manager removed some data from the training sessions. The protocol buys the SPY ETF at the close of the second-to-last trading session of the month and sells it at the close of the third trading session of the next month. The protocol looks at the daily returns at the start of the month, using auction prices as a substitute for actual prices, without setting a stop loss or target price. The manager used a system called vectorbt to record orders, with a fee of 5 basis points and a slippage of 5 basis points. The manager also tracked the cash balance, making sure it never went negative, and checked that the final balance was correct to within $0.00001. The manager counted all the times the protocol entered the market, not just the times it partially filled an order. The manager also looked at the drawdown, which includes the difference between the high and low prices, but not all the small changes in price during the trading session. The manager used historical prices for the ETF that had been adjusted for dividends and splits, and saved the results of two tests, one with a 10 basis point error and one with a 2/20 basis point sensitivity. The manager did not check if the protocol worked with 15-minute or extended-hours trading sessions, or if it made a profit from whole shares, or if it executed trades at the exact right time. The manager noted that recent good results cannot override the training data, and that there is no current plan to enter the market. The manager cited two sources: a research paper from Purdue University and a website with stock market data. The manager archived all the inputs, code, and results in 291 parts and verified that they could be read back exactly. This is an adaptation of a trading strategy for an ETF, not an exact copy of a strategy from the Center for Research in Security Prices, and it has not been approved for deployment.

A news scan found the headline about a slow trend in the LTC/USD market. The trading rule is based on a 480-observation prior high breakout above the 2400 simple moving average, which is a calculation to determine the direction of the market. The rule also states to exit below the 240 prior low. The system used is called Vectorbt, which is a long-only system with a 25 basis point fee per side and 10 basis point baseline and 30 basis point stress slippage, where basis point refers to a unit of measurement for interest rates and fees. The development of this system took place from 2023 to 2025 after a gap of more than 3 hours in the market, specifically on August 26, 2024, for SOL and February 6, 2023, for LTC. The holdout period, which is a period of time used to test the system, was from January 1 to September 7, 2026, with 5999 bars of data and 1 missing hour. The results of the system were a return of 9.7854% and a drawdown of 12.6753%, where drawdown refers to the decrease in value of the investment. The stress return was 9.1291% and the stress drawdown was 12.6753%. The premises for closing the trade were insufficient, and the daily independent cross frequency calibration failed. The venue stop, target, and profit management with 15-minute fills were untested. The conclusion was to allocate zero funds to this system. The returns for BTC, SOL, and LTC included one open hypothetical trade, while ETH had all closed trades but the return turned negative under stress. The raw quarter Kelly for ETH was 1.332%, but it is naive and cannot be supported. The other currencies lacked both win and loss groups. A strong bull counterexample with slower turnover captured large persistent trends, but this only applies after new independent validation. The current candidate had all four daily entries false. The fees and slippage are assumptions, and the actual LTC roundtrip was 1.1145%, which exceeds the stress approximation of 1.1% before extra impact. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars and https://docs.alpaca.markets/us/docs/crypto-fees. The code used was test_hourly.py and hourly_results.json.

A news scan found the headline about a slow trend in SOL/USD hourly holdout due to too few trades and no venue exit validation. The fixed rule note 827/831 states that there should be a 480-observation prior high breakout above the 2400 SMA, and an exit below the 240 prior low. The next observed open should only happen if the adjacent bar is exactly 1 hour. This uses a vectorbt long-only strategy with a 25bp fee per side, 10bp baseline, and 30bp stress slippage. The development of this strategy took place from 2023 to 2025 after the last gap of more than 3 hours, which occurred on August 26, 2024, for SOL, and February 6, 2023, for LTC. The sweep of 360/480/720 was only used during development. The holdout period was from January 1 to September 7, 2026, with 5999 bars and 1 missing hour. The previously seen 2026 market is a descriptive chronological holdout, and the future paper window is untouched. The results show a buy and hold percentage of -16.64, a return percentage of 13.39, and a drawdown percentage of 19.74. The stress return is 12.26%, and the drawdown is 20.38%. There are several premises that were closed due to insufficient data, failed independent cross frequency calibration, and untested venue stop, target, profit management, and 15-minute fills. The conclusion is zero allocation. The returns for BTC, SOL, and LTC include one open hypothetical trade, while ETH has all closed trades but a return of +0.6572% turns to -0.5434% under stress. The raw ETH quarter Kelly is 1.332%, but it is naive and cannot support it. The others lack both win and loss groups. A strong bull counterexample with slower turnover captures large persistent trends, but it only applies after new independent validation. The current candidate has all four daily entries as false. Fees and slippage are assumptions, and the actual LTC round trip is 1.1145%, which exceeds the 1.1% stress approximation before extra impact. The source of this data is https://runway.adityaperswal.workers.dev/data/crypto-bars and https://docs.alpaca.markets/us/docs/crypto-fees. The code test_hourly.py and hourly_results.json were used. There is no proposal.

A news scan found the headline about an automated trading fund manager's notes. The manager used a fixed rule, note 827/831, which means the fund looks at the price of ETH/USD over 480 observations to see if it breaks out above its prior high and the 2400 simple moving average, or SMA, which is a calculation of the average price of a stock over a certain period of time. The fund exits a trade if the price goes below the prior low. The next trade is only made if the next observed open is exactly one hour later. The fund uses a system called Vectorbt, which is a backtesting and trading platform, with a long-only strategy, meaning it only buys assets, and charges a 25 basis point fee per trade, where a basis point is 0.01%, and assumes a 10 basis point baseline and 30 basis point stress slippage, which refers to the difference between the expected and actual price of a trade. The fund was developed from 2023 to 2025 after a gap of more than 3 hours in the data, which occurred on August 26, 2024, for SOL, and February 6, 2023, for LTC. The fund only uses a certain set of data, called the holdout, for testing, which includes 5999 bars of data from January 1 to September 7, 2026, but is missing one hour of data. The results of the test show a return of 0.6572% and a drawdown of 16.0014%, which is the maximum loss of the fund during the testing period. The stress return, which is the return of the fund under stressful market conditions, is -0.5434%, and the drawdown is 16.5838%. The fund's premises, or assumptions, were not met, including insufficient trades and untested venue stop and target management. Therefore, the conclusion is to allocate zero funds to this strategy. The results also show that the fund's returns are not stable under stressful conditions, and the fees and slippage assumptions may not be accurate. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars and https://docs.alpaca.markets/us/docs/crypto-fees, and the code used to test the strategy is test_hourly.py and hourly_results.json.

A news scan found the headline about a slow trend in the BTC/USD market. The trading rule is based on a 480-observation prior high breakout above the 2400 simple moving average, which is a calculation of the average price of a security over a certain period of time, in this case 2400 hours. The rule also includes an exit below the prior low. The next observed open is only considered if the adjacent bar is exactly 1 hour. The trading system used is called Vectorbt, which is a backtesting and trading library, with a long-only strategy, a 25 basis point fee per side, and a 10 basis point baseline and 30 basis point stress slippage. Slippage refers to the difference between the expected price of a trade and the actual price. The system was developed from 2023 to 2025 after a gap of more than 3 hours in the market data, which occurred on August 26, 2024, for SOL, and February 6, 2023, for LTC. The holdout period, which is a period of time used to test the system, was from January 1 to September 7, 2026, with 5999 bars of data and 1 missing hour. The results of the system were a return of 18.79%, a drawdown of 5.89%, and a win rate of 100%. The stress return was 18.08% and the drawdown was 6.27%. However, the premises for the system were not met, including insufficient trades and untested venue stop and target management. Therefore, the conclusion was to allocate zero funds to this system. The returns for BTC, SOL, and LTC include one open hypothetical trade, while ETH had all trades closed but the return was negative under stress. The raw quarter Kelly, which is a measure of the system's performance, for ETH was 1.33% but this is considered naive and not reliable. The system lacks both win and loss groups for all currencies except ETH. A strong bull counterexample with slower turnover captures large persistent trends, but this only applies after new independent validation. The current candidate for the system has all four daily entries as false. The fees and slippage are assumptions, and the actual roundtrip fee for LTC was 1.1145%, which exceeds the stress approximation of 1.1%. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars and https://docs.alpaca.markets/us/docs/crypto-fees. The code used to test the system is test_hourly.py and the results are in hourly_results.json.

A preregistered note, number 830, was written before returns were calculated, using a specific method: T-5 open and T-1 close, excluding events that happened within a seven-day period, and using data from 2021 to 2024 for training and from 2025 to September 4, 2026, for testing, with 16 and 6 samples, respectively. The 22 SEC Item 2.02 dates matched the issuer's calls, which can be found at https://investor.scholastic.com/financial-information/quarterly-results. The analysis used adjusted daily data from 1425 bars, obtained from https://runway.adityaperswal.workers.dev/data/stock-bars. The vectorbt long-only strategy was used, with fees of 5 basis points and a slip of 10 basis points on each side, and the results were checked against independently compounded trades. The training return was -13.8619%, with a drawdown of 17.7126%, a win rate of 50%, and a mean return of -0.8216%. In contrast, the testing return was 6.3868%, with a drawdown of 4.7942%, a win rate of 66.667%, and a mean return of 1.06968% over 5 sessions. A bootstrap analysis with 20,000 samples and a seed of 908 gave a mean 95% confidence interval of [-0.60133, 3.41469]%. A mechanical holdout quarter Kelly criterion of 12.3659% was not adopted, and the training Kelly criterion was 0. A sensitivity analysis using a 1 ATR stop and a 1.5 ATR target gave a testing return of 3.01377%, with a drawdown of 7.74758%, and a mean return of 0.547997%. A stress test with a 25 basis point slip gave a return of 1.17611%, with a mean return of 0.24680%. The analysis used a daily worst-order stop-first and gap-open stop, with a capped target. A descriptive T3/T10 sweep of all 24 configurations and observations from 9479 to 9484 did not select a holdout winner. The baseline analysis for the September Q1 subgroup from 2021 to 2024 showed all 4 negative results, versus a return of 0.7623% for 2025, with only 1 out-of-sample result for the same quarter. The analysis has several limitations, including the availability of historical advance notices, early pre-announcements, corporate actions, and cash dividends, as well as the use of partial IEX data and unvalidated 15-minute and extended quotes. Additionally, the daily marked drawdown can miss intraday lows. This is not an executable strategy. A separate gross diagnostic using a generic CLI SMA 10/50 with 16 samples gave a return of -16.54625%, with a drawdown of 65.46458%, and a win rate of 18.75%. The conclusion is that there should be no allocation, as the plausible positive holdout does not repair the failed training and wide uncertainty. The analysis will be frozen prospectively from September 17 open to September 23 close, and this is a paper-only analysis.

A news scan found the note: Rule 827/831 was completed with a close above the prior 20 daily highs and the 100-day Simple Moving Average, which is a calculation of the average price of a stock over a certain period of time. The rule also included an exit close below the prior 10 daily lows. Signals were shifted to the next open, using a long-only strategy with 25 basis points fees per side and 10 basis points slippage per side as a baseline, and 30 basis points stress. The development of this rule took place from 2023 to 2025, with a long gap in the data after August 26, 2024, for the SOL symbol. The development only considered neighbors of 15, 20, and 30 days. The rule was evaluated from January 1 to September 7, 2026, with a starting cash amount, and a warm-up period from the development data. Since the 2026 market history has already been seen, this is a descriptive chronological holdout, not an untouched authorization. No stop, target, or 15-minute enforcement was simulated. The current candidate fails to meet the requirements for a fresh entry and sample gates before implementation. The baseline data includes the symbol LTC/USD, a timeframe of 1 day, and a window of holdout, with a lookback of 20 days and slippage of 0.001. The data starts on January 1, 2026, and ends on September 7, 2026, with 250 bars and an original 1346 bars. The return percentage is 9.088471452870035, and the drawdown percentage is 10.059013636356118. The stress return is 8.436331222614681%, and the drawdown is 10.239075671118192%. The marked return includes one open hypothetical trade, but the closed win rate is not sufficient to estimate a durable edge. No live holding is implied. The premises for rejecting allocation include an incomplete payoff distribution, exits that differ from the required venue, and a positive mark-to-market that cannot replace validation. The conclusion is to reject allocation and retain a future prospective fixed rule. A gap filter removes all data before the last historical gap of more than 3 bars and never joins the SOL year outage. Existing hourly gaps are audited separately. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars, and the fees are https://docs.alpaca.markets/us/docs/crypto-fees. The code test_rule.py and raw results rule_results.json are also available.

A news scan found the headline about a slow trend in SOL/USD. The trading rule was preregistered as note 827/831, which means it was planned in advance. The rule was to buy when the price closed above the highest price of the last 20 days and the 100-day simple moving average, which is the average price over 100 days. The rule was to sell when the price closed below the lowest price of the last 10 days. The signals were shifted to the next open, which means the decision to buy or sell was made at the start of the next trading day. The trading system used a long-only strategy, which means it only bought and never sold short, and it used a system called vectorbt. The system also accounted for fees of 25 basis points, or 0.25%, per trade, and slippage of 10 basis points, or 0.10%, per trade, which is the difference between the expected price and the actual price. The system was tested from 2023 to 2025, but only used data from after August 26, 2024, for the SOL cryptocurrency. The system was also tested on neighboring days of 15, 20, and 30 days. The test started with a fixed amount of cash on January 1, 2026, and used a warm-up period from the development data. The test used a descriptive chronological holdout, which means it used existing market data to test the system, rather than trying to predict future market data. The test did not simulate stop-loss or target prices, and it did not enforce any rules for 15-minute intervals. The current candidate, or trading system, failed to meet the requirements for a fresh entry and sample gates before implementation. The baseline results were: the symbol was SOL/USD, the timeframe was 1 day, the window was a holdout, the lookback days were 20, the slippage was 0.001, the start date was January 1, 2026, the end date was September 7, 2026, there were 250 bars of data, the original number of bars was 928, 185 bars were removed due to a prior long gap, there were no missing bars, the buy-and-hold percentage return was -16.64%, the return percentage was 11.43%, the drawdown percentage was 12.83%, there was 1 trade, the win rate percentage was 0%, the average win percentage was null, the average loss percentage was 8.06%, the expectancy percentage was -8.06%, the raw quarter Kelly was null, there was 1 open trade, and the returns were -0.0806. The stress return was 10.76% and the drawdown was 13.00%. The marked return included one open hypothetical trade, but the closed win rate was not enough to estimate a durable edge or Kelly. The premises for rejecting allocation were: the payoff distribution was incomplete, the exits differed from the required venue, and a positive mark-to-market cannot replace validation. The conclusion was to reject allocation and retain a future prospective fixed rule. The gap filter removed all data before the last historical gap of more than 3 bars, and never joined the SOL year outage. Existing hourly gaps were audited separately. The source of the data was https://runway.adityaperswal.workers.dev/data/crypto-bars, and the fees were from https://docs.alpaca.markets/us/docs/crypto-fees. The code was tested in a file called test_rule.py, and the raw results were in a file called rule_results.json.

A news scan found the note: Rule 827/831 was completed with a close above the prior 20 daily highs and the 100-day Simple Moving Average, which is a calculation of the average price of a stock over a certain period of time. The exit strategy is to close below the prior 10 daily lows. Signals were shifted to the next open, using a long-only strategy with 25 basis points fees per side and 10 basis points slippage per side as a baseline, and 30 basis points stress. This strategy was developed from 2023 to 2025, with a long gap in SOL after August 26, 2024. The development neighbors were 15, 20, and 30 days, and the fixed 20 was evaluated from January 1 to September 7, 2026, starting with cash and warming up from development. Since the 2026 market history has already been seen, this is a descriptive chronological holdout, not an untouched authorization. No stop, target, or 15-minute enforcement was simulated. The current candidate fails fresh entry and sample gates before implementation. The baseline is a set of data that includes the symbol ETH/USD, a timeframe of 1 day, and a window of holdout. The results show a buy and hold percentage of -16.04, a return percentage of 3.91, and a drawdown percentage of 7.24. The stress return is 3.29% and the drawdown is 7.43%. The marked return includes one open hypothetical trade, and the closed win rate is not enough to estimate a durable edge. No live holding is implied. The premises for rejecting allocation are: 1, the incomplete payoff distribution; 2, exits differ from the required venue; and 3, positive mark-to-market cannot replace validation. The conclusion is to reject allocation and retain a future prospective fixed rule. A gap filter removes all data before the last historical gap of more than 3 bars, and never joins the SOL year outage. Existing hourly gaps are audited separately. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars, and the fees are https://docs.alpaca.markets/us/docs/crypto-fees. The code test is test_rule.py, and the raw results are rule_results.json.

A news scan found the note: Rule 827/831 was completed with a close above the prior 20 daily highs and the 100-day Simple Moving Average, which is a calculation of the average price of a stock over a certain period of time. The exit strategy is to close below the prior 10 daily lows. The signals were shifted to the next open, using a long-only strategy with 25 basis points fees per side and 10 basis points slippage per side as the baseline, and 30 basis points stress. The development period was from 2023 to 2025, with a long gap in SOL after 2024-08-26. The development neighbors were 15, 20, and 30 days only, with a fixed 20 evaluated from January 1 to September 7, 2026, starting with cash, and a warm-up from development. Since the 2026 market history has already been seen, this is a descriptive chronological holdout, not an untouched authorization. No stop, target, or 15-minute enforcement was simulated. The current candidate fails the fresh entry and sample gates before implementation. The baseline is a set of parameters, including symbol, timeframe, window, lookback days, slippage, start and end dates, and number of bars. The results show a buy and hold percentage of -9.62%, a return percentage of 14.41%, a drawdown percentage of 5.30%, and a win rate percentage of 100%. The stress return is 13.72% and the drawdown is 5.49%. The marked return includes one open hypothetical trade, but the closed win rate cannot estimate a durable edge or Kelly, which is a formula to determine the optimal size of a bet. No live holding is implied. The premises for rejecting allocation are: 1, the incomplete payoff distribution; 2, the exits differ from the required venue; and 3, the positive mark-to-market cannot replace validation. The conclusion is to reject allocation and retain the future prospective fixed rule. A gap filter removes all data before the last historical gap of more than 3 bars and never joins the SOL year outage. Existing hourly gaps are audited separately. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars and the fees are https://docs.alpaca.markets/us/docs/crypto-fees. The code is test_rule.py and the raw results are rule_results.json.

A news scan found the headline about a breakout baseline audit. The command node ran a backtest on the symbols BTC/USD, ETH/USD, and SOL/USD using the breakout strategy with a lookback of 20 days and 1345 days of data, starting on September 8. The test did not include a separate holdout or explicit fees and slippage, and it did not use the new slow-rule implementation. The strategy was to go long when the 20-day close was high and to stay flat otherwise, not entering at the 20-day high and exiting at the 10-day low. The results were: BTC had a return of 56.0766%, a drawdown of 15.9694%, a win rate of 30.3030%, 99 trades, and a buy-and-hold return of 374.3371%; ETH had a return of 28.6762%, a drawdown of 24.9433%, a win rate of 28.3951%, 81 trades, and a buy-and-hold return of 105.0607%; and SOL had a return of 838.0652%, a drawdown of 25.1510%, a win rate of 45.6522%, 46 trades, and a buy-and-hold return of 818.0603%. However, the SOL trade from July 6, 2023, to August 27, 2024, had a long data gap, which means there is missing history. This is a problem because the premise of the test is that the reported performance depends on observed prices, and missing history cannot justify exits. Additionally, costs and a holdout are required for a valid test. Therefore, the conclusion is that these headline figures should be rejected as trading evidence. A separate test using vectorbt with costs and a gap filter will follow. This is a valid inference, but it does not prove that the slow rule fails. The historical endpoint pagination and current 2026 familiarity also limit independent validation. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars, and the raw CLI result is saved in cli.json.

A news scan found the headline about a fixed breakout for the stock ticker BWEN. The data used for this analysis came from a website and included 1002 adjusted bars of stock data from September 8, 2022, to September 4, 2026. A command line interface was run with the following parameters: symbols BWEN, strategy breakout, lookback 20, days 1460, stop percentage 5, and target percentage 8. The results showed a gross return of -0.581118%, a drawdown of 25.928124%, and 32.653061% of trades were winners. This analysis was done for diagnostic purposes only and did not take into account costs. Another analysis was done using vectorbt version 1.1.0, which is a backtesting framework, with a fixed 20-day close breakout strategy, and the results were trained through December 31, 2025, with a return of 25.191988% and a drawdown of 24.545056%. The holdout period from January 1, 2026, to September 4, 2026, showed a return of -41.566941% and a drawdown of 48.55146%. The average win was 8.465371% and the average loss was 8.003927%. When stress testing with a slippage of 0.3% on each side, the return was -42.332839%. Other diagnostic lookbacks of 10 and 40 days also failed the holdout period with returns of -39.277357% and -45.131608% respectively. The stress testing for these lookbacks showed returns of -44.307145% and -45.525128%. No winning parameters were chosen. The baseline was specified before the independent test, but the ticker was selected with current fundamentals. This was not a prospective universe holdout. The daily open, high, low, and close data and the stop priority do not reproduce the 15-minute venue, entry day order, or the partial take profit. The limitations of this analysis cannot justify upgrading a severe rejected result. This is a reject rule and not all possible BWEN strategies. The reproduction of the local research and test results are preserved in the file test_bwen.py and the record observation 9303 or the nearest result id carries all metrics.

A news scan found the headline 'QIM90708 full sensitivity and closing-price audit: rejection survives endpoint changes'. The audit examined the performance of 4 ETFs, including SPY, EEM, EFA, and TLT, with 3 splits, 5 rules, and 6 one-way costs of 0, 1, 2, 5, 10, and 20 basis points. The main results showed holdout returns at 10 basis points: SPY -77.7773%, EEM -71.3673%, EFA -70.4248%, and TLT -59.6151%. At 5 basis points, all 20 ETF/rule holdout cells were negative. The TLT main 2 basis point holdout return was +0.8522% over 5 years, with a training return of -21.0091% and a 2026 return of -3.937%. The ROD 2 basis point holdout return was +3.769%, with a training return of -17.864% and a recent return of -3.526%. No cheap cost variant was adopted. The audit excluded early closes using the exchange_calendars 4.13.2 reconstructed NYSE calendar, which includes more than the 13 observed bars, to prevent extended trading from mimicking a full session. A discrepancy was found between the 30-minute close and daily close for SPY, with a maximum of 279.40 basis points on March 12, 2020, and similar discrepancies were found for EEM and EFA. The official daily close diagnostic for SPY holdout at 10 basis points was -77.1045%, with a gross return of -6.4085%. The audit concluded that the tested transfers were rejected, with no stop or target adaptation or parameter retuning. The sources for the data include https://runway.adityaperswal.workers.dev/data/stock-bars and https://www.nyse.com/trade/hours-calendars, as well as primary papers in analysis 660.

A recent test of a trading strategy for the TLT ETF, which is a type of investment fund, showed the following results. The test used historical data from 2016 to 2026, up to September 4, and only considered full trading days on the New York Stock Exchange. The strategy was tested with a 5 basis point fee, which is a type of trading cost, and 5 basis point slippage, which is the difference between the expected and actual price of a trade, on each side of the trade. The results showed a return of -16.151854%, which means the strategy lost value, and a maximum drawdown, which is the largest loss from peak to trough, of 16.151854%. The strategy had a win rate of 1.176471%, a mean return of -0.206994%, an average win of 0.028136%, and an average loss of 0.209793%. These results are based on simulated trades, not actual account performance. The test did not include same-day stop or target orders, or guarantees of auction fill, and excluded intrabar drawdown, which is the loss within a single trading period. The conclusion is to reject this strategy due to its poor performance. The test was run with six different cost variants, but changing the costs did not improve the results. The original research papers used for this test can be found at https://smallake.kr/wp-content/uploads/2015/01/SSRN-id2440866.pdf and https://academicweb.nd.edu/~zda/intramom.pdf. This test was a simplified version of the original research and did not validate the strategy for future use.

A news scan found the headline 'QIM90708 TLT pdt holdout: costed final-half-hour cash-only replication'. This is a report about a trading strategy test. The test used historical data from 2016 to 2020 to train and from 2021 to 2025 to validate, and also included recent data up to September 4, 2026. The test was done on full New York Stock Exchange sessions, using a calendar and stock bar data from a specific website. The strategy was tested with fees of 5 basis points and slippage of 5 basis points on each side of a trade. The test results showed a return of -55.191299%, with 11.350294% of trades being winners, and an average win of 0.135632% and an average loss of 0.194284%. The test did not include same-day stop or target orders, and did not guarantee auction-fill orders. The results are simulated and not actual account performance. The test concluded that this strategy should be rejected, and that changing costs did not improve the results. The original research papers used for this test can be found on specific websites, and this test is a transfer of a strategy to an ETF, not a full replication of a research paper. The historical market period used for the test is not a pristine example of prospective validation.

A news scan found the report of an automated trading fund manager. The report is about a trading strategy called QIM90708, which involves buying and selling the TLT ETF. The strategy was tested using historical data from 2016 to 2026, with the most recent data up to September 4, 2026. The test used only full New York Stock Exchange sessions and was calibrated with a tool called exchange_calendars. The data was also adjusted using a method called SIP bars, which is a way of adjusting stock prices to account for certain market activities. The strategy was tested with a 5 basis point fee and 5 basis point slippage, which is a measure of the difference between the expected price of a trade and the actual price. The results showed a return of -60.584141%, with a maximum drawdown of -60.584141%. The strategy had 11.735537% winning trades, with an average win of 0.128209% and an average loss of -0.191110%. These results are based on simulated trades and do not reflect actual account performance. The report concludes that this strategy should be rejected, as changing the costs of the trades did not improve the results. The report cites two academic papers as sources, but notes that the historical market period used in the test is not a perfect representation of future market conditions.

The QIM90708 TLT strategy was tested using historical data from 2016 to 2020 for training and 2021 to 2025 for holdout, with additional recent data up to September 4, 2026. The test only used full New York Stock Exchange sessions and was calibrated with exchange calendars version 4.13.2, using adjusted SIP bars from https://runway.adityaperswal.workers.dev/data/stock-bars. The strategy was tested with 5 basis point fees and 5 basis point slippage on each side of the trade, using a long-only direction and all available strategy capital, with cash reconciled independently. The results showed a return of -59.615132%, with a closed wealth drawdown of 59.615132%, 572 trades, a win rate of 11.363636%, a mean return of -0.158261%, an average win of 0.126794%, and an average loss of 0.194806%. These results are simulated full-capital returns, not actual account performance. The test did not include same-day stop or target orders, or auction-fill guarantees, and excluded intrabar drawdown. The conclusion is to reject this costed transfer, with an actual size of 0. All six cost variants were recorded, but changing costs did not select a winner. The original sources are https://smallake.kr/wp-content/uploads/2015/01/SSRN-id2440866.pdf and https://academicweb.nd.edu/~zda/intramom.pdf, and this is an ETF long-only transfer, not a full paper replication. The historical market period used is not a pristine prospective validation.

A trading strategy called QIM90708 TLT was tested. The test used historical data from 2016 to 2020 and held out data from 2021 to 2025, with recent data up to September 4, 2026. The test only used full New York Stock Exchange sessions and was calibrated with a tool called exchange_calendars version 4.13.2. The data was adjusted using SIP bars from a website. The test had fees of 5 basis points and slippage of 5 basis points on each side of a trade. The test only considered long positions, used all available capital, and reconciled cash independently. The results showed a return of -71.383763%, with a maximum drawdown of 71.383763%. The test had 659 trades, with a win rate of 4.552352%, a mean return of -0.189575%, an average win of 0.119127%, and an average loss of 0.204299%. These results are simulated and not actual account performance. The test did not guarantee same-day stops or targets, and did not account for intrabar drawdown. The conclusion is to reject this costed transfer, with a size of 0. Six different cost variants were tested, but changing costs did not select a winner. The original sources of the strategy are two academic papers, and this test is a transfer of an ETF long-only strategy, not a full replication. The historical market period used is not a pristine prospective validation.

A news scan found the report of a trading simulation. The simulation used historical data from 2016 to 2020 to train and 2021 to 2025 to test, with additional recent data up to September 4, 2026. It only considered full New York Stock Exchange sessions and adjusted for certain market factors. The simulation assumed a 5 basis point fee and 5 basis point lippage, which is a measure of the difference between the expected price of a trade and the actual price, on each side of a trade. The strategy was long-only, meaning it only considered buying, and used all available capital. The results showed a return of -72.035476%, with a maximum drawdown of 72.035476%. The simulation had 649 trades, with a win rate of 4.776579%, a mean return of -0.195917%, an average win of 0.145722%, and an average loss of 0.213055%. These results are based on simulated full-capital returns, not actual account performance. The simulation did not guarantee same-day stops or targets and excluded intrabar drawdown, which is a measure of the decline in value of a trade during a single trading period. The conclusion was to reject this costed transfer, with no actual size. Six different cost variants were recorded, but changing costs did not select a winner. The original sources of the data were from academic papers. This simulation was for an ETF long-only transfer, not a full replication of a research paper. The historical market period used was not a pristine prospective validation, meaning it was not a perfect test of the strategy's potential performance in the future.

A recent test of a trading strategy called QIM90708 EFA had the following results. The test used historical data from 2016 to 2026, up to September 4, and only considered full New York Stock Exchange sessions. The data was adjusted using a tool called exchange_calendars version 4.13.2 and stock bars from a website. The test assumed a 5 basis point fee and 5 basis point lippage, which is a type of trading cost, on each buy and sell. The test only considered long positions, meaning it only bought stocks, and used all available capital. The results showed a return of -14.042114 percent, with a maximum drawdown of -14.042114 percent. The test had 91 trades, with a win rate of 6.593407 percent. The average win was 0.210324 percent and the average loss was 0.192614 percent. These results are simulated and do not reflect actual account performance. The test did not use same-day stop or target prices, and did not guarantee auction-fill, which is a type of order. The test also excluded intrabar drawdown, which is a type of risk measure. Based on these results, the strategy was rejected. The test was run with six different cost variants, but changing the costs did not improve the results. The original research papers used for this test can be found at https://smallake.kr/wp-content/uploads/2015/01/SSRN-id2440866.pdf and https://academicweb.nd.edu/~zda/intramom.pdf. This test was a simulation of an ETF long-only transfer, not a full replication of a research paper. The historical market period used for the test is not a pristine prospective validation, meaning it is not a perfect test of the strategy's potential performance.

The automated trading fund manager analyzed the EFA product using a costed final-half-hour cash-only replication method. The analysis involved frozen notes 651 and 653 with a positive result, and the earliest signal allowed by the maximum three trades per rolling five sessions. The entry point was at 15:30 bar open and the exit point was at the same 30-minute bar close, resulting in a flat position on the same day. The training period was from 2016 to 2020, the holdout period was from 2021 to 2025, and the recent period was from 2026 through September 4. The analysis only considered full NYSE sessions and was calibrated with exchange calendars version 4.13.2, using adjusted SIP bars from a specific website. A 5 basis point fee and 5 basis point slippage were applied to each side of the trade, with a long-only direction and all available strategy capital, and cash was reconciled independently. The results showed a return of -64.124066%, a closed wealth drawdown of 64.124066%, 538 trades, a win rate of 5.018587%, a mean return of -0.190258%, an average win of 0.160302%, and an average loss of 0.208781%. These results are simulated full-capital returns and not actual account performance. The analysis did not include same-day stop or target, or auction-fill guarantee, and intrabar drawdown was excluded. The PDT throttle was applied per isolated ETF, not as a combined account allocation. The conclusion was to reject this costed transfer, with an actual size of 0. All six cost variants were recorded, but changing costs did not select a winner. The original sources were from specific research papers, and this analysis was an ETF long-only transfer, not a full paper replication. The inspected historical market period was not a pristine prospective validation.