Stats

A trading strategy for Ethereum (ETH) was tested using a system called Vectorbt. The strategy was to buy when the closing price exceeded the highest closing price of the prior 20 calendar days, and sell when this was no longer true, using the exact opening price. The test used hourly data from 480 bars, trained from 2024 to 2025, with the SOL period starting on August 26, and a cash reset on January 1. The test also included a holdout period from 2026 through September 12. The strategy included fees of 25 basis points on each side, credited to the asset equivalent buy fee of 0.0025/0.9975, and slippage of 10 basis points/30 basis points stress. The daily open, high, low, and close (OHLC) data was used from 24 complete hours, with missing calendar windows remaining unavailable and no filling. The baseline train return was -69.973193%, with a drawdown of 70.504557%, 197 trades, and a win rate of 17.766%. The holdout return was -20.782205%, with a drawdown of 25.913218%, 45 trades, and a win rate of 13.333%. The stress test result was -33.831787%. A high threshold control holdout resulted in -11.419497% with 21 trades. All 15/20/30 day close/high cost cells recorded 26572-83, with a known holdout range of -38.410 to -9.158%, but none were selected. The premise was that same-close gross gains are not evidence for a costed next-open rule. A counterexample showed that daily SOL remained positive, but hourly disagreement and missing required 3R/partial/15min policy prevented adoption. This short-hold toggle is not a prior 10 low exit Donchian or managed ETH pullback. The conclusion was to reject this strategy as a live replacement. The data sources were https://runway.adityaperswal.workers.dev/data/crypto-bars and https://docs.alpaca.markets/us/docs/crypto-fees.

A news scan found the text describing the C91300 ETH daily costed close-screen control and HIGH sensitivity. This is about an automated trading system. Vectorbt is a tool used for backtesting, which means testing a trading strategy on past data. The system goes long, or buys, when the closing price exceeds the highest closing price of the prior 20 calendar days, and exits when this is no longer true, following the exact opening price. It uses hourly data from 480 bars. The system was trained on data from 2024 to 2025, with the SOL period starting on August 26, and the cash reset on January 1. The known holdout period is from 2026 through September 12. There is no claim that this is an untouched new out-of-sample, or OOS, test. Fees of 25 basis points, or 0.25%, are credited to the asset equivalent buy fee of 0.0025/0.9975, and slippage of 10 basis points/30 basis points is stressed. Daily open, high, low, and close, or OHLC, data is from 24 complete hours. Missing calendar windows remain unavailable, with no filling. The baseline train return is 5.185949%, the drawdown is 31.646448%, there are 41 trades, and the win rate is 29.268%. The holdout return is -21.109976%, the drawdown is 29.758847%, there are 13 trades, and the win rate is 15.385%, with an open rate of 1. The stress test result is -25.256777%. The HIGH threshold control holdout result is -16.281467% with 8 trades. All 15/20/30 day close and high cost cells recorded from 26572 to 83, with a known holdout range of -25.257% to -7.238%, were considered, but none were selected. The premise is that same-close gross cumulative log returns, or CLI, gains are not evidence for a costed next-open rule. A counterexample is that daily SOL remains positive, but hourly disagreement and missing required 3R/partial/15 minute policy prevent adoption. This short-hold toggle is not a prior 10 low exit Donchian or managed ETH pullback. The conclusion is to reject this as a live replacement. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars and https://docs.alpaca.markets/us/docs/crypto-fees.

A news scan found the headline about a trading strategy for Bitcoin, which is based on a set of rules. The strategy uses a concept called a'vector' to decide when to buy or sell Bitcoin. It goes long, meaning it buys, when the closing price exceeds the highest closing price of the prior 20 calendar days, and exits when this condition is no longer true, following the exact opening price of the next hour. This strategy is tested using 480 hours of data from 2024 to 2025, with the training period starting on August 26 and a cash reset on January 1. The results from 2026 are held out and not used to train the strategy, up to September 12. The strategy includes fees of 25 basis points, or 0.25%, on each side, credited to the asset equivalent buy fee of 0.0025 and 0.9975, and slippage of 10 basis points and 30 basis points for stress testing. The daily high, low, open, and close prices are used from 24 complete hours, with missing calendar windows remaining unavailable and not filled. The baseline train return is -78.978048%, the drawdown is 78.978048%, with 208 trades and a win rate of 14.423%. The holdout return is -25.349461%, the drawdown is 26.444395%, with 57 trades and a win rate of 12.281%, and an open rate of 0. The stress test return is -40.568953%. A high threshold control holdout return is -12.189031% with 28 trades. All 15, 20, and 30 day close and high cost cells recorded from 26572 to 83, with a known holdout range of -40.569% to -8.841%, were considered, but none were selected. The idea that same-close gross gains are not evidence for a costed next-open rule is the premise. A counterexample is the daily SOL, which remains positive, but hourly disagreement and missing required 3R, partial, and 15-minute policy prevent adoption. This short-hold toggle strategy is not a prior 10 low exit Donchian or managed ETH pullback. The conclusion is to reject this strategy as a live replacement. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars and https://docs.alpaca.markets/us/docs/crypto-fees.

A news scan found the text describing a trading strategy for Bitcoin (BTC). The strategy uses a system called Vectorbt, which buys when the closing price exceeds the highest closing price of the prior 20 calendar days and sells when this condition is no longer true, using the exact opening price for the sale. This strategy was tested, or trained, from 2024 to 2025, starting on August 26, with the cash balance reset on January 1, and the results from 2026 up to September 12 were used to evaluate the strategy without interfering with it. The strategy includes fees of 25 basis points, or 0.25%, on each buy and sell, with a slippage of 10 basis points for buying and 30 basis points for selling. The daily high, low, open, and close prices were used from 24 complete hours of data, without filling in missing calendar days. The baseline training return was -2.113871%, with a drawdown of 26.189310%, 45 trades, and a win rate of 28.889%. The holdout return was -10.194852%, with a drawdown of 17.371016%, 18 trades, and a win rate of 22.222%, starting with no open positions. The stress test return was -16.433561%. A high threshold control had a holdout return of 5.123294% with 8 trades. All 15, 20, and 30-day close and high price cells were recorded from 26572 to 83, with a known holdout range of -17.855% to 5.689%, but none were selected. The premise is that same-close gross gains are not evidence for a costed next-open rule. A counterexample is the daily SOL, which remains positive, but hourly disagreement and missing required 3R/partial/15-minute policy prevent adoption. This strategy is not a prior 10 low exit Donchian or a managed ETH pullback. The conclusion is to reject this strategy as a live replacement. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars and https://docs.alpaca.markets/us/docs/crypto-fees.

A diagnostic and parameter stability test was run on the CNQ stock using a command line interface, or CLI, with the following parameters: the node was set to /app/agent/dist/backtest.js, the symbol was CNQ, the strategy was breakout, the lookback period was 20 days, the number of days was 2100, the stop percentage was 6, and the target percentage was 18. The test yielded 1439 daily bars from the IEX exchange, 106 trades, a return of 23.1053%, a drawdown of 21.2717%, and a win percentage of 39.6226%. For comparison, a buy and hold strategy would have returned 104.1599%. The CLI uses signal exits and does not account for fees or slippage, which is a measure of the difference between expected and actual trade prices. This test is not the same as the frozen 3R strategy or the held out test, so it is not sufficient for making trades. An independent audit using vectorbt, a backtesting library, was also run with SIP, or securities information processor, data and explicit costs. The test results showed sensitivity to different parameters: training base returns were 5.375%, 15.6616%, and 4.6417% for different parameter sets, while holdout returns were 20.9743%, 15.3639%, and -7.8131%. Stress tests, which simulate difficult market conditions, showed returns of -6.8202%, 8.9260%, and -0.8591% for training and 17.7518%, 12.7409%, and -9.7288% for holdout. The results indicate that the parameters are fragile, meaning small changes can greatly affect performance. The L20 parameters should be kept frozen, and the better holdout results for L10 should not be chosen. The data source is https://runway.adityaperswal.workers.dev/data/stock-bars, and the scripts and cached bars are available in the research folder of this workspace, with persistence via recorded recipe and manifest.

A preregistered trading strategy, called 2804, was tested before results were known. The strategy used a prior 20-day close breakout and a 50-day Simple Moving Average, which is a calculation of the average price of a stock over a certain period, to decide when to enter a trade. The next day's opening price was used to enter the trade, with a baseline entry of 5 basis points, which is a unit of measurement for interest rates and investment returns, and a stop loss of 2 Average Daily Ranges, which is a measure of how much a stock's price moves on average, and a target of 6 Average Daily Ranges. The trade was limited to a maximum of 40 sessions. The test was run from January 1, 2025, to September 11, 2026. The test used a dividend-adjusted proxy, which takes into account the impact of dividends on the stock price, rather than raw cash entitlement. The test also used a Vectorbt long-only cash portfolio, which is a type of investment portfolio that only holds cash and long positions, with fees of 1 basis point and slippage of 5 basis points on each side of the trade. The test included open and close marks, entry day exits, and conservative stop losses on the first day of ambiguity, and gap open exits. A 15 basis point stress fix was applied to the baseline barriers and the same entry and exit dates were used. An earlier diagnostic test, 2808, was superseded. The report includes returns on marked open positions where present, and the win rate is calculated on closed trades only. A manual reconciliation error of less than 4e-16 was found. The test did not include validation of 2R quarters or 15-minute/off-hours trades, and did not claim to execute corporate actions. The main result, L20, was positive but based on a small sample. Sensitivity tests for L10 and L40 were not selected. The data source was https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=CNQ&timeframe=1Day&feed=sip&adjustment=all. There is no order, and the current price of 50.07 is less than the prior 20-day close of 51.33.

A preregistered strategy, number 2804, was tested before results were known. The strategy used a prior 20-day close breakout and a 50-day Simple Moving Average, which is a calculation of the average price of a stock over 50 days. The next open was used as the entry point, with a baseline entry of 5 basis points, which is 0.05%, and an entry stop loss of 2 Average Daily Ranges, which is a measure of the average daily price movement, and a target of 6 Average Daily Ranges. The maximum number of sessions was 40. The test was run from January 1, 2021 to December 31, 2024. The test included a warm-up period in 2020 and used dividend-adjusted prices, which means the prices were adjusted to reflect the payment of dividends. The test also included a 1 basis point fee and 5 basis points of slippage, which is the difference between the expected price and the actual price, on each side of the trade. The test used open and close marks, entry day exits, and a conservative stop loss on the first day of ambiguity, and exited on gap opens. A 15 basis point stress fix was applied to the baseline barriers and the same entry and exit dates were used. An earlier diagnostic test, number 2808, was superseded. The report return includes the marked open position where present, and the win rate is based on closed trades only. A manual reconciliation error of less than 4e-16, which is a very small number, was found. The test did not include validation of 2R quarters or 15-minute or off-hours trades, and did not claim to execute corporate actions. The main test, L20, was positive but the sample size was small. The test did not select L10 or L40 sensitivity. The data source was https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=CNQ&timeframe=1Day&feed=sip&adjustment=all. There is no order, and the current price is 50.07, which is less than the prior 20-day price of 51.33.

A news scan found the headline is not available, but a data line shows the results of a required CLI run. The run had the following returns: for BTC, the return was 46.460751%, the maximum drawdown was 14.692428%, the win rate was 34.328358%, and there were 67 trades; for ETH, the return was 41.521926%, the maximum drawdown was 24.943256%, the win rate was 31.481481%, and there were 54 trades; for SOL, the return was 72.151727%, the maximum drawdown was 18.795394%, the win rate was 52.777778%, and there were 36 trades. This was a diagnostic test, not an executable policy, using a same-close prior 20 day policy, with no fees and no holdout, and including a partial day on September 13. The managed ETH studies with IDs 2883, 2884, 2897, and 2899 were retained unchanged, with a training return of 42.312262%, a maximum drawdown of 37.521903%, and 7 wins out of 13 trades. The known 2026 holdout return was 19.206747%, the maximum drawdown was 14.542745%, and there were 3 wins out of 3 trades, with 1 open trade, and a stress test return of 17.549943%. The daily control return was -1.925730% over 2 trades. The policy used a 20 day high, then a half average true range pullback within 3 days, with a next exact open, a 2 average true range stop, and a 20 day timeout. The fees were 25 basis points, with a 10 basis point slip, and 15 minute point checks, with a quarter of the position at 2 times the risk, then a stop entry, and a residual target net 3 times the risk. The daily disagreement and training mean confidence interval was between -3.77% and 12.80%. The missing 15 minute bounds were retained from a prior audit and were not new quote proof. On September 11 at 22 UTC, an entry was made for ETH, but it did not result in a trade, with a size of 0. This record retries two failed writes, but neither failure was counted as a record. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars and https://docs.alpaca.markets/us/docs/crypto-fees.

A test of Stitch Fix stock control, which is a system to manage stock prices, showed positive results with only two reused holdout events. The test looked at 10 consecutive earnings reports from AMC, from March 2024 to June 2026, and found that the median absolute change in stock price was 16.578014%. On September 11, the stock price was $2.95, and on October 16, it was $2.5. The test also considered a $0.12 cost on September 11 and a $0.5 cost on September 10, plus an assumed $0.05 per share cost. The results showed that the stock control system was effective, with 2 out of 10 trades covered at a 20% rate and 4 out of 10 trades covered at a 40% rate. However, the test also found that the call option premium exceeded the conviction risk by $50, and the put option failed 40% of the time. The test used data from option contracts and stock bars, and historical sources recorded from 26537 to 46. The test also used a momentum indicator, which showed a 41.383224% return over 20 days, with a 55.802626% drawdown and a 28.125% win rate. The test also used a vectorbt fixed lag system, which showed a 14.174754% return with a 14.475874% drawdown and a 37.5% win rate. The test found that the holdout period, which is a period of time used to test the system, cannot be used to validate the option premium decay, spread, or 3R target. A news scan found the headline 'Stitch Fix Announces Date for Fourth Quarter and Full Fiscal Year 2026 Financial Results and Conference Call' on the company's website, which stated that the earnings report would be released after market close on Wednesday, September 23, 2026. The test also found that the option expiry date was 23 days later, on October 16, and the default exit date was September 22. The sources used for the test included option contracts and stock bars data from https://runway.adityaperswal.workers.dev/data/stock-bars and https://runway.adityaperswal.workers.dev/data/option-bars.

The automated trading fund manager reran a required command line interface, which is a way to interact with a computer program using text commands. The command was: node /app/agent/dist/backtest.js --symbols FRO --strategy breakout --lookback20 --days2080 --stop-pct10 --target-pct30. This means the program was testing a strategy on the stock symbol FRO, using a breakout method, looking back 20 days, over 2080 days, with a 10 percent stop loss and a 30 percent target gain. The result was 1426 daily bars of data from the IEX exchange and 117 trades. The command line interface includes rules for when to exit a trade, percentage barriers, but does not account for holding onto a trade independently or the cost of making trades. This is different from another policy, called L20/SMA50/2ADR/40-session SIP, which had a negative result. This negative result cannot be replaced with a positive result from another test or hidden. No trades were made based on this command line interface test. The data came from https://runway.adityaperswal.workers.dev/data/stock-bars.

A fund manager made notes about a trading strategy called Frozen specification note2629, which was clarified in note2631. The strategy checks if the daily closing price is higher than the average of the last 50 days and the previous 20 days. If it is, the strategy buys the stock at the next scheduled opening time on the New York Stock Exchange at 9:30. The strategy uses a stop-loss, which is a price at which to sell the stock to limit losses, set at 2 times the average daily range of the last 20 days. The target price is 3 times the range. The strategy will sell the stock after a maximum of 40 sessions. If the stock price reaches 2 times the range, the strategy will sell a quarter of the stock and move the stop-loss to the break-even point. The strategy will sell the rest of the stock at the observed price plus any additional costs. The strategy does not consider dividend payments, only the stock price. There have been 16 closed trades and 1 open trade, with a win rate that does not include the open trade. The strategy was trained on historical data, including one trade with missing data from March 8, 2022. The data comes from https://runway.adityaperswal.workers.dev/data/stock-bars. The results of the strategy will be used to decide when to execute the next trade, but do not predict future performance, including a forthcoming 6.93% distribution.

A news scan found the text of a specification note for an automated trading fund manager. The note explains that the fund will enter a trade when the daily close exceeds the prior 20 closes and the 50-day simple moving average, which is a calculation of the average price of a stock over 50 days. The fund will enter the trade at the next scheduled New York Stock Exchange opening at 9:30, with a stop loss at 2 times the prior average daily range, and a target at 3 times the risk. The fund will also use a breakeven stop, which means that if the trade is profitable, the stop loss will be moved to the entry price, and the trade will be closed at a quarter of the target, which is 2 times the risk. The fund will exit the trade at the observed sampled price plus slippage, which is the difference between the expected price and the actual price, and never at the extreme price within a bar, which is a unit of time. The fund will also reconcile manual cash orders within a small margin of error, and will account for fees of 1 basis point per side, and slippage of 5 basis points and 15 basis points. The fund will handle early closes and will not account for dividend credits. The training data contains 31 closed trades and 0 open trades, with a win rate that excludes the open trade. The data also contains one trade with missing entry-day samples from March 8, 2022, but the envelope rules, which are rules that determine the range of prices, do not indicate a missed exit or partial trigger. The historical dates have already been viewed for other rules, and this is a chronological holdout for the frozen implementation, not globally unseen evidence. The costs can alter the barriers and dates, which explains the non-monotone variant behavior. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars. The positive results will advance the execution watch, but are not a forecast for the forthcoming 6.93% distribution.

A news scan found the specification note for R91208 FRO, which includes clarification notes 2629 and 2631. The raw SIP daily close exceeds the prior 20 closes and the 50-day simple moving average, which is a calculation of the average price over the last 50 days, also known as SMA50. When this happens, the system enters a trade at the next scheduled New York Stock Exchange opening at 9:30, which is a specific time when the market opens, with a stop loss set at 2 times the prior average daily range, which is the average difference between the high and low prices of the last 20 days, and a target of 3 times the range, which is the difference between the high and low prices. The maximum number of sessions for the trade is 40. The system uses actual slipped entry, which means it takes into account any delays in entering the trade, and samples subsequent 15-minute opening prices during regular trading hours, which is the time when the market is open. If the stop loss is triggered before the target, there will be no partial management of the trade. The trade exits at the observed sampled price plus slippage, which is the difference between the expected and actual prices, but never at the extreme price within a bar, which is a unit of time. The final exit is at the scheduled close minus 15 minutes on session 40. The system uses full-equity fractional long-only orders, which means it only buys and never sells short, and reconciles manual cash within a very small margin of 1e-9. The fee is 1 basis point per side, which is a unit of measurement for interest rates, and the slippage is 5 basis points and 15 basis points for stress testing. The calendar handles early closes, which means it takes into account any days when the market closes early. The system uses zero dividend credit and static raw barriers, which means it only looks at the price of the stock and not the total return including dividends. The account has 16 closed trades and 1 open trade, and the win rate excludes the open trade. The training data contains one trade with missing entry-day samples from March 8, 2022, but the envelope rules, which are used to determine the stop loss and target, did not trigger a missed exit or partial trigger. The historical dates have already been viewed for other rules, and this is a chronological holdout for the frozen implementation, not globally unseen evidence. Costs can alter barriers and dates, which explains non-monotone variant behavior. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars. Positive results will advance the execution watch, but this is not a forecast for the forthcoming 6.93% distribution.

A fund manager's note explains the rules for an automated trading system. The system checks if the daily closing price is higher than the previous 20 closes and the 50-day simple moving average, which is a calculation of the average price over 50 days. If it is, the system enters a trade at the next scheduled opening time on the New York Stock Exchange, which is 9:30. The system uses a stop-loss order, which is an order to sell a stock when it falls to a certain price, set at 2 times the average daily range of the previous 20 days. The target price is 3 times the range, and the trade can last up to 40 sessions. The system also considers slippage, which is the difference between the expected price and the actual price of a trade. If the stop-loss is triggered before the target price, the system does not use partial management, which means it does not sell part of the trade. The system exits the trade at the observed price plus slippage, but never at the highest or lowest price of the session. The system uses a type of order called vectorbt, which is a way of placing trades, and reconciles the trades with the cash in the account. The system also considers fees, which are costs of trading, and slippage. The system does not consider dividend credits, which are payments made to shareholders, and only looks at the price of the stock. The training data includes 31 closed trades and 0 open trades, and the win rate does not include the open trade. One trade is missing entry-day samples, but the system's rules mean that the missed exit and partial triggers are not possible. The system does not have any missing scheduled bars. The historical dates have already been used for other rules, and this is a test of the frozen implementation, not a test of the system on new data. Costs can affect the system's performance and alter the dates of trades, which can explain why the results are not always consistent. The data comes from https://runway.adityaperswal.workers.dev/data/stock-bars. The positive results mean that the system will be watched more closely, but they do not predict future performance, including the forthcoming 6.93% distribution.

A test of nine past AMC earnings reports from February 2024 to June 2026 found that the median absolute change in stock price was 5.543964% and the mean change was -0.908716%. On September 11, the stock price was $39.02, and on October 16, the call option price was $0.30 and the put option price was $0.40. Assuming a $0.05 per share roundtrip reserve, the required return would be 16.222450% or -11.455664%. However, only 1 out of 9 tests were successful, which is 11.11% and below the 40% threshold. Without the reserve, 2 out of 9 tests were successful. The test used data from option contracts and stock bars. The AMC earnings report is scheduled to be released on September 15, 2026, after the market closes. The test also used a vectorbt24 prespecified cell with lags of 2, 3, and 5 sessions, and directions of long and short. The test found that the T3 short train had a return of 2.471362% and a drawdown of 4.871260%, with a win rate of 57.142857%. The T2 short train had a return of 2.337936%, and the T5 short train had a return of 1.717228%. However, the T5 short train cannot be selected as a fresh out-of-sample winner because the known holdout had already been seen. The longs all had negative returns, with the T3 long train having a return of -8.144169% and a holdout return of -1.318695%. A required CLI diagnostic node test succeeded with 251 bars, 7 trades, a return of -14.664032%, a drawdown of 21.044477%, and a win rate of 42.857143%. The conclusion is that there is no entry, and the last two holdouts will be preserved and not retuned on a monthly shortfall.

A news scan found the text describing the C91208 trading fund manager's notes. The manager is reusing experiments 2883, 2884, 2897, and 2899, which were previously audited. The policy for these experiments is to buy when the price breaks out above the highest price of the last 20 days, then pulls back by half of the average true range within 3 days. The manager then sets a stop loss at 2 times the average true range over 20 days and a timeout of 20 days. There is a 0.25% entry fee and a 1% exit fee. The manager also allows for a 0.1% slip on each side and a 0.3% stress test. The experiments showed a 42.31% return from 2024 to 2025, with a 37.52% drawdown, and 7 out of 13 trades were winners. In the known 2026 holdout period, the return was 19.21%, with a 14.54% drawdown, and all 3 trades were winners, with one trade still open. The stress test showed a 17.55% return. The daily training showed a -27.13% return, while the holdout period showed a -1.93% return. The manager notes that 18 missing 15-minute checks are not a concern. The manager concludes that there is no need for a new order and will preserve the current candidate without selecting new parameters or inflating the sample size. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars and https://docs.alpaca.markets/us/docs/crypto-fees.

A command was given to run a test on two stocks, FRO and STNG, using a strategy called breakout. This strategy looks at the prices of the stocks over the past 20 days and decides what to do based on that. The test ran for 2080 days, with a stop loss of 10 percent and a target gain of 30 percent. The test used a method that makes decisions based on signals and also considers percentage stops and targets. It did not account for the costs of actually buying or selling the stocks and did not set aside any data to check the results independently. This test was a diagnostic tool, not a way to exactly copy a specific investment policy. The results of this test and another tool were kept, even though they were different, because the methods used were different. The data for the test came from https://runway.adityaperswal.workers.dev/data/stock-bars.

A required command line interface audit was run with the strategy breakout, looking back 20 days, on the currencies Bitcoin/USD, Ethereum/USD, and Solana/USD. The results were: Bitcoin had a return of 46.460751%, a drawdown of 14.692428%, a win rate of 34.328358%, and 67 trades; Ethereum had a return of 40.761782%, a drawdown of 24.943256%, a win rate of 30.909091%, and 55 trades; Solana had a return of 72.151727%, a drawdown of 18.795394%, a win rate of 52.777778%, and 36 trades. The audit inspected the strategies, including the prior 20 close threshold, same close entry, no explicit fees, and no held out window, with the current partial daily bar included. Solana had 748 bars versus 981 major bars. The premise of the gross performance is diagnostic only. A counterexample was found in Ethereum, which had a high cost and a next open daily loss of 4.8672% and an hourly loss of 0.4895%, and these also lacked a managed 3R. The conclusion is that there will be no cross-policy borrowing of command line interface returns for orders. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars. The command line interface and raw statistics were saved in observations, but this is not an adoption decision. A news scan found the headline is not applicable here, this is a raw script output.

A command was given to run a test on a trading strategy using the stocks FRO and STNG. The test used a strategy called breakout, looking back 20 days, over 2080 days, with a stop loss of 10 percent and a target gain of 30 percent. The test returned 1426 sets of stock price data and used a method called signal-driven exits, as well as stop losses and target gains based on percentages. This test did not include costs of executing trades or a separate set of data to test the results. It was a diagnostic test, not a copy of a specific investment policy, and cannot validate that policy. The data came from https://runway.adityaperswal.workers.dev/data/stock-bars. The results of this test and another method were kept, even if they were not the same, to show the differences between the two methods.

A news scan found the headline 'C91206 SOL hourly frozen Donchian comparator with known holdout and all sensitivity cells'. This is a report about an automated trading system. The system is based on two premises. The first premise is a trading rule that buys when the price breaks out above the highest price in the last 20 calendar days and sells when the price falls below the lowest price in the last 10 days. The system uses hourly data and considers fees of 25 basis points and a 10 basis point slip on each side, with a stress test of 30 basis points. The system was trained on data from 2024 to 2025, starting with SOL data from August 26, 2024, and the cash was reset on January 1. The system was tested on a holdout period from 2026, which was already researched, through September 12, 2026, at 06:00 UTC. The training period had a return of 9.334791%, a maximum drawdown of 36.795757%, 8 trades, a win rate of 50%, and a mean confidence interval of -4.938225% to 8.002702%. The holdout period had a marked return of 1.322187%, a maximum drawdown of 24.525746%, 4 closed trades, a win rate of 0%, a mean return of -5.897838%, and a confidence interval of -7.941673% to -3.062724%. There was also 1 open trade. The stress test return was -0.484906%. The second premise tested all combinations of 9 lookback periods and 3 simple moving average cells, with holdout base costs ranging from -2.935467% to 19.188094%. The mean confidence interval was calculated using 5000 independent and identically distributed trade bootstraps, but this was not adjusted for the small sample size and multiple comparisons. One issue with this approach is that open profits can mask closed losses, and a strict rolling simple moving average becomes unavailable after gaps, so the number of trades is not a valid measure of performance. The conclusion is that no winning parameter was chosen from the holdout period and no current late entry is authorized. The data source is https://runway.adityaperswal.workers.dev/data/crypto-bars and the fees are described at https://docs.alpaca.markets/us/docs/crypto-trading. The report covers all 216 cell figures and observations from 26302 to 26325.

A news scan found the daily frozen Donchian comparator results for SOL. The premise was to buy when the price broke out above the high of the past 20 calendar days and sell when it fell below the low of the past 10 days, with a 25 basis point fee and 10 basis point slip on each side, and a stress test of 30 basis points. The daily high, low, open, and close prices were reconstructed from 24 complete hours. There were no stops, targets, or partial management, as this was a diagnostic comparator, not an executable policy. The training period was from 2024 to 2025, starting on August 26, 2024, with the cash reset on January 1, and the known holdout period was from 2026 through September 1, 2026, at 12:00 UTC. The train return was -10.120708%, with a drawdown of 42.459529%, 7 trades, a win rate of 42.857143%, and a mean confidence interval of -8.221019% to 5.529944%. The holdout period had a marked return of 4.581108%, a drawdown of 17.406778%, 3 closed trades, a win rate of 0%, a mean return of -4.373566%, and a confidence interval of -5.151662% to -3.875026%, with 1 open trade. The stress return was 3.12758%. All 9 lookback periods of 15, 20, and 30 days, with simple moving averages of 0, 100, and 200 days, were retained. The holdout base cost range was 0 to 19.561908%. The mean confidence interval was calculated using 5000 independent and identically distributed trade bootstraps, without adjusting for small sample size and multiple comparisons. A counterexample noted that marked open profits can mask closed losses, and strict rolling simple moving averages become unavailable after gaps, with 0 trades not being validation. The conclusion was that no winning parameter was chosen from the holdout, and no current late entry was authorized. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars, and the fees are listed at https://docs.alpaca.markets/us/docs/crypto-trading. All 216 cell figures and 26302-26325 observations were considered.

A news scan found the following trading results for Ethereum (ETH): the trading strategy used a Donchian comparator, which is a technical analysis tool, with a prior 20 calendar-day high breakout and a prior 10-day low exit. It also used a fill following the open, long cash only, with fees of 25 basis points and a slip of 10 basis points on each side, and stress of 30 basis points. The hourly strategy used 480 hours for the high and 240 hours for the low. There were no stops, targets, or partial management, as this was a diagnostic comparator, not an executable policy. The training period was from 2024 to 2025, with Sol starting on August 26, 2024, and the cash was reset on January 1. The known holdout period was from 2026 through September 12, 06:00 UTC, and the results for 2026 were already researched. The train return was 16.493028%, with a drawdown of 41.889093%, 14 trades, a win rate of 42.857143%, and a mean confidence interval of -6.826896% to 12.704494%. The holdout return was -0.489511%, with a drawdown of 22.644888%, 5 closed trades, a win rate of 40%, a mean of 1.326191%, and a confidence interval of -5.247297% to 10.986248%. The stress return was -2.654456%. All 9 lookback cells with 15, 20, and 30 periods, and SMA 0, 100, and 200 were retained. The holdout base cost range was -12.836788% to 13.949464%. The mean confidence interval was calculated using 5000 independent and identically distributed trade bootstraps, but the results were not adjusted for small N and multiple comparisons. A counterexample noted that marked open profit can mask closed losses, and a strict rolling SMA becomes unavailable after gaps, and N0 is not validation. The conclusion was that no winning parameter was chosen from the holdout, and no current late entry was authorized. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars, and the fees are explained at https://docs.alpaca.markets/us/docs/crypto-trading. All 216 cell figures were observed from 26302 to 26325.

A news scan found the following text: Premise 1 is a preregistered trading strategy that buys when the price breaks out above the highest price in the last 20 calendar days and sells when the price falls below the lowest price in the last 10 days. It only buys at the next day's opening price and only uses cash, with fees of 25 basis points and a 10 basis point slip on each side, and a stress test of 30 basis points. The daily high, low, open, and close prices are reconstructed only from complete 24-hour periods. This strategy does not use stops, targets, or partial management, and is for diagnostic comparison only, not for actual use. The strategy was trained on data from 2024 to 2025, with the SOL data starting on August 26, 2024, and the cash was reset on January 1. The holdout period, which is a known period used to test the strategy, is from 2026 through September 1 at 12:00 UTC, and the results for 2026 were already researched and not new. The training period had a return of -24.66415%, a maximum drawdown of 50.545173%, 12 trades, a win rate of 41.666667%, and a mean confidence interval of -9.572975% to 7.246572%. The holdout period had a marked return of -4.867204%, a maximum drawdown of 17.451183%, 3 closed trades, a win rate of 0%, a mean return of -5.049061%, and a confidence interval of -9.724302% to -2.115529%. The stress test return was -6.189414%. Premise 2 retained all 9 combinations of 15, 20, and 30-day lookbacks and simple moving averages of 0, 100, and 200 days. The holdout base cost range was -12.839698% to 11.294846%. The mean confidence interval was calculated using 5000 independent and identically distributed trade bootstraps, but the results are not adjusted for multiple comparisons. A counterexample is that an open profit can mask closed losses, and a strict rolling simple moving average becomes unavailable after gaps, and the number of trades is not validation. The conclusion is that no winning parameter was chosen from the holdout period and no current late entry is authorized. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars, and the fees are described at https://docs.alpaca.markets/us/docs/crypto-trading. All 216 cell figures are based on observations from 26302 to 26325.

A news scan found the following report: Premise 1 is a trading strategy that was planned in advance. It buys when the price of Bitcoin breaks out above the highest price of the last 20 calendar days, and sells when the price falls to the lowest price of the last 10 days. The strategy only buys and sells at the opening price of the next hour, and it only uses cash, with fees of 25 basis points and a 10 basis point slip on each side, and a stress test of 30 basis points. This strategy uses hourly data, with 480 hours for the high price and 240 hours for the low price. There are no stop-loss orders or profit targets, and the strategy does not manage positions partially. This is a diagnostic tool, not a strategy that can be used by a trader, and it is different from a previous strategy used for Ethereum. The strategy was tested on data from 2024 to 2025, starting on August 26, 2024, for the cryptocurrency SOL, and the cash was reset on January 1. The data from 2026, up to September 1, 06:00 UTC, was already known and used to test the strategy, and it is not new, unseen data. The return on investment during the test period was 30.449462%, with a maximum drawdown of 34.232668%, and 13 trades were made, with a win rate of 30.769231%, and a mean confidence interval of -5.327741% to 13.821846%. In the holdout period, the return was 17.935419%, with a maximum drawdown of 12.151445%, and 3 trades were closed, with a win rate of 33.333333%, and a mean of 0.4253%, and a confidence interval of -2.628821% to 6.359096%, and 1 trade was still open. The stress test return was 16.296286%. Premise 2 tested all combinations of 9 different lookback periods and 3 different moving average parameters, and retained all the results. The holdout period had a base cost range of 0 to 17.935419%. The mean confidence interval was calculated using 5000 independent and identically distributed trade bootstraps, but the results were not adjusted for the small number of trades and multiple comparisons. One counterexample is that a marked open profit can hide closed losses, and a strict rolling moving average becomes unavailable after gaps, and the number of trades is not validation. The conclusion is that no winning parameter was chosen from the holdout period, and no current late entry is authorized. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars, and the fees are described at https://docs.alpaca.markets/us/docs/crypto-trading. All 216 cell figures are based on observations from 26302 to 26325.

A news scan found the headline 'C91206 BTC daily frozen Donchian comparator with known holdout and all sensitivity cells'. The report discusses a trading strategy for Bitcoin, denoted as BTC, which is a type of digital currency. The strategy is based on a premise that involves a 20-calendar-day high breakout and a 10-day low exit, with fees of 25 basis points and a slip of 10 basis points on each side, and a stress test of 30 basis points. A basis point is a unit of measurement, equal to 0.01%. The daily high, low, open, and close prices, referred to as OHLC, are reconstructed only from complete 24-hour periods. The strategy does not include stops, targets, or partial management, and is used for diagnostic comparison purposes only. The training period for the strategy is from 2024 to 2025, with the SOL, a type of cryptocurrency, starting on August 26, 2024. The cash is reset on January 1, and the holdout period, which is a period of time used to test the strategy, is from 2026 through September 1, 2026, at 12:00 UTC. The return on investment during the training period is 32.167365%, with a drawdown, which is a decline in value, of 32.754193%. The strategy had 10 trades, with a win rate of 30%, and a mean confidence interval, which is a measure of uncertainty, of -6.573376% to 17.652511%. During the holdout period, the return on investment is 18.257988%, with a drawdown of 10.252513%, and 2 closed trades, with a win rate of 50%. The mean confidence interval for the holdout period is -4.104569% to 4.199843%. The stress return, which is a measure of the strategy's performance during stressful market conditions, is 17.081765%. The report also discusses a second premise, which involves retaining all 9 lookback periods of 15, 20, and 30 days, with simple moving averages, referred to as SMA, of 0, 100, and 200 days. The holdout base cost range is from 0 to 18.257988%. The report concludes that no winning parameter was chosen from the holdout period, and no current late entry is authorized. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars, and the fees are discussed at https://docs.alpaca.markets/us/docs/crypto-trading. The report includes 216 cell figures and 26302-26325 observations.