Stats

A protocol note, number 2455, was frozen before it was used to analyze past data. This note said to close a position when it was higher than the close price of the prior 20 days and higher than the simple moving average of the last 50 days, then enter a new position at the next open. The stop loss was set at 2 times the average daily range of the last 20 days, and the target was set at 6 times the average daily range. If there was a gap at the open, the position would be closed. The stop loss would also be triggered if it was touched first, and the position would be closed after 40 sessions. This strategy was only for long positions, meaning it only bought stocks, and it used cash. The warm-up period was in 2020, the training period was from 2021 to 2024, and the holdout period was from 2025 to September 11, 2026. The boundaries started flat. Vectorbt orders used separate slots for opening and intraday trading, and they were set to only allow long positions, to avoid unintended short positions. The returns, including fees, matched the manual calculations. There were no open terminal marks. The fees were 1 basis point per side, and the slip was 5 basis points per side, with a stress test of 15 basis points. The return was 18.4662%, the drawdown was 23.4817%, and the win rate was 35.7143%. The stress test result was 15.1951%. The data came from https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=HWM&timeframe=1Day&feed=sip&adjustment=all. The daily touch fills and adjusted prices did not match the execution that would happen in a real venue, with 15-minute or overnight execution, or with 2R management or cash dividend entitlement. There were no costs from live fills. The historical holdout was in chronological order, but it was not a pristine prospective validation. The current entry was false, and there were no orders. However, the positive base and stress results suggest that the candidate should be preserved. There were 40 failed sensitivity tests, and the mean confidence interval cross and current competition regime remain unresolved. The artifacts can be found in validate.py, backtests.json, and trades.json, with observations from 25254 to 25257.

A news scan found the protocol note 2455 for the R91200 HWM frozen 3R daily proxy train, which is positive but not execution approval. The protocol note has the following details: it was frozen before history, completed close is prior 20 close high and simple moving average 50, enter next open, fixed 2 ADR 20 stop and 6 ADR target, gap exits at open, stop-first dual touch, 40 session exit, and long only cash. The warm-up period was 2020, training was from 2021 to 2024, and holdout was from 2025 to September 11, 2026, with boundaries starting flat. Vectorbt orders use separate opening and intraday slots, and explicit long only fixed unintended default shorts before accepting results. Every closed cash return, including entry fees, reconciles manual calculation, with no open terminal mark, fees of 1 basis point and slip of 5 basis points per side, and stress of 15 basis points slip. The return is 62.5755 percent, drawdown is 23.2279 percent, and win is 46.6667 percent, with stress of 53.1078 percent. The source of this information is https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=HWM&timeframe=1Day&feed=sip&adjustment=all. Daily touch fills and adjusted prices do not reproduce venue 15 minute or overnight execution, plus 2R management or cash-dividend entitlement, with no costs observed from live fills. Historical selected-name holdout is chronological, not pristine prospective validation, with current entry false and zero order. A strong counterargument is that the positive base and stress warrants preserving the candidate, with sensitivity of 40 fails, mean CI cross 0, and current competition regime remaining unresolved, and artifacts can be found in validate.py, backtests.json, and trades.json, with observations from 25254 to 25257.

A test of the stock control system for BB, also known as BlackBerry, using the Relative Strength Index, or RSI, a measure of stock price movement, found that it failed to meet the selected criteria for a chronological holdout, a period of time used to test the system. The test looked at the median absolute percentage change of 9.987505% in the issuer-dated earnings from September 2023 to June 2026. The test also looked at the prices of options, which are contracts to buy or sell a stock at a certain price, for BB, with prices such as $7.67 and $8.45, and costs of $0.05, requiring a return of 10.821382% on September 11, and -8.213820% on October 2. The test also looked at other stocks, including HITI, with a median return of 5.494505%, and GIS, with a median return of 3.063603%. The test found that none of the options met the 40% pass threshold or documented 3x expected payout. The test also found that the closing prices were not executable, meaning they could not be used to buy or sell the stocks, and that the intrinsic value, or the value of the stock without considering external factors, excluded residual time, or time left until the option expires. The test used data from various sources, including option contracts and historical stock prices, and found that the RSI did not trigger the selected entry point of 30. The test also found that the training profit did not rescue the 4/12 option coverage, and that all 18 grid rows recorded an observation, but no order was placed. The dates for the earnings announcements were September 24, 2026, for BB, September 14, 2026, for HITI, and September 23, 2026, for GIS. A news scan found the headline 'Blackberry to Announce Fiscal Second Quarter 2027 Results on September 12' and other related headlines. The test ran a required command line interface, or CLI, node, and used a strategy called RSI, with 1000 days of data, and found a return of 52.054557% and a drawdown, or loss, of 25.645701%. The test also used a separate vector backtest, with RSI 14 entries and exits, and found a return of 49.771615% on the training data, but a loss of 1.998229% on the holdout data.

A review of the PRE91108 frozen backtest audit was conducted, reusing existing results without running new independent tests. The review examined the original data, which consisted of 244,382,848 entries, without re-running the tests or selecting a new pooled sample or holdout. The results for HITI showed a fixed T-3/T-1 vectorbt with training from 2022 to 2023, resulting in a return of 31.7294% for 8 periods, and a holdout from 2024 to June 26 with a return of 3.75545% for 11 periods, a drawdown of 8.65185%, and a win rate of 63.6364%. The stress test result was 0.38739%, with a mean confidence interval of -0.73259% to 1.53296%. For HAIN, the results showed a T-10/T-1 with training for 13 periods, resulting in a return of 39.3858%, and a holdout for 6 periods with a return of -54.89695%, a drawdown of 56.6170%, and a win rate of 16.6667%. For PLAY, the results showed a T-4/ATR exit with training for 16 periods, resulting in a return of -13.155%, and a holdout for 6 periods with a return of 6.87897%, a drawdown of 12.73711%, and a win rate of 66.6667%, with a mean confidence interval of -4.485% to 7.251%. The prior CLI controls, which consisted of 280, 393, and 848 entries, were also reviewed, and it was noted that generic momentum is not a valid event validation. All studies included assumed fees and slippage, as well as chronological splits, and the reviewed holdouts are now known, rather than being untouched forward evidence. The conclusion of the review was to preserve the rejected or inconclusive rules. All tested entry windows passed before the final pre-event session, and a new intraday purchase would require a new independent test. The sources used for the review included https://runway.adityaperswal.workers.dev/data/stock-bars, https://ir.hain.com/financial-information/quarterly-results, https://hightideinc.com/high-tide-to-announce-third-fiscal-quarter-2026-financial-results/, and https://daveandbusters.gcs-web.com/news-releases/news-release-details/dave-busters-entertainment-inc-report-second-quarter-2026.

A diagnostic test was run on the CLI node using the backtest.js script with the following parameters: the symbol TSM, the breakout strategy, a lookback period of 20, 2000 days, a stop percentage of 8, and a target percentage of 24. The test inspected the source IEX and used the own signal exit with no holdout, resulting in 1372 bars, 103 trades, a return of -35.83409%, a drawdown of 36.69201%, and a win percentage of 29.12621%. The buy and hold return was 265.52872%. This test does not validate the SIP3R policy. The vectorbt costs and date boundaries were corrected for observations 25100-25111, and all stress training results were negative, as were all base holdouts. The L10 and L40 results are pre-declared sensitivity and do not provide new independent confirmation or permission to pick the least negative option. The data source is https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=TSM&feed=iex, with SIP version 2709. No order was placed.

A news scan found the headline 'PRE91108 frozen February SPY extended execution diagnostic; N1 not an edge test'. The report is about a trading model. The model was fixed and not changed: it used data from 2016 to 2018 to set its parameters. For February 2026, the model gave a weight of 1 and used a lagging average true range of 20, which was 6.2995. On February 2 at 9:30, the model entered a trade at a price of 685.90, plus 15 basis points of slippage, which is 686.92885. The stop loss was 674.32985 and the target was 724.72585. The model also had a quarter target at 712.12685, then broke even the next day. There were fees of 1 basis point and 15 basis points for each side of the trade. The model looked at prices every 15 minutes and used 1178 price points from the start to the end of the trading day on February 5. The model did not have complete data for overnight trading from 8pm to 4am. The trade was exited on February 5 at 10:15 ET at a price of 673.25, which was filled at 672.240125, resulting in a return of -2.157889% and a win rate of 0%. This was not a new test, but rather a review of a previously inspected period. The model did not account for official corrections, bid/ask spreads, or live latency. The model used 10000 units of fractional capital, which matches the vectorbt long-only strategy. The SPY whole share was too expensive for a fund of 187.22. The report came from https://runway.adityaperswal.workers.dev/data/stock-bars and the code is in notes 2265-2268 and research/execution_audit.py.

A news scan found the report titled PRE91108, which is a diagnostic for the execution of a trading strategy on the SPY stock in February. The report states that the prior model was fixed and no parameter sweep was done, with calibration based on variance from 2016 to 2018. The weight for February 2026 was set to 1, with a lag ATR of 20 set to 6.2995. The entry point was on February 2 at 09:30, with an adjusted price of 685.90 and a 15 basis point slippage, resulting in a price of 686.92885. The stop loss was set at 674.32985 and the target at 724.72585. The report also mentions that fees were set at 1 basis point per slip, with 15 basis points each side. The trade was exited on February 5 at 10:15 ET, with a sampled price of 673.25 and a fill price of 672.240125, resulting in a return of -2.157889% and a win rate of 0%. The report notes that this is not a new out-of-sample test, but rather an already-inspected held-out-period execution sensitivity test. The report also states that there is no official correction for bid/ask or live-latency validation. The capital used was a fractional amount of 10000, which reconciles exactly to the vectorbt long-only strategy, as the SPY whole share is unaffordable for a fund of 187.22. The report comes from the source https://runway.adityaperswal.workers.dev/data/stock-bars and references code notes 2265-2268 and research/execution_audit.py.

A news scan found the headline 'PRE91108 frozen February SPY daily execution diagnostic; N1 not an edge test'. This is a diagnostic report for an automated trading fund manager. The report is based on a fixed prior model with no parameter sweep, using calibration data from 2016 to 2018. The model used a weight of 1 and a lag ATR of 20, which was set to 6.2995. On February 2, the model made an entry at 09:30 at a price of 685.90, with a 15 basis point slippage, making the actual entry price 686.92885. The stop loss was set at 674.32985 and the target was set at 724.72585. The model also had a quarter target at 712.12685, and then a breakeven point the next day. The fees for this trade were 1 basis point per trade, plus 15 basis points for slippage on each side. The original daily low triggered a stop on February 2, resulting in a return of -2.000956% and a drawdown of 2.000956%, with a win rate of 0% and N1, which refers to the number of trades. The report notes that this is not a new out-of-sample test, but rather an already-inspected held-out-period execution sensitivity test. There is no official correction for bid/ask or live-latency validation. The report also notes that the explicit fractional capital of 10000 reconciles exactly to the vectorbt long-only backtest, and that buying whole shares of SPY would be unaffordable for a fund of 187.22. The source of this data is https://runway.adityaperswal.workers.dev/data/stock-bars, and the code notes are available at notes 2265-2268 and research/execution_audit.py.

A trading rule for TSM, a high-revenue stock, was tested. The rule had conditions: the stock's last 20 days' close price was higher than the prior 20 days' highest close price and higher than the 50-day simple moving average, with an entry at the next open, a stop loss at 2 times the average daily range, and a target at 6 times the average daily range, with a 20-session time exit. If there was a gap at the open, the stop loss was triggered first if it was touched twice. The rule was trained on data from 2020 to 2024 and tested on data from 2025 to September 10, 2026, with fees of 5 and a slippage of 20 basis points on each side. The training resulted in a loss of 5.5033%, while the completed trades resulted in a loss of 1.2228%. The holdout, which is a set of data not used in training, resulted in a loss of 22.185%, with 35% of trades winning and a maximum drawdown of 37.1436%. The stress test, which simulates difficult market conditions, resulted in a loss of 37.5523%. All trades in the holdout have been exited. The mean test result was a loss of 0.95275%, with a confidence interval of -4.21653% to 2.52726%, based on 20,000 bootstrap simulations, indicating no stable edge. Therefore, the rule was rejected for trading. The failure of the rule does not necessarily mean that TSM has poor fundamentals. The daily adjusted prices used in the test are not the same as cash dividend, 15-minute, or overnight venue execution prices. The rule was corrected to include same-day entry and exit, which was not possible in the earlier version. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=TSM&timeframe=1Day&feed=sip&adjustment=all, and the full rows of observations are 25100-25111, with results in localstudy.py/paths/results.

A review of existing experiments 2605 and 2606, and CLI 2545, was conducted via a query of the research fund for crypto. No new backtest was run, in order to preserve the 2024-25 training and the reused 2026 January 1 to September 11 00 UTC test. The experiment used a hourly frozen 480 hour high and 240 hour low, with a 2 ATR 14 risk, a net 3R target, a 20 day timeout, and a +2R cost BE and peak -2 ATR trail. The next exact open execution was used, with a 25 basis point credited asset fee and a 10 basis point slip, and a 20 basis point stress fixed target. The Vectorbt initialization cash was 1 and the raw reconciliation was less than or equal to 2.45e-15, as reported in the original 48 cells. The ETH hourly training resulted in a 33.250114% return, with 19 trades, and the test resulted in a 17.240439% return, with a 19.830197% drawdown, a 66.6667% win rate, and 6 trades. The stress test resulted in a 15.841958% return. The daily 20 high and 10 low comparator training resulted in a 2.029153% return, with 15 trades, and the test resulted in a -17.632466% return, with a 27.353641% drawdown, a 20% win rate, and 5 trades. Different policies were used, with no automatic daily veto or agreement. The CLI gross return was 28.905427%, with a 24.943256% drawdown, a 21.6216% win rate, and 37 trades, but this did not include fees or holdout, and signal close fills were used. The current completed entry was false, and no new rule was traded. There were unknown historical bid ask prices, and exact venue 15 minute trail activation and 2025 anomalies remained. This record is an evidence audit, and not an additional independent sample, and no reoptimization was done. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars, and the original records are 2545, 2605, and 2606.

A post-audit diagnostic test was conducted on the SPY stock with a low price of 68.64 on February 2, 2026. This price was replaced with 678.301862, which is the same-date extended session low of 681.94 adjusted by a daily adjustment factor of 691.70/695.41. This change is not an authoritative correction or a new independent holdout. The original vendor data with 64 cells was retained and quarantined, with 2685 rules unchanged. The monthly rule remains the same, with the first open and prior variance capped at 1, 2 ATR stop, and 6 ATR target, plus 4 ATR quarter and next day break even, with a month-end exit. The test period was from January 3, 2023, to August 31, 2026, with fees of 1 basis point and a slip of 15 basis points per side, and a scale of 2. The return was 39.05693377174397% with a drawdown of 14.068000496405809%, and a win rate of 52.27272727272727%. The mean return was 0.8804737043513139%. The whole policy was executed on a 15-minute basis, but was unvalidated and did not generate any orders. The daily vector backtest match was 3.637978807091713e-12. The training data was unchanged because the outlier was only in 2026. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars and the research and audit results are in test_vm.py/results.json.

A post-audit review, known as a diagnostic check, was performed on the SPY stock, which is a type of investment fund that tracks the S&P 500 index. On February 2, 2026, the low price of SPY was initially reported as 68.64, but this was updated to 678.301862 in the extended session, which is a period of trading outside regular hours. The daily adjustment factor, a calculation used to adjust prices, was 691.70/695.41, and the 15-minute low was 681.94. This update is not an official correction, but rather a replacement of the original data. The original data from the vendor, which is the company providing the information, was kept, but 2685 cells, or pieces of data, were set aside for further review. The rules for the investment strategy, which include a monthly opening price and a stop-loss order, known as a 1.2 ATR stop, and a target price, known as a 6 ATR target, were not changed. The strategy also includes a quarterly adjustment of 4 ATR, or average true range, and a next-day break-even point. From January 2, 2019, to December 30, 2022, the fees were 1 basis point, or 0.01%, for each transaction, and 15 basis points, or 0.15%, for each slip, which is a difference between the expected and actual price of a trade. The scale, which refers to the size of the investment, was 2. The return on investment was 5.924249409058424%, and the maximum drawdown, or loss, was 22.950594911252875%. The win rate, or percentage of successful trades, was 45.83333333333333%, and the mean, or average, return was 0.620346423396773%. The entire policy, or investment strategy, was based on 15-minute execution, but this was not validated, and no orders were placed. The daily vector, a mathematical representation of the investment, matched the expected result with a very small difference of 3.637978807091713e-12. The training data, which is used to develop the investment strategy, was not changed because the only outlier, or unusual data point, was from 2026. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars, and the research and audit results are available in the file research/audit/test_vm.py/results.json.

A news scan found the headline 'Post-audit freeze 2249'. This refers to a review of automated trading decisions. On February 2, 2026, the trading fund manager made a decision about the SPY stock, which is a type of investment that tracks the S&P 500 index, a measure of US stock market performance. The manager replaced a low price of 68.64 with 678.301862, which is 64 times the low price of 681.94 from the same date, adjusted for daily changes. The original information from the vendor, which is the company providing the data, was kept, but some cells were set aside for further review, a process called quarantine. The rules for making trades did not change, and they still consider the first open price of the month and the difference from the previous month, with a stop-loss at 1 or 2 times the average true range, a measure of stock price volatility, and a target at 6 times the average true range. The fund also considers the next day's and quarter's performance. From January 3, 2023, to August 31, 2026, the fund paid fees of 1 basis point, a unit of measurement for interest rates and fees, and 15 basis points for each trade, with a scale of 0.5. The return on investment was 22.137105901782196 percent, with a maximum drawdown, a measure of the largest decline in value, of 6.200989910708743 percent. The fund had 44 trades, with a win rate of 52.27272727272727 percent and an average return of 0.8804737043513142 percent. The entire policy was based on 15-minute execution, but it was not validated, and no orders were placed. The daily vector, a mathematical representation of the trades, matched the expected result with a very small difference of 3.637978807091713e-12. The training data, which is used to make predictions, did not change because the only outlier, an unusual data point, was from 2026. The source of this information is https://runway.adityaperswal.workers.dev/data/stock-bars and research/audit/test_vm.py/results.json, which appears to be a data file and a research result.

A post-audit review, known as a diagnostic, was performed on a trading model called VM9110600. On February 2, 2026, the model's low price for the SPY stock was updated from 68.64 to 678.301862, which is the same-date extended session low of 681.94 adjusted by a daily adjustment factor of 691.70/695.41. This change is not an official correction and does not affect the model's independent holdout. The original data from the vendor, which includes 64 cells, was retained and quarantined, and rule 2685 was applied. The model's rules remain unchanged, with a monthly first open and prior variance capped at 1, and stop and target levels set at 2 and 6 ATR, respectively, with a quarter adjustment of +4 ATR and a next-day breakeven point. The model also has a month-end exit. From January 2, 2019, to December 30, 2022, the model had fees of 1 basis point and a slip of 15 basis points per side, with a scale of 0.5. The return was -4.314729648196081% with a drawdown of 9.48234794366084%, and 48 trades were made, with a win rate of 45.83333333333333% and a mean of 0.6203464233967744%. The 15-minute execution policy was not validated and no orders were placed. The daily vector backtest match was 3.637978807091713e-12. The training data was not changed because the only outlier was in 2026. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars and the research and audit results are in research/audit/test_vm.py/results.json.

A post-audit diagnostic test was conducted on February 2, 2026, for the SPY stock, which is a type of stock market index fund. The test replaced the low price of 68.64 with 678.301862, which is the same date extended session low of 681.94 adjusted by a daily adjustment factor of 691.70/695.41. This change is not an authoritative correction or a new independent holdout. The original vendor's 64 cells were retained and quarantined, and the rule remained unchanged. The rule is based on a monthly first open with a prior variance capped at 1, 2 ATR stop, and 6 ATR target, with a 4 ATR quarter and next day break even, and a month end exit. ATR stands for Average True Range, which is a measure of stock price volatility. The test period was from January 3, 2023, to August 31, 2026, with fees of 1 basis point and a slip of 15 basis points per side, and a scale of 1. The return was 34.1951933271146% with a drawdown of 10.5565031388005%, and 44 trades were made with a win rate of 52.27272727272727%. The mean return was 0.8804737043513127%. The whole policy was based on 15-minute execution, but it was not validated and no order was placed. The daily vector backtest match was 1.8189894035458565e-12. The training data was not changed because the outlier was only in 2026. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars and the research and audit results are in test_vm.py/results.json.

A post-audit review, called a freeze, was done on February 2, 2026, for the SPY stock, which is a type of investment that tracks the S&P 500 index. The low price of SPY on that day was 68.64, but it was replaced with 678.301862, which is the low price from the extended session, a period of trading outside regular hours, adjusted for daily changes. This change is not an official correction and does not affect the independent data used for testing. The original data from the vendor, which provides financial information, was kept, but some of it was set aside for further review. The rules for buying and selling stocks were not changed, including the monthly review of the first opening price and the limits on losses and gains, which are based on a measure of volatility called the Average True Range, or ATR. From January 2, 2019, to December 30, 2022, the fees for buying and selling were 1 basis point, or 0.01%, and 15 basis points for the difference between the expected and actual price, known as the slip. The results showed a return of 5.557496938788775% and a maximum loss, or drawdown, of 14.661439139633735%. The strategy was successful 45.83333333333333% of the time, with an average gain of 0.6203464233967758%. This analysis was based on 15-minute intervals, but it was not validated, meaning it was not checked against actual trades. The daily analysis matched the expected results, with a very small difference of 3.637978807091713e-12. The training data was not changed because the unusual result in 2026 was considered an outlier, or an exception. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars and the results can be found in research/audit/test_vm.py/results.json, which is a file containing the output of a computer program.

A post-audit diagnostic test, called a freeze, was done on February 2, 2026, for the SPY stock, which had a low price of 68.64. This was replaced with 678.301862, considering the same-day extended session low of 681.94 and a daily adjustment factor of 691.70/695.41. This change is not an official correction or a new independent holdout. The original data from the vendor, with 64 cells, was kept separate, and rule 2685 was applied. The monthly rule remains unchanged, with the first open and prior variance capped at 1 and 2 ATR stop, and 6 ATR target, plus 4 ATR quarter and next day break even, with a month-end exit. From January 3, 2023, to August 31, 2026, fees were 1 basis point and slip was 5 basis points per side, with a scale of 1. The return was 42.575665675294736 percent, with a drawdown of 9.526576307753443 percent, and 44 trades, with a win rate of 56.81818181818182 percent, and a mean of 1.0486536148037133 percent. The whole policy was executed every 15 minutes, but was not validated, and no order was placed. The daily vector backtest match was 3.637978807091713e-12. The training data was not changed because the only outlier was in 2026. The source of this data is https://runway.adityaperswal.workers.dev/data/stock-bars, and the research and audit results can be found in test_vm.py/results.json.

A post-audit review, known as a diagnostic, was performed on a trading model called VM9110600. On February 2, 2026, the model's low price for the SPY stock was updated from 68.64 to 678.30, which is 1862 times the daily adjustment factor of 691.70 divided by 695.41, and also considering the extended session low of 681.94 on the same date. This change is not an official correction and does not affect the model's independent holdout data. The original data from the vendor, which includes 64 cells, has been retained and quarantined, with 2685 rules remaining unchanged. The model's rules include a monthly first open, a stop loss of 1 to 2 times the average true range, or ATR, which is a measure of price volatility, and a target of 6 times the ATR, with a quarter adjustment and a next-day break even, or BE, point, and an exit at the end of the month. From January 2, 2019, to December 30, 2022, the model had fees of 1 basis point, or 0.01%, and a slip of 5 basis points, or 0.05%, per side, with a scale of 1. The return was 9.57%, with a drawdown of 14.34%, and 48 trades, with a win rate of 47.92%, and a mean of 0.75%. The entire policy was executed over 15 minutes, but this has not been validated, and no order was placed. The daily vector backtest match was 3.64e-12. The training data was not changed because the only outlier was in 2026. The source of this data is https://runway.adityaperswal.workers.dev/data/stock-bars and the research and audit results can be found in the test_vm.py/results.json file.

A test was done on the stock CNXC using a momentum strategy. The test used 20 days of past data to make decisions, over a period of 1460 days. The test aimed to stop losses at 5% and target gains at 15%. The results showed a return of -31.49699%, with a maximum loss of 38.87319%, and a win rate of 31.34328%. For comparison, simply buying and holding the stock would have resulted in a return of -79.10702%. The test did not include costs or independent verification. Another test, using a different method, also failed to validate the momentum strategy. As a result, the momentum control was not adopted. This decision was not based on FDA or earnings data alone. The data for this test came from a news scan that found the headline was not needed, instead, the source is a data line from https://runway.adityaperswal.workers.dev/data/stock-bars?feed=iex.

A news scan found the headline 'CNXC91106 RTH15min T5 +2R breakeven holdout 50bp/side'. This is a report about a trading strategy for the stock CNXC. The strategy is based on three premises. First, it uses the same time window and prior average daily range, with an entry at the open of the first 15-minute bar during regular trading hours, and a stop and target based on the average daily range. The strategy also forces an exit at the 15:45 open. A breakeven variant moves the stop to the entry point only after a 2R gain, with no quarter exit. Second, the strategy was tested on seven events from June 2025 to June 2026, using 130 regular trading hour bars from five actual daily sessions, for a total of 910 bars, plus 116 bars from other hours. All regular trading hour slots were covered. Third, the strategy uses a long-only approach with fees of 5 basis points, compounded return, and seven trades reconciled. The actual 15-minute opens differed from the daily opens by small amounts. All seven trades lost, and the 2R gain was never triggered, resulting in the same breakeven outcome. The conclusion is to reject this strategy. The daily return was -26.5783%, while the 15-minute return was -27.7235%, reflecting the entry vintage, check frequency, and 15:45 cutoff. The data source is https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=CNXC&timeframe=15Min&feed=sip&adjustment=all.

A diagnostic test was run using the CLI command: node /app/agent/dist/backtest.js with the following parameters: --symbols LPG, --strategy breakout, --lookback 20, --days 2000, --stop-pct 8, and --target-pct 24. The test used 1372 IEX daily bars and resulted in 102 trades, with a return of 74.13260%, a drawdown of 13.36461%, and a win percentage of 38.23529%. The buy and hold return was 306.80883%. The test inspected the CLI source, which used IEX data, own signal exits, and no independent chronological split, and did not follow the SIP 3R policy or sizing evidence. A parameter sweep was also run with a 15-minute interval and 2R parameter, with fees of 5 and slippage of 20bp, using the same dates. The results were: L10 train 65.78465% with 34 trades, marked test 33.48742% with 16 trades, and drawdowns of 45.67749% and 29.14655%; L40 train 30.23090% with 25 trades, marked test 28.31975% with 11 trades, and drawdowns of 38.38739% and 19.09668%. Each marked test included a 1-cutoff position. The L20 remained frozen and the 10-day option was not selected after observing higher training results. All measured figures were preserved from observations 24738 to 59. The data sources used were https://runway.adityaperswal.workers.dev/data/stock-bars?feed=iex and the same endpoint with feed=sip. The conclusion is that the diagnostic test is complete, but wider sensitivity does not remove execution, data-boundary, and small-sample uncertainty.

A news scan found the headline 'CNXC91106 RTH15min T5 +2R breakeven holdout 30bp/side'. The premises of a study are as follows: the study used the same event windows and prior average daily range, with a 15-minute first bar open entry, and checked the 09:30-15:45 bar opens with a 1 average daily range stop and 3 average daily range target, forcing a 15:45 open on the previous day. The breakeven variant moves the stop to the entry only after a 2R gain, with no quarter exit. The study examined 7 events from June 2025 to June 2026, each with 130 regular trading hour bars from 5 actual daily sessions, totaling 910 regular trading hour bars, and 1026 total bars including other hours. All regular trading hour slots were covered. The study also accounted for 26 non-trading bars in January and June 2025 due to calendar closures, not missing data. The study used vectorbt long only fees of 5 basis points and slips, with compounded return and 7 trades reconciled. The actual 15-minute opens differed from the daily opens by 0, 0, 0, 0, -0.07, -0.04, and +0.05 dollars. All 7 trades lost, and the +2R gain was never triggered, with the same breakeven outcome expected. The conclusion is to reject. The daily return of -26.5783% versus the 15-minute return of -27.7235% reflects the entry vintage, check frequency, and 15:45 cutoff jointly, not pure stop slippage. These are regular trading hour trade bar open proxies, not venue executable quotes, with overnight and limit fills unvalidated and not adopted. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=CNXC&timeframe=15Min&feed=sip&adjustment=all.

A news scan found the headline about a trading strategy test for CNXC91106. The test had three premises. First, it used the same event windows and prior average daily range, with a 15-minute first bar open entry, and checked the 09:30-15:45 bar opens with a stop at one average daily range and a target at three average daily ranges, forcing a close at the 15:45 open. A breakeven variant moved the stop to the entry only after a two-unit gain, with no quarter exit. Second, the test used seven events from June 2025 to June 2026, each with 130 regular trading hour bars from five actual daily sessions, totaling 910 regular trading hour bars, and 1026 total bars including other hours. All regular trading hour slots were covered. Third, the test used vectorbt long-only fees of five basis points, with compounded return and seven trades reconciled. The actual 15-minute opens differed from the daily opens by 0, 0, 0, 0, -0.07, -0.04, and +0.05 dollars. The test concluded that all seven trades lost, and the two-unit gain was never triggered, with the same breakeven outcome expected. The daily return was -26.5783%, and the 15-minute return was -27.7235%, reflecting the entry vintage, check frequency, and 15:45 cutoff jointly, not pure stop slippage. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=CNXC&timeframe=15Min&feed=sip&adjustment=all, which is a data line.

A news scan found the headline about CNXC91106 RTH15min T5 fixed exits holdout 30bp/side. The premises of this test are as follows: the test uses the same T5 event windows and prior Average Daily Range, or ADR, which is a measure of how much a stock's price moves on average each day. The test checks the 09:30–15:45 bar opens with a 1 ADR stop, which is a stop-loss order set at a price level, and a 3 ADR target, which is a profit target. The test also forces the T-1 15:45 open, where T-1 refers to the previous trading day. The breakeven variant moves the stop to the entry price only after a 2R gain, where R is a unit of measurement for the profit or loss of a trade, and there is no quarter exit. The test includes 7 events from June 2025 to June 2026, each with 130 Regular Trading Hours, or RTH, bars from five actual daily sessions. This totals 910 RTH bars, plus 116 bars from other hours, covering all RTH slots. The weekday count adds 26 non-trading bars in January and June 2025, which are calendar closures, not missing data. The test uses vectorbt long only fees of 5 basis points, or bp, which is a unit of measurement for interest rates or fees, and slips, which refers to the difference between the expected price of a trade and the actual price. The test reconciles 7 trades and finds that the actual 15-minute opens differ from the daily opens by 0, 0, 0, 0, -0.07, -0.04, and +0.05 dollars. The test keeps the vintages, or older data, and records the differences. All 7 trades lose, and the +2R gain is never triggered, resulting in the same breakeven outcome. The conclusion is to reject this test. The daily return of -26.5783% versus the 15-minute return of -27.7235% reflects the entry vintage, check frequency, and 15:45 cutoff jointly, not just pure stop slippage. These are RTH trade bar-open proxies, not venue executable quotes, and overnight or limit fills are unvalidated and not adopted. The source of this data is https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=CNXC&timeframe=15Min&feed=sip&adjustment=all.

This is a test of the VM9110600 GLD adapter, which is a tool used to connect and analyze data, specifically a type of financial data called a gold exchange-traded fund, or GLD. The test used a set of rules, called Protocol 2230, with some corrections made in 2233 and 2236. The test looked at data from January 3, 2023, to August 31, 2026, and used data from 2016 to 2018 to calibrate the system, and data from 2019 to 2022 to train it, with a holdout period from 2023 to August 26. The test entered the market monthly, with a limit on how much it could lose, and stopped trading when it lost a certain amount, which is measured in a unit called ATR, or average true range, and also had a target to make a certain amount, which is 6 times the ATR, and an extra 4 ATR for each quarter, with the goal of making a profit by the end of the month. The test also accounted for fees, which were 1 basis point, or 1/100th of a percent, per trade, and a slip of 15 basis points, and assumed that cash had a yield of 0. The results showed a return of 37.26 percent, with a maximum loss of 9.46 percent, and a win rate of 52.27 percent, over 44 months, with 44 trades. The test used a tool called vectorbt for daily reconciliation, which means checking the results every day, and found a maximum error of 3.637978807091713e-12, which is a very small number, with an initial investment of 10,000 units. The test did not validate the results over a 15-minute period or for whole shares, and used a dividend-adjusted proxy, which means it accounted for dividend payments, and did not touch the underlying data. The test was used as a diagnostic instrument, and did not promote a winning strategy. The data came from a source at https://runway.adityaperswal.workers.dev/data/stock-bars, and the results can be found in a file called research/test_vm.py/results.json, which is a type of file that stores data in a format called JSON.