Notes

A controlled experiment using a trading strategy was completed from 2024 to 2025 and reused from January 1, 2026, to September 11, 2026. The strategy had parameters including a 20 high and 10 low, with a 2 ATR (Average True Range, a measure of price volatility) of 14, a net 3R fixed 10 basis point target, and 25 basis point credited fees. It also had a 20-day timeout and next timestamp open checks, with an optional 2RBE (Risk-Reward Ratio, a measure of potential profit versus potential loss) entry and 2 ATR trail. The experiment used a 24-hour matching system to mask daily gaps and a 20-day window, but did not exactly replicate the prior strategy. The results were as follows: the symbol was BTC, the source was aggregate, the signal and execution were daily, and the part was holdout, with a trail of false and a slip of 10 basis points. The return was 7.421736699555104, the drawdown was 12.22313155668493, the win rate was 40, the sample size was 5, the mean was 1.7028789856190385, the average win was 9.890046523951035, and the average loss was 3.7552327066022926, with no zero volume entries or exits. The results were reconciled with Tradewealth and vectorbt, with marks not executable quotes, within a margin of 1e-10. https://runway.adityaperswal.workers.dev/data/crypto-bars

A controlled experiment using a trading strategy was completed from 2024 to 2025 and reused from January 1, 2026, to September 11, 2026. The strategy used a 20-day high and 10-day low, with a 14-day average true range, a fixed 10 basis point target, and 25 basis point credited fees. It also had a 20-day timeout and checked for the next timestamp open. The results showed a return of -6.7946%, a drawdown of 35.8169%, and a win rate of 41.1765%. The average win was 10.3560 and the average loss was 7.0545. A news scan found the trade results in a JSON format, which is a type of data format. The results were based on 17 samples and had a mean of 0.1145. The experiment used a daily signal and daily execution, with a 20 basis point slip. The trade wealth was reconciled using vectorbt, a type of trading software, within a 1e-10 margin, which means the results were very precise. The experiment also marked quotes that were not executable, meaning they could not be traded. https://runway.adityaperswal.workers.dev/data/crypto-bars

A controlled experiment using past data was completed from 2024 to 2025 and reused from January 1, 2026, to September 11, 2026. The experiment had 20 high and 10 low points, with a 2 average true range over 14 days, a fixed target of 3 times the risk, fees of 25 basis points, and a 20-day time limit. The next timestamp was used to open trades, with an optional 2 times risk entry and a 2 average true range trail. The experiment matched data over 24 hours, including daily gaps and a 20-day window, but did not exactly replicate the prior strategy. The results were: symbol BTC, aggregate source, daily signal, daily execution, train part, trail true, 10 basis point slip, return of -3.571165459343084, drawdown of 33.795926445638194, win rate of 41.17647058823529, 17 samples, mean of 0.314946366538795, average win of 10.576970886297655, average loss of 6.868470797292406, 0 zero volume entries, and 0 zero volume exits. The trade wealth was reconciled with vectorbt within a very small margin of error, and marks were made for quotes that were not executable, which means they were not able to be traded. https://runway.adityaperswal.workers.dev/data/crypto-bars

A controlled experiment to test trading strategies was completed from 2024 to 2025 and reused from January 1, 2026, to September 11, 2026. The experiment used a set of rules, including a 20-day high and 10-day low, a 14-day average true range, a fixed target of 3 times the risk, and fees of 25 basis points. The experiment also used a 20-day timeout and checked for the next timestamp open. The results showed a return of 0.8466, a drawdown of 31.538, and a win rate of 41.176. The average win was 11.653 and the average loss was 7.054. The experiment used a symbol, which refers to a unique identifier for a stock or currency, in this case, Bitcoin, and a source, which refers to where the data comes from, in this case, an aggregate of multiple sources. The signal and execution were both daily, meaning the strategy was based on daily data and trades were executed daily. The part of the experiment was the train, or training, phase, and the trail, which refers to a type of stop-loss order, was set to false. The slip, which refers to the difference between the expected and actual price of a trade, was 20 basis points. A news scan found the data line showing the results of the experiment. The results were reconciled with Tradewealth and vectorbt, which are tools used to analyze and execute trades, and the marks, or prices, were not executable quotes, meaning they were not actual prices that could be traded. https://runway.adityaperswal.workers.dev/data/crypto-bars

A controlled experiment to understand how a trading strategy works was completed from January 1, 2024, to September 11, 2026. The strategy used a 20-day high and 10-day low, with a 14-day average true range, a fixed target of 10 basis points, and fees of 25 basis points. The strategy also had a 20-day timeout and checked for the next timestamp open. The experiment used a daily signal and daily execution, and it did not exactly replicate a prior strategy. The results of the experiment showed a return of 4.334386391934331, a drawdown of 29.523597061664862, and a win rate of 41.17647058823529. The experiment was based on 17 samples, with a mean of 0.8501424740596831, an average win of 11.876732861705525, and an average loss of 6.868470797292406. There were no instances of zero volume entries or exits. A news scan found the headline 'Tradewealth reconciles vectorbt within 1e-10; marks not executable quotes', which is a statement about the accuracy of the trade data. https://runway.adityaperswal.workers.dev/data/crypto-bars

A controlled experiment using a trading strategy was completed from 2024 to 2025 and reused from January 1, 2026, to September 11, 2026. The strategy had 20 high and 10 low settings, with a 2 average true range over 14 periods, a fixed 3 return target of 10 basis points, and credited fees of 25 basis points. It also had a 20-day timeout and checked for the next timestamp open, with an optional 2 risk basis points entry and 2 average true range trail. The strategy matched the full 24 hours, masking daily gaps and 20-day windows, but did not exactly replicate the prior strategy. The results were as follows: the symbol was BTC, the source was native matched, the signal was calendar based, and the execution was hourly, with the part being holdout and the trail being true, and a slip of 20 basis points. The return was 10.805339966340298, the drawdown was 12.20466350739664, the win rate was 50, the sample size was 6, the mean was 1.886559398914069, the average win was 6.82552265047458, and the average loss was 3.052403852646442, with no zero volume entries or exits. The trade wealth reconciles with vectorbt within a very small margin of error, and marks quotes that are not executable. https://runway.adityaperswal.workers.dev/data/crypto-bars

A controlled experiment using a trading strategy was completed from 2024 to 2025 and reused from January 1, 2026, to September 11, 2026. The strategy had parameters including 20 high and 10 low, 2 ATR (Average True Range, a measure of price volatility) with a 14 period, a net 3R fixed 10 basis point target, 25 basis points credited for fees, and a 20 day timeout. It also included next timestamp open checks and an optional 2RBE (Risk-Reward Ratio, a measure of potential profit versus potential loss) entry with a 2 ATR trail. The experiment matched the full 24 hour period, considering daily gaps and a 20 day window, but did not exactly replicate the prior strategy. The results showed a return of 12.143017178330773%, a drawdown of 11.866958490428214%, and a win rate of 50%. The sample size was 6, with a mean of 2.0905369031711847, an average win of 7.039387988844603, and an average loss of 2.8583141825022347. There were no zero volume entries or exits. A news scan found the headline was not applicable, this is raw data. The trade wealth reconciles with vectorbt within a very small margin of error, and marks quotes that are not executable. https://runway.adityaperswal.workers.dev/data/crypto-bars

A controlled experiment using a trading strategy was completed from 2024 to 2025 and reused from January 1, 2026, to September 11, 2026. The strategy used a 20-day high and 10-day low, with a 14-day average true range, a fixed target of 3 times the risk, and fees of 25 basis points. The experiment had a 20-day timeout and checked for the next timestamp open. The results showed a return of 10.84%, a drawdown of 12.20%, and a win rate of 50%. The average win was 6.84% and the average loss was 3.05%. The experiment used a hourly execution and a calendar signal, with a 20 basis point slip. A news scan found the trade details as follows: symbol is BTC, source is native_matched, signal is calendar, execution is hourly, part is holdout, trail is false, and slip is 20 basis points. The trade wealth was reconciled with vectorbt within a small margin of error, and unexecutable quotes were marked. https://runway.adityaperswal.workers.dev/data/crypto-bars

A controlled experiment using a trading strategy was completed from 2024 to 2025 and reused from January 1, 2026, to September 11, 2026. The strategy had 20 high and 10 low settings, with a 2 ATR 14, which means a measure of price volatility, and a fixed target of 10 basis points, with 25 basis points credited for fees. The strategy also had a 20-day timeout and checked for the next timestamp open, with an optional 2RBE, or risk-based entry, and a 2 ATR trail. The experiment matched the BTC symbol, using a native matched source, with a calendar signal and hourly execution, on a holdout part, with a 10 basis point slip, and no trail. The results showed a return of 12.1745%, a drawdown of 11.867%, and a win rate of 50%, with a sample size of 6, a mean of 2.0954, an average win of 7.0492, and an average loss of 2.8583. The trade wealth reconciles with vectorbt, a trading software, within a small margin of error, and marks quotes that are not executable. https://runway.adityaperswal.workers.dev/data/crypto-bars

A controlled experiment using a trading strategy was completed from 2024 to 2025 and reused from January 1, 2026, to September 11, 2026. The strategy used a 20-day high and 10-day low, with a 2 ATR (Average True Range, a measure of price volatility) of 14 days, a fixed target of 3R (a risk-reward ratio), and credited fees of 25 basis points (a unit of measurement for interest rates). The strategy also had a 20-day timeout and checked for the next timestamp open. Additionally, it used an optional 2RBE (Risk-Reward Ratio Entry) and a 2 ATR trail. The experiment matched the native BTC symbol from the calendar signal source, with an hourly execution, and was part of the train trail. The results showed a return of 18.3044%, a drawdown of 29.1461%, and a win rate of 36.8421%. The average win was 12.5858% and the average loss was 5.1617%. The experiment had 19 samples and a mean of 1.3769%. There were no zero-volume entries or exits. The results were reconciled with Tradewealth and vectorbt, with a precision of 1e-10, and marked non-executable quotes. https://runway.adityaperswal.workers.dev/data/crypto-bars

A controlled experiment using a trading strategy was completed from January 1, 2024, to September 11, 2026, with a holdout period. The strategy used 20 high and 10 low values, with a 2 ATR (Average True Range, a measure of price volatility) of 14, a fixed target of 3R (three times the risk), and credited fees of 25 basis points (a unit of measurement for interest rates). The strategy also had a 20-day timeout and checked for the next timestamp open, with an optional 2R (two times the risk) entry and 2 ATR trail. The results were matched to a 24-hour calendar, which masked daily gaps and a 20-day window. The results do not exactly replicate the prior strategy. The results are as follows: the symbol is BTC, the source is native_matched, the signal is calendar, the execution is hourly, and the part is train, with a trail of true and a slip of 10 basis points. The return was 22.89%, the drawdown was 28.29%, and the win rate was 36.84%. The sample size was 19, the mean was 1.58, the average win was 12.81, and the average loss was 4.97. There were no zero-volume entries or exits. The trade wealth reconciles with vectorbt within a margin of 1e-10, and marks quotes that are not executable. https://runway.adityaperswal.workers.dev/data/crypto-bars

A controlled experiment to understand how a trading strategy works was completed from 2024 to 2025 and reused from January 1, 2026, to September 11, 2026. The strategy used 20 high and 10 low values, with a 2 ATR 14, which is a measure of price movement, and a fixed target of 10 basis points, or 0.1%, with fees of 25 basis points, or 0.25%, credited. The strategy also had a 20-day timeout and checked the next timestamp open. An additional 2R BE, or risk-to-reward ratio, was optional for entry, with a 2 ATR trail. The strategy matched the market 24 hours a day, using masks to account for daily gaps and a 20-day window, but did not exactly replicate the prior strategy. The results for the symbol BTC, using a native matched signal on a calendar basis, executed hourly, showed a return of 18.304427465635207%, a drawdown of 29.14607882770095%, and a win rate of 36.84210526315789%. The sample size was 19, with a mean of 1.376876439365853, an average win of 12.585845226133358, and an average loss of 5.161688686248525. There were no entries or exits with zero volume. The results were reconciled with Tradewealth and vectorbt, with a margin of error of 1e-10, and marked any non-executable quotes. https://runway.adityaperswal.workers.dev/data/crypto-bars

A controlled experiment using a trading strategy was completed from 2024 to 2025 and reused from January 1, 2026, to September 11, 2026. The strategy had 20 high and 10 low parameters, with a 2 average true range over 14 periods, a fixed net target of 3 times the risk, and fees of 25 basis points. The experiment also had a 20-day timeout and checked for the next timestamp open. The results showed a return of 22.89%, a drawdown of 28.29%, and a win rate of 36.84%. The average win was 12.81 and the average loss was 4.97. A comparison with another system, called vectorbt, showed that the results were very close, with a difference of less than 1e-10. This means that the trades were accurately recorded and reconciled. https://runway.adityaperswal.workers.dev/data/crypto-bars

The stressregime5_15_test had a return of -7.285553826338653 percent, with a drawdown of 22.12081821678159 percent and a compound annual growth rate, or CAGR, of -1.3260567890593156 percent, from January 4, 2021, to September 10, 2026. The win rate was 43.90243902439025 percent, based on 41 samples. The average winning trade was 3.0731734554743024 percent, and the average losing trade was 2.6346836582518964 percent. The mean trade percentage was -0.12879516929893167 percent, with a 95 percent confidence interval, or CI95, of -1.202347268207001 to 0.9499598378533135 percent. The half Kelly, a measure of optimal bet size, was 0. Reasons for trades included time, with 19 occurrences, stop, with 17 occurrences, gap stop, with 3 occurrences, and target, with 2 occurrences. There were 3 partials and 3 same day trades. The vbt, or vectorized backtest, was reconciled, with 85 orders, over 16 trade months. The market had a beta, or volatility measure, of 0.21150043775354208, an alpha, or excess return, of -4.019717940559546 percent per year, and a correlation, or corr, of 0.3462150404455502. This is a daily proxy, with all 20 cells reconciled, based on training data from 2016 to 2020 and test data from 2021 to September 10, 2026, with no globally untouched claim found in the inspected history. https://runway.adityaperswal.workers.dev/data/stock-bars ; research/results.json

A controlled experiment using a trading strategy was completed from January 1, 2026, to September 11, 2026. The strategy used a 20-day high and 10-day low, with a 2 ATR (Average True Range, a measure of price volatility) 14-day setting, a fixed 10 basis point target, and credited fees of 25 basis points. The strategy also had a 20-day timeout and checked for the next timestamp open. Additionally, it used an optional 2RBE (Risk-Reward Ratio, a measure of potential profit to potential loss) entry with a 2 ATR trail. The experiment matched the native strategy daily, with a 24-hour mask to account for daily gaps and a 20-day window. The results showed a return of 0.7557, a drawdown of 13.80, and a win rate of 40. The average win was 6.65 and the average loss was 3.95. There were no zero-volume entries or exits. The trade wealth was reconciled with vectorbt, a trading software, within a 1e-10 margin, and non-executable quotes were marked. https://runway.adityaperswal.workers.dev/data/crypto-bars

A controlled experiment using a trading strategy was completed from January 1, 2026, to September 11, 2026. The strategy used a 20-day high and 10-day low, with a 2 ATR (Average True Range, a measure of price volatility) 14-day setting, and a fixed 10 basis point target, with 25 basis point credited fees. The strategy also had a 20-day timeout and checked for the next timestamp open. Additionally, it used an optional 2RBE (Risk-Reward Ratio, a measure of potential profit versus potential loss) entry and a 2 ATR trail. The experiment used a 24-hour matching system to replicate daily gaps and a 20-day window, but did not exactly replicate the prior strategy. The results of the experiment were as follows: the symbol was BTC, the source was native_matched, the signal was daily, the execution was hourly, and the part was holdout, with a trail and a 10 basis point slip. The return was 1.7683071497686775, the drawdown was 13.283045511581715, and the win rate was 40. The sample size was 5, the mean was 0.49244691366545407, the average win was 6.863966344067074, and the average loss was 3.7552327066022926. There were no zero volume entries or exits. The results were reconciled with Tradewealth and vectorbt, with a margin of error of 1e-10, and marked any non-executable quotes. https://runway.adityaperswal.workers.dev/data/crypto-bars

The stressregime5_15_train model was trained from 2016-01-04 05:00:00+00:00 to 2020-12-31 05:00:00+00:00. The return was -3.5512104631624575, with a drawdown of 31.12832801616634 and a compound annual growth rate, or CAGR, of -0.7211216796557229. The win rate was 48.837209302325576, based on 43 samples. The mean trade percentage was -0.0052382715649152995, with an average win percentage of 3.03150452279356 and an average loss percentage of 2.9039473025434606. The 95% confidence interval, or CI95, was between -0.7943843415420596 and 0.8182585213220581. The half Kelly, a measure of optimal bet size, was 0. There were 19 stops, 23 time-based exits, and 1 target-based exit. The model had 2 partials and 7 same-day trades. The vbt, or vector-based trading, orders were reconciled, with 88 orders placed over 18 trade months. The market had a beta, or volatility measure, of 0.30074884783154426, an alpha, or excess return, of -4.575633852454907, and a correlation, or corr, of 0.4014145703729052. This is a daily proxy report, with all 20 cells reconciled, covering training from 2016 to 2020 and testing from 2021 to September 10. The history was inspected and no globally untouched claim was found. https://runway.adityaperswal.workers.dev/data/stock-bars ; research/results.json

A controlled experiment using a trading strategy was completed from 2024 to 2025 and reused from January 1, 2026, to September 11, 2026. The strategy had 20 high and 10 low settings, with a 2 ATR (Average True Range, a measure of price volatility) of 14, a fixed target of 10 basis points (a unit of measurement for interest rates), and credited fees of 25 basis points. It also had a 20-day timeout and checked for the next timestamp open. Additionally, it had an optional setting to add 2 times the risk of the trade to the entry point, with a 2 ATR trail. The experiment used a 24-hour matching system to mask daily gaps and a 20-day window, but did not exactly replicate the prior strategy. The results of the experiment were as follows: the symbol was BTC, the source was native_matched, the signal was daily, the execution was hourly, and the part was holdout, with a trail setting of false and a slip of 20 basis points. The return was 5.62680428188711, the drawdown was 13.801790657242508, the win rate was 40, the sample size was 5, the mean was 1.3429724939270815, the average win was 9.27872672706187, and the average loss was 3.9475303281627774. There were no zero volume entries or exits. The trade wealth was reconciled using vectorbt within a margin of 1e-10, and marks were not executable quotes. https://runway.adityaperswal.workers.dev/data/crypto-bars

A controlled experiment using a trading strategy was completed from 2024 to 2025 and reused from January 1, 2026, to September 11, 2026. The strategy had 20 high and 10 low settings, with a 2 ATR (Average True Range, a measure of price volatility) of 14, a fixed target of 3R (a risk-reward ratio) with a 10 basis point target, and credited fees of 25 basis points. It also had a 20-day timeout and checked for the next timestamp open. Additionally, it had an optional 2RBE (Risk-Reward Ratio Entry) and an entry 2ATR trail. The experiment used a 24-hour matching mask to account for daily gaps and a 20-day window. The results showed a return of 6.688373802888781, a drawdown of 13.283045511581715, and a win rate of 40. The sample size was 5, with a mean of 1.5458617339334224, an average win of 9.497503394736995, and an average loss of 3.7552327066022926. There were no zero-volume entries or exits. The trade wealth was reconciled using vectorbt, a trading software, within a margin of 1e-10, and non-executable quotes were marked. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the headline 'REV9110400 frozen cell stressregime5_5_test'. The test results for the stressregime5_5_test model are as follows: the test period started on 2021-01-04 05:00:00+00:00 and ended on 2026-09-10 04:00:00+00:00. The return was -1.261381525933425, the drawdown was 19.26179018524704, and the Compound Annual Growth Rate, which is a measure of investment return, was -0.22376183456347398. The win rate, which is the percentage of winning trades, was 48.78048780487805. There were 41 samples, and the mean trade percentage, which is the average percentage change per trade, was 0.02547047064335189. The average win percentage was 2.949690005810315, and the average loss percentage was 2.759500515229946. The mean confidence interval, which is a measure of uncertainty, was between -1.034219621696518 and 1.1220167616599663. The half Kelly, which is a measure of optimal bet size, was 0.004317482615661311. The reasons for trade closures were time, which means a trade was closed after a certain time, with 19 occurrences, stop, which means a trade was closed when it reached a certain price, with 17 occurrences, gap stop, which means a trade was closed due to a large price gap, with 3 occurrences, and target, which means a trade was closed when it reached a certain profit, with 2 occurrences. There were 3 partial closures, which means a trade was partially closed, and 3 same-day closures, which means a trade was closed on the same day it was opened. The vbt, which stands for vectorized backtest, was reconciled, and there were 85 vbt orders. The trade months, which is the number of months the model was trading, was 16. The market beta, which is a measure of volatility, was 0.2111427235089168, the alpha annual percentage, which is a measure of excess return, was -2.9020391488859394, and the alpha t HAC5, which is a measure of statistical significance, was -0.7889268021568044. The correlation, which is a measure of relationship between two variables, was 0.3450909432183879. This is a daily proxy, and all 20 cells were reconciled. The model was trained on data from 2016 to 2020 and tested on data from 2021 to September 10, and the history was inspected, with no globally untouched claim found. https://runway.adityaperswal.workers.dev/data/stock-bars ; research/results.json

A controlled experiment using a trading strategy was completed from 2024 to 2025 and reused from January 1, 2026, to September 11, 2026. The strategy had 20 high and 10 low settings, with a 2 ATR, or average true range, of 14, a fixed net target of 3R, and credited fees of 25 basis points. It also had a 20-day timeout and checked for the next timestamp open, with an optional 2RBE, or risk-based entry, and a 2 ATR trail. The strategy was matched for 24 hours, using masks for daily gaps and a 20-day window, but did not exactly replicate the prior strategy. The results for the BTC symbol, using a native matched source, with a daily signal and hourly execution, and a train part, with a trail, and a 20 basis point slip, were a return of -15.198785883777022, a drawdown of 40.689361015708016, and a win rate of 38.88888888888889. The sample size was 18, with a mean of -0.4530213330605443, an average win of 9.916863380139858, and an average loss of 7.052038877824438. There were no zero volume entries or exits. The results were reconciled with Tradewealth and vectorbt, with a margin of error of 1e-10, and marked non-executable quotes. https://runway.adityaperswal.workers.dev/data/crypto-bars

A controlled experiment using a trading strategy was completed from 2024 to 2025 and reused from January 1, 2026, to September 11, 2026. The strategy had 20 high and 10 low settings, with a 2 ATR 14, which means a measure of price volatility, and a fixed 10 basis point target, with 25 basis points credited for fees. It also had a 20-day timeout and checks for the next timestamp open, with an optional 2RBE, or risk-based entry, and a 2 ATR trail. The experiment used a 24-hour matching system to mask daily gaps and a 20-day window, but did not exactly replicate the prior strategy. The results for the BTC symbol, using a native matched signal, with daily signals and hourly execution, and a train trail, showed a return of -12.09%, a drawdown of 38.70%, and a win rate of 38.89%. The sample size was 18, with a mean of -0.25, an average win of 10.14, and an average loss of 6.87. There were no zero volume entries or exits, and the results were reconciled with Tradewealth within a very small margin of error, marking some quotes as not executable. https://runway.adityaperswal.workers.dev/data/crypto-bars

The stressregime5_5_train model was trained from 2016-01-04 05:00:00+00:00 to 2020-12-31 05:00:00+00:00 with a slippage of 5 basis points. The model had a return of 3.141303177443411, a drawdown of 29.208956587023938, and a compound annual growth rate, or CAGR, which is a measure of an investment's annual growth rate, of 0.6210063886076256. The win rate was 48.837209302325576, with 43 samples, a mean trade percentage of 0.15273155105519037, an average win percentage of 3.2610566105679957, and an average loss percentage of 2.814306005752487. The mean confidence interval, or CI, which is a range of values within which a result is expected to lie, at 95% was between -0.6414368975212927 and 0.987275693369416. The half Kelly criterion, which is a formula used to determine the optimal bet size, was 0.02341749458752701. The reasons for trades were gap stop 2 times, time 23 times, stop 17 times, and target 1 time. There were 3 partials, 5 same day trades, and 89 vbt orders, which are orders executed through a specific trading system, over 18 trade months. The market had a beta, which is a measure of volatility, of 0.33097778080530876, an alpha, which is a measure of an investment's performance, of -3.7009603880671085 percent per year, and a correlation, or corr, which is a measure of how much two variables move together, of 0.4389026919695314. This is a daily proxy report, and all 20 cells were reconciled. The training period was from 2016 to 2020, and the test period was from 2021 to September 10. The history was inspected, and there were no globally untouched claims found. A news scan found the headline 'Daily proxy', which is a type of financial report. https://runway.adityaperswal.workers.dev/data/stock-bars ; research/results.json

A controlled experiment using a trading strategy was completed from 2024 to 2025 and reused from January 1, 2026, to September 11, 2026. The strategy had 20 high and 10 low settings, with a 2 ATR, or average true range, of 14, which is a measure of price volatility. The goal was to make a net return of 3 times the risk, with a fixed target of 10 basis points, or 0.1%, and fees of 25 basis points, or 0.25%, were credited. Trades were stopped after 20 days, and the next trade was opened at the next timestamp. There was also an option to add 2 times the risk and entry 2 ATR trail, which is a way to adjust the stop-loss level. The experiment used a 24-hour matching system to mask daily gaps and a 20-day window, but it did not exactly replicate the prior strategy. The results of the experiment were: symbol BTC, source native matched, signal daily, execution hourly, part train, trail false, slip 20 basis points, return -8.435%, drawdown 35.958%, win rate 38.888%, sample 18, mean 0.055, average win 11.223, average loss 7.052, zero volume entries 0, and zero volume exits 0. The trade wealth was reconciled using vectorbt, a trading software, within a small margin of error of 1e-10, and marks were made for quotes that were not executable. https://runway.adityaperswal.workers.dev/data/crypto-bars

A controlled experiment using a trading strategy was completed from 2024 to September 11, 2026. The strategy used a 20-day high and 10-day low, with a 2 ATR (Average True Range, a measure of price volatility) of 14 days, a fixed target of 3R (a risk-reward ratio), and fees of 25 basis points (a unit of measurement for interest rates). The experiment also used a 20-day timeout and checked for the next timestamp open. The results showed a return of -5.0789%, a drawdown of 33.81%, and a win rate of 38.89%. The average win was 11.45 and the average loss was 6.87. The experiment used a daily and hourly trading signal and did not replicate a prior strategy exactly. A news scan found the trade details: symbol is BTC, source is native_matched, signal is daily, execution is hourly, part is train, trail is false, slip is 10 basis points, and there were no zero volume entries or exits. https://runway.adityaperswal.workers.dev/data/crypto-bars