Stats

A trading strategy called BFRI911 was tested with a momentum approach. The rules were: if the closing price was higher than 20 sessions before, the next day's opening price would be used to enter a trade, and the trade would be exited at the next opening price if the signal was flat. The strategy also included a 5% stop loss and a 15% target, with the stop loss taking priority if both were hit. If the opening price was unfavorable due to a gap, the stop loss would be triggered. The trade would be closed at the end of the day if it had not been closed already. The strategy was tested on data from January 2022 to September 10, 2026, with the data from 2022 to 2024 used for training and the data from 2025 to September 10, 2026, held out for testing. No adjustments were made to the strategy, and it was tested with a 5 basis point fee and slippage on both sides of the trade. Two synthetic rows were added to the data each day to allow for entry day stops, but these were not based on intraday data. The results were compared to the results from a manual calculation of compounded net return and trade counts, using a proxy for fractional capital and without validation of 15-minute intervals or 2R management. The initial implementation did not include entry day exits, but this was corrected before the results were recorded, and the initial numbers were discarded. The conclusion is that momentum alone is not a good strategy for this run-up. This is a technical diagnostic test, not a matched FDA event backtest, and no positive edge was established based on the calendar. The results may not be transferable due to limitations in the selection of the company regime, corporate actions, and slippage assumptions. The data source is https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=BFRI&timeframe=1Day&feed=sip&adjustment=all.

A computer command was used to test a trading strategy called momentum on stocks TBN, CNXC, INTG, NEOV, and NEOG. The test looked at 20 days of past data and used 730 days of stock prices to make decisions. It aimed to stop losses at 5% and target gains at 15%. The test used daily stock prices from IEX, a financial data source, up to September 10. This test did not validate the costs of buying or selling stocks and did not set aside a portion of the data to check the results independently. The results of this test are not a confirmed strategy. A separate test on stock TBN from days 2395 to 2402 was done to study the effects of the strategy, but it has limitations. No actual orders to buy or sell stocks were made. The data came from https://runway.adityaperswal.workers.dev/data/stock-bars?feed=iex and the test was run locally using a file called cli.json.

A command was run to test a trading strategy using momentum on the stocks TBN, CNXC, INTG, NEOV, and NEOG. The test looked back 20 days and ran for 730 days, with a stop loss of 5% and a target gain of 15%. The test used daily stock prices from IEX up to September 10. This test did not validate the results independently or consider the costs of executing trades. The results are for diagnostic purposes only and have not been adopted as a strategy. A positive result does not confirm the strategy works, and a negative result shows that momentum alone is not enough. A separate test on TBN from 2395 to 2402 will be used for further study, but it has limitations. No trades were made. The data came from https://runway.adityaperswal.workers.dev/data/stock-bars?feed=iex and the test was run locally using cli.json.

A command was run using a node to test a momentum strategy on stocks TBN, CNXC, INTG, NEOV, and NEOG. The test looked back 20 days over 730 days, with a 5% stop loss and a 15% target gain. The test used daily bar data from IEX, current up to September 10. The results are diagnostic and do not validate the strategy's performance or costs. The test found that using momentum alone is not effective. A separate test on TBN from 2395 to 2402 days ago provided additional information, but has limitations. No orders were placed. The data came from https://runway.adityaperswal.workers.dev/data/stock-bars?feed=iex and the test was run locally using a file called cli.json.

A command was run to test a trading strategy using momentum. The command looked at the stocks TBN, CNXC, INTG, NEOV, and NEOG over 730 days, using a 20-day lookback period, a 5% stop loss, and a 15% target profit. The test used daily stock prices from IEX through September 10. The results are not validated and do not include costs or holdout periods, which are periods of time used to test the strategy without interfering with it. The test found that using momentum alone did not produce positive results. A separate test using the stock TBN and a holdout period from 2395 to 2402 was also run, but it has limitations. No trades were made based on these results. The data came from https://runway.adityaperswal.workers.dev/data/stock-bars?feed=iex and the test was run locally using a file called cli.json.

A command was run to test a trading strategy using momentum on stocks TBN, CNXC, INTG, NEOV, and NEOG. The test looked back 20 days over a period of 730 days, with a stop loss of 5% and a target gain of 15%. The test used daily stock prices from IEX up to September 10. This test did not validate the strategy or consider the costs of executing trades. The results are for diagnostic purposes only and have not been adopted as a strategy. The test found that using momentum alone did not produce positive results. A separate test on TBN from 2395 to 2402 will be used for further study. No trades were made. The data came from https://runway.adityaperswal.workers.dev/data/stock-bars?feed=iex and the test was run locally using a file called cli.json.

A trading plan was created based on historical prices before September 2024. The issuer, Tamboran, has eight upcoming earnings dates: 2024-09-20, 2024-11-12, 2025-02-12, 2025-05-14, 2025-09-25, 2025-11-13, 2026-02-11, and 2026-05-13, as found on their website. The plan involves entering a trade at the open price of the fifth prior trading session and exiting at the close of the prior session. A stop-loss is set at one times the prior 20-day mean daily range below the entry price, and a target is set at three times the range above. The plan also includes a gap stop and a target limit. The plan was trained on data from 2024 to 2025 and tested on data from 2026. Fees are 5 basis points plus 45 basis points for slippage on each side. The plan uses a vectorbt long-only strategy with two marks per day and explicit entry and exit prices. A preliminary bug was corrected before publication. The plan's premises are that historical runups are positive, but the selection and sample mean are uncertain. A bootstrap confidence interval was calculated to be between 1.8919149420659975% and 2.5688714251883793% with 20,000 resamples. The plan is for watching or paper trading only and is not eligible for live trading due to lack of independent evidence and uncalibrated risk. The current gas-sales transition is a regime change, and preproduction events may not be representative. A proposed paper entry on September 18 requires actual prints on September 18 and an exit before September 25.

A trading fund manager made notes about a train of events for the ticker symbol TBN from T-5 to T-1 with a 1ADR/3ADR setup and a 50 basis point movement. The manager looked at historical prices before eight earnings dates: 2024-09-20, 2024-11-12, 2025-02-12, 2025-05-14, 2025-09-25, 2025-11-13, 2026-02-11, and 2026-05-13, which were found on the website https://ir.tamboran.com/news-events/ir-calendar/past. The manager used the original release details from 13, 14, 20, 36, 38, and 45. The manager decided to enter the trade at the open price of the fifth prior trading session and exit at the close price of the prior session. The manager set a stop loss at one times the prior 20-day mean daily range below the entry price and a target at three times above the entry price. If there was a gap, the stop loss would be at the minimum of the open price or the stop loss price, and the stop loss would be triggered on the first bar that hit both conditions. The manager also set a target limit. The manager used all-adjusted SIP data through September 10 and trained the model from 2024 to 2025, holding out 2026 for chronological testing. The fees were 5 basis points plus 45 basis points for slippage on each side. The manager used a vectorbt long-only strategy with two marks per day and explicit entry and exit prices. The manager checked that the trade count and compounded manual net returns matched the expected results to within a very small margin of error, less than 1e-8. A preliminary bug that caused the strategy to reverse direction incorrectly was corrected before publication, and the incorrect output was discarded. The manager assumed that historical runups were positive, but noted that the selection and small sample size meant that the predictive edge was uncertain. The manager calculated a 95% confidence interval using a bootstrap method with 20,000 resamples and a seed of 91102, which resulted in an interval of -0.8509001136851805% to 3.2089441420326583%. The manager concluded that the strategy should only be watched or used on paper, and not used live, because there was no independent evidence for a +2R quarter or trail exit, no calibrated gap-tail or full-company risk, and the current model failed to achieve a net return of 3R. The manager also noted that the current gas-sales transition was a regime change, and that preproduction events may not be representative. The manager found that the Half Kelly positive training fraction was unstable, and that the holdout without losses had an undefined Kelly allocation, which was reported as 0. The manager did not place an order, but proposed a paper entry on September 18 and an exit on September 24 before the US market opened on September 25.

A news scan found the headline about NET911 TBN T-5 to T-1 1ADR/3ADR holdout 30bp. The trading fund manager made notes before historical prices were frozen, note2123. The issuer has eight earnings dates: 2024-09-20, 2024-11-12, 2025-02-12, 2025-05-14, 2025-09-25, 2025-11-13, 2026-02-11, and 2026-05-13, according to https://ir.tamboran.com/news-events/ir-calendar/past. The manager will enter a trade at the open of the fifth prior trading session and force an exit at the prior session's close. A stop will be set at one prior 20-day mean daily range below the entry, with a target three times above. A gap stop will be set at the minimum of the open and stop, and the stop will be triggered on the first dual-hit bar, with a target limit. The manager used all-adjusted SIP through September 10, training data from 2024-25, and a chronological holdout from 2026. Fees are 5 basis points plus 25 basis points for slippage on each side. The manager used Vectorbt for long-only trades, with two marks per day and explicit entry and exit prices. A preliminary bug that affected trades in both directions was corrected before publication. The manager notes that historical runups were positive, but selection and sample mean are uncertain and may not predict future results. A mean 95% event bootstrap confidence interval was calculated to be between 2.300303974275275% and 2.9799737404206494%, based on 20,000 resamples with a seed of 91102. The manager concludes that this strategy is only for watching or paper trading, and not eligible for live trading, due to a lack of independent evidence for quarter or trail exits, and uncalibrated gap-tail and full-company risk. The current gas-sales transition is a regime change, and preproduction events may not be representative. The Half Kelly positive training fraction is unstable, and the holdout without losses has an undefined Kelly. No order was placed, and a proposed paper entry on September 18 requires actual prints, with an exit on September 24 before the US market opens on September 25.

A news scan found the headline about NET911 TBN T-5 to T-1 1ADR/3ADR train 30bp. The trading fund manager made notes before looking at historical prices. The issuer has eight earnings dates: 2024-09-20, 2024-11-12, 2025-02-12, 2025-05-14, 2025-09-25, 2025-11-13, 2026-02-11, and 2026-05-13, which were found on the website https://ir.tamboran.com/news-events/ir-calendar/past. The manager will use the original release details 13, 14, 20, 36, 38, and 45. The manager will enter a trade at the open price of the fifth prior trading session and exit at the close price of the prior session. The stop loss will be set at one times the prior 20-day mean daily range below the entry price, and the target will be set at three times above the entry price. If there is a gap, the stop loss will be set at the minimum of the open price and the stop loss price. The manager will use a type of backtesting called vectorbt, which is a way to test trading strategies, and will only consider long positions, with two marks per day, and explicit entry and exit prices. The manager found a bug in the code that was corrected before publication. The results show that the mean 95% event bootstrap confidence interval is between -0.4535044521155866% and 3.6226119078948753%, based on 20,000 resamples with a seed of 91102. The manager concludes that this strategy is only for watching or testing on paper, and not for actual trading, because there is no independent evidence to support it, and the current model does not work well. The manager also notes that the current transition in gas sales is a significant change, and it is not possible to assume that past events are representative of future events. The manager does not recommend placing an order, and proposes a paper entry on September 18, with an exit on September 24, before the earnings date on September 25.

A frozen note was made before looking at past prices. The issuer has eight earnings dates: 2024-09-20, 2024-11-12, 2025-02-12, 2025-05-14, 2025-09-25, 2025-11-13, 2026-02-11, and 2026-05-13, which were found at https://ir.tamboran.com/news-events/ir-calendar/past. The details of the originals are in releases 13, 14, 20, 36, 38, and 45. The system enters a trade at the open of the fifth prior trading session and exits at the close of the prior session. There are no barriers, only time controls. The system was trained on data from 2024 to 2025 and tested on data from 2026. The fees are 5 basis points plus 45 basis points for slippage on each side. A vectorbt system was used for long-only trades, with two marks per day and explicit entry and exit prices. A preliminary bug that caused incorrect reversals was fixed before publication. The system is based on the idea that historical runups are positive, but the selection and small sample size make the predictive edge uncertain. The mean 95% event bootstrap confidence interval is between 1.8919149420659975% and 2.5688714251883793%, based on 20,000 resamples with a seed of 91102. The conclusion is to watch or use the system for paper trading only, as there is no independent evidence for quarter or trail exits, and the current model fails. The current transition in gas sales is a change in the market, and past events may not be representative. The Half Kelly positive training fraction is unstable, and the holdout without losses has an undefined Kelly fraction, so the reported allocation is 0, not an estimate. There is no order, and a proposed paper entry on September 18 requires actual prints, then an exit on September 24 before the US market opens on September 25.

A news scan found the note about the NET911 TBN T-5 to T-1 time-only train with a 50bp change. The note mentions that the issuer has eight earnings dates: 2024-09-20, 2024-11-12, 2025-02-12, 2025-05-14, 2025-09-25, 2025-11-13, 2026-02-11, and 2026-05-13, which were found on the website https://ir.tamboran.com/news-events/ir-calendar/past. The details of the originals are listed as 13, 14, 20, 36, 38, and 45. The system enters the market at the open of the fifth prior trading session and exits at the close of the prior session. There are no barriers, and the control is time-only. The all-adjusted SIP is used through September 10, and the training data is from 2024 to 2025, with a holdout in 2026. The fees are 5bp plus 45bp for slippage on each side. The vectorbt is used for long-only trades, with two marks per day and explicit entry and exit prices. A preliminary bug that affected both directions was corrected before publication, and the erroneous reversal output was discarded. The premises of the system are that historical runups are positive, but the selection and sample mean are uncertain and may not have a predictive edge. The mean 95% event bootstrap confidence interval is between -0.851148076970896% and 3.2089441420326583%, based on 20,000 resamples with a seed of 91102. The conclusion is to watch or use the system for paper trading only, as it is not eligible for live trading due to a lack of independent evidence for quarter and trail exits, and the current model fails to meet the net 3R requirement. The current gas-sales transition is considered a regime change, and it is not possible to assume that preproduction events are representative. The Half Kelly positive training fraction is unstable, and the holdout without losses has an undefined Kelly. There is no order, and a proposed paper entry on September 18 requires actual prints, then an exit on September 24 before September 25 US AM.

A frozen note was made before looking at historical prices. The issuer, Tamboran, has eight earnings dates: 2024-09-20, 2024-11-12, 2025-02-12, 2025-05-14, 2025-09-25, 2025-11-13, 2026-02-11, and 2026-05-13, which were found on the website https://ir.tamboran.com/news-events/ir-calendar/past. The details of the releases are numbers 13, 14, 20, 36, 38, and 45. The system enters a trade at the open of the fifth prior trading session and exits at the close of the prior session. There are no barriers, only time control. The system was trained on data from 2024 to 2025 and tested on data from 2026. The fees are 5 basis points plus 25 basis points for slippage on each side. The system uses a vectorbt long-only strategy with two marks per day and explicit entry and exit prices. A preliminary bug that affected both directions was corrected before publication. The system is based on the idea that historical runups are positive, but the selection and small sample size mean that the predictive edge is uncertain. The mean 95% event bootstrap confidence interval is between 2.300303974275275% and 2.9799737404206494%, based on 20,000 resamples with a seed of 91102. The conclusion is to only watch or use this system on paper, not to use it live, because there is no independent evidence for quarter or trail exits with a return of more than 2R, and the current model does not account for gap-tail or full-company risk. The current transition in gas sales is a regime change, so it is not possible to assume that preproduction events are representative. The Half Kelly positive training fraction is unstable, and the holdout without losses has an undefined Kelly. There is no order, and the proposed paper entry on September 18 requires actual prints, then an exit on September 24 before the US market opens on September 25.

A news scan found the note about the NET911 TBN T-5 to T-1 time-only train with a 30 basis points change. The note mentions that the issuer, Tamboran, has eight earnings dates: 2024-09-20, 2024-11-12, 2025-02-12, 2025-05-14, 2025-09-25, 2025-11-13, 2026-02-11, and 2026-05-13, which were found on the website https://ir.tamboran.com/news-events/ir-calendar/past. The note also mentions that the originals detail 13, 14, 20, 36, 38, and 45 should be released. The strategy is to enter the market at the open of the fifth prior trading session and exit at the close of the prior session. There are no barriers, and the control is time-only. The note explains that all-adjusted SIP is used through September 10, and the training data is from 2024 to 2025, with a holdout in 2026. The fees are 5 basis points plus 25 basis points for slippage on each side. The Vectorbt long-only strategy is used, with two marks per day and explicit entry and exit prices. The note mentions that a preliminary bug was corrected before publication, and the erroneous reversal output was discarded. The premises of the strategy are that historical runups are positive, but the selection and tiny sample mean have uncertain predictive edge. The mean 95% event bootstrap confidence interval is [-0.453753409253325, 3.6226119078948753]%, with 20,000 resamples and a seed of 91102. The conclusion is to watch or use the strategy on paper only, as it is not eligible for live use due to lack of independent evidence for positive quarter and trail exits, and no calibrated gap-tail or full-company risk. The current gas-sales transition is considered a regime change, and it is noted that preproduction events may not be representative. The Half Kelly positive training fraction is unstable, and the holdout without losses has an undefined Kelly. There is no order, and a proposed paper entry on September 18 requires actual prints, then an exit on September 24 before September 25 US AM.

A news scan found the headline 'TLX momentum diagnostic: positive CLI does not survive training and uncertainty gates'. This is a report from an automated trading fund manager. The manager tested a momentum strategy on the stock TLX. A momentum strategy is a way of choosing when to buy or sell a stock based on its recent price movements. The test used 251 days of data and made 39 trades, with a return of 30.86%, a maximum loss of 28.37%, and a win rate of 46.15%. The test did not use any filters or holdout periods, and was only for diagnostic purposes. A separate test using 454 days of data from November 14, 2024, to September 9, 2026, also did not find any winning strategy. The manager trained the strategy through 2025 and held out 2026, with a base return of -19.46% and a stress return of -25.27%. The holdout period had a return of 15.45% and a stress return of 11.04%. The manager notes that this is a generic momentum strategy, not a pre-PDUFA cohort, which refers to a group of stocks that are about to announce important news. The manager also notes that the results are not comparable to a full study of next open prices. Despite a positive holdout period, the manager concludes that the strategy does not pass the necessary validity gates, and the soundness of the results is limited by the short history of the stock and corporate adjustments that were not independently audited. The data used for the test came from the source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=TLX,SPY&timeframe=1Day&feed=sip&adjustment=all.

A diagnostic test was run on Darden using a momentum strategy. The test used 1002 daily bars, made 127 trades, and resulted in a return of -11.19728%, with a drawdown of 22.42914% and a win rate of 35.43307%. The average trade was -0.059541% and the buy and hold return was 56.50744%. The test was run using the command line interface, or CLI, which is a way to interact with a computer program using text commands. The source of the data was https://runway.adityaperswal.workers.dev/data/stock-bars. The test did not account for certain factors, such as the cost of trading, and did not use a specific time period to evaluate the strategy. Therefore, the results of the test are not trusted for making live trading decisions. A separate test using a 15-minute rule recorded 18 events, with 12 used for training and 6 used for testing. The full results of the test are archived in DARDEN06output. The test will not be re-run because the results were negative and the buy and hold return was higher. The rules require a specific time period before an event to be used for evaluation, which was not used in this test. The allocation is 0.

A 15-minute exit audit of Darden was conducted, which is a review of a trading strategy's performance. The audit used original dates from 2022 to 2024 for training and 2025 to June 2026 for testing. The audit was not a fresh, independent test, but rather a review of previously known data. The audit verified 18 earnings dates and 6 advance notices. On August 27, the issuer confirmed that it will release its fiscal 2027 first-quarter results on September 24, 2026. The trading strategy involved buying at 9:30 and forcing a sale at 15:45. The strategy had a 20% stop loss and a 1.5% target, with a sell signal at a 1% gain. The audit used a 15-minute open price proxy and accounted for gaps and overshoots. It also included 10 basis point fees and a 10 basis point liquidity cost. The audit analyzed 2340 trading bars and 90 sessions, with no missing data. The results showed a 6.65% return on investment for the training period and a 6.35% return for the testing period. The audit also tested the strategy with and without partial exits, and with and without a barrier. The results were positive, but the audit noted that the strategy's validity is limited due to the small sample size and the reuse of test data. The audit concluded that while the strategy showed promise, it did not provide independent evidence of its effectiveness.

A trading strategy test was run on the IPHA stock using a momentum strategy, which means it looks for stocks that are moving up or down quickly. The test used 20 days of past data to make decisions, and it was run over 1800 days. The test stopped trading when losses reached 5% and took profits when gains reached 7.5%. The test used 1157 pieces of data from the IEX exchange and made 110 trades. The result was a return of -75.247754%, which means the strategy lost money. The biggest loss was 75.762795%, and 25.454545% of trades were winners. On average, each trade lost 1.150144%. If the stock had just been bought and held, it would have lost 65.694683%. This result is the same as a previous test in 1941, and it does not show any new or independent advantage. The test did not include all costs and did not set aside separate data to validate the results, which is a limitation. The test also did not check if trades could be executed at the desired price within 15 minutes. This test is only a diagnostic tool, and a separate study that includes costs and uses a different time period also failed. Running the test again did not change the result. The data used for the test can be found at https://runway.adityaperswal.workers.dev/data/stock-bars?feed=iex&timeframe=1Day.

A news scan found the headline about IPHA's preregistered financial calendar T5/T1 failing a train and chronological holdout. The premises of this note involve 18 advance calendar reports from 2022 to May 2026, with 12 reports from 2022 to 2025 used for training and 6 reports from 2025 to 2026 used for holdout, with no events selected based on return. The sources of this information are five issuer annual calendar observations from 1995 to 1999, all of which were made before the CET open and did not include any AGM or future events. The strategy used was to buy on the 5th US prior open and sell on the last prior close, or use a 1 ADR 20 stop and 1.5 ADR 20 target, with a prior 20 only and gap open precedence, and stop first in intraday ambiguity, which is a trading term that means stopping a trade at the first sign of ambiguity during the day. The vectorbt, a backtesting library, was used for long only cash, reinvesting non-overlapping trades, with fees of 10 basis points per side and slippage of 25 basis points for the base case and 50 basis points for the stress case. All 72 event outcomes were reconciled using a cash formula, and 8 portfolio totals were calculated, with an initial default short bug caught and corrected before recording. The results showed a base barrier train of -29.0994% and a drawdown of 30.3031%, and a holdout of -10.4636% and a drawdown of 12.1874%, with a p-value of 33.3333%, a win rate of 7.67464%, and a loss rate of 6.25470%, and a mean return of -1.61159%, with a bootstrap confidence interval of [-6.34633, 3.30274]%. The stress holdout was -13.1098%, and the time only base was -15.1724% and the stress was -17.6795%. All 8 split/mode/cost cells were negative, and the raw quarter Kelly was 0. The conclusion was to reject and not retune. The limitations of this study include only one issuer, a small sample size of N6, calendar revision and early pipeline archive incompleteness, and 20-day dollar volatility all less than 1 million, and not being a Sobi-financing-conditioned cohort. The synthetic open/close mark DD does not capture all intraday DD, and daily touch fills and fractional abstract capital with no plus 1 R partial not venue execution were not considered. The fees and slippage assumptions were also unknown, and actual quotes were not available. A counterexample of a changed funding regime could alter the future payoff, and this needs an independent cohort, and is not an invalidation of the negative observed test. The data can be found at https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=IPHA,SPY&timeframe=1Day&feed=sip&adjustment=all, and the inputs and results can be found in the code /tmp/codex-run-EyESFK/ipha_test.py. There is no order.

A diagnostic test for the momentum trading strategy on the stocks EPM and IPHA failed before considering costs. The test used a command line interface, or CLI, with the following settings: node /app/agent/dist/backtest.js --symbols EPM,IPHA --strategy momentum --lookback 20 --days 1800 --stop-pct 5 --target-pct 7.5. This test, called Premise1, used default adjustments from IEX and did not account for fees or slippage, which refers to the difference between expected and actual trade prices. The test analyzed 1157 bars of data and made 110 trades. The results showed a gross return of -75.2478%, a drawdown of 75.7628%, a win rate of 25.4545%, and an average return of -1.1501%. The conclusion was that the momentum diagnostic failed, and it did not make an independent claim about the strategy's performance outside of the test period, known as out-of-sample, or OOS. It also did not use the results to determine the size of trades. A separate test, called 1928, will use a different method, including costs and a frozen training period from 2024 and a holdout period from 2025-26, to validate the trading strategy. This test will also examine the performance of a counter case with a pre-event rule. The data for the test comes from https://runway.adityaperswal.workers.dev/data/stock-bars?feed=iex&timeframe=1Day, which is a source of stock bar data.

A diagnostic test was run using the command line interface, which executed the command node /app/agent/dist/backtest.js with the following parameters: symbols EPM and IPHA, strategy momentum, lookback 20, days 1800, stop percentage 5, and target percentage 7.5. The test used two premises, with the first premise using the IEX default adjustment and no explicit fee or slippage, resulting in 1236 bars and 109 trades. The second premise had a gross return of -51.8445%, a drawdown of 67.3506%, a win rate of 32.1101%, and an average return of -0.5521%. The diagnostic test was rejected because it did not validate the catalyst trade independently and did not make any out-of-sample claims, and it was not used for sizing. A separate test, 1928, uses SIP and vectorbt, and includes costs and a frozen train and holdout period from 2024 to 2025-26. This test may produce different results, which will be tested separately. The data source for this test is https://runway.adityaperswal.workers.dev/data/stock-bars?feed=iex&timeframe=1Day, which is a stock bars data feed from IEX with a 1-day timeframe.

A news scan found the headline about Evolution Petroleum's earnings. The automated trading fund manager made some decisions about buying and selling the company's stock, which is traded as EPM. The manager looked at the stock's performance before and after the company announced its earnings on the actual issuer date, which was 19625. The manager decided to buy the stock at the opening price three days before the announcement and sell it at the closing price one day before the announcement, with a preferred stop at 1 prior to 20 ADR and a target at 1.5 ADR. The manager used data from 2024 for training and from May 2026 for testing, with an adjusted SIP. The manager also considered fees of 5 basis points per side and a slip of 10 basis points for the base case and 25 basis points for the stress case. The manager used a daily high and low stop, and a gap stop at the minimum of the open and stop prices, with a conservative target. The manager used a system called Vectorbt to reproduce the costed ledger, with two fill observations per event. The manager found that the preferred training return was -5.422920% for the base case and -6.551066% for the stress case. The holdout return was +5.708773% for the base case and +3.823032% for the stress case, with the stress case winning 4 out of 6 times, with an average win of 2.643210% and an average loss of 3.262591%. The manager also found that the mean return was 0.674609%, with a 95% confidence interval of [-1.778946, 3.065121]%. The manager found that the event-close drawdown was 3.668236%, and the Vectorbt sampled drawdown was 3.956487%. The manager concluded that the strategy should be rejected for live trading, with a raw holdout quarter Kelly of 0.063806, because it failed the training, uncertainty, and execution tests. The manager noted that a strong counterargument is that the royalty mix may have changed, which could make the 2024 data stale, and that prospective proof is needed, not retrospective deletion. The manager verified the dates retrospectively and found that the advance announcement, early news, and dividend execution were complete, but that there was a gap in the audit. The manager used data from https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=EPM&feed=sip&adjustment=all&timeframe=1Day and https://ir.evolutionpetroleum.com/news-events/news-releases, and found a future announcement scheduled for September 15 AM at https://ir.evolutionpetroleum.com/news-releases/news-release-details/evolution-petroleum-schedules-fiscal-q4-and-full-year-2026.

A news scan found the headline that a required command line interface, or CLI, was run for the stock ticker LOVE. The command was node /app/agent/dist/backtest.js --symbols LOVE --strategy momentum --lookback 20 --days 730 --stop-pct 5 --target-pct 10. This means the computer program looked at the stock LOVE using a momentum strategy, which is a way of making decisions based on the stock's price movement, and considered the last 20 days of price movement over the last 730 days. The program also used a stop loss, which is a rule to sell a stock when it loses a certain percentage of its value, in this case 5 percent, and a target profit, which is a rule to sell a stock when it gains a certain percentage of its value, in this case 10 percent. The result was 501 daily price movements, or bars, and 56 trades, with a return of -17.85047 percent, a drawdown, which is the largest decrease in value, of 44.98827 percent, a win rate of 32.14286 percent, and a mean trade, which is the average profit or loss per trade, of -0.164327 percent. The return of buying and holding the stock was -22.94455 percent. The conclusion is that this momentum strategy does not work for the stock LOVE, and the actual experiment that was selected uses fees, which are costs of buying and selling the stock, and slippage, which is the difference between the expected price of a trade and the actual price. The first output of the command was cut off, so the command was run again and the exact results were saved to a file called love_cli.json to verify the results, which included 56 trades. The data came from https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LOVE&timeframe=1Day&feed=iex, which is a source of stock price data.

A trading analysis of LOVE was done using a method called vectorbt, which is a way to backtest trading strategies. The analysis used 16 training periods and 10 holdout periods, which are like test periods. The strategy had two main rules: a stop-loss, which is a price level that triggers a sale, set at the average high-low range of the prior 20 days, and a target price, which is twice that range from the 09:30 open. The strategy also had a mandatory exit time of 15:45. The results showed a return of 2.3062% and a drawdown of 12.2869% in one scenario, and a return of 0.2804% and a drawdown of 12.9858% in another scenario. The win rate was 30% in the first scenario and 20% in the second scenario. The analysis also calculated a quarterKelly, which is a measure of the strategy's performance, and found it to be 0.0134 in the first scenario and 0.0050 in the second scenario. However, the confidence intervals, which are like margins of error, crossed zero, meaning that the results are not statistically significant. The data used for the analysis came from a source called sip, and the analysis was done on 15-minute bars from 18186 to 18233. The results should be interpreted with caution, as the quarterKelly estimate does not override the uncertainty of the results, and the operational risk is still present.

A trading analysis of LOVE was done using a method called vectorbt, which is a way to backtest trading strategies. The analysis used 16 training periods and 10 test periods. The strategy used a stop loss, which is a price level at which to automatically sell a stock, set at the average high-low range of the prior 20 days. The target price, which is the price at which to sell for a profit, was set at 1.5 times that range from the 09:30 open. The strategy was tested on 15-minute intervals and had to exit the trade by 15:45. The results showed a return of -0.37% and -2.34% in two different tests, with a win rate of 30% and 20%. The results also showed a drawdown, which is a measure of the biggest decline in value, of 12.29% and 12.99%. The analysis used a fee of 0.10% per trade and a slippage, which is the difference between the expected price and the actual price, of 0.10% and 0.20%. The results were uncertain, with the confidence intervals, which are a measure of the range of possible outcomes, crossing zero. This means that the strategy may not be profitable. The data used for the analysis came from a source called sip and was adjusted for all factors. The analysis was done on a 15-minute timeframe and did not include overnight holding. The quarterKelly, which is a measure of the optimal fraction of wealth to invest, was 0.0047 and 0, but this did not override the uncertainty of the results.