Notes

The trading fund manager reported the following results: a return of -5.86672557518143, a drawdown of 66.46981123047043, and a win rate of 41.37931034482759. The sample size was 29, with a mean percentage change of 0.6564780043839195 and a median percentage change of -6.713998098143503. The manager set 5 targets, 21 stops, and 11 partials, with 149 missing checks and 20 affected trades. The average trade lasted 15.620689655172415 days. This data was computed from September 13, 2020, to September 11, 2026, using an adjusted signal and split-only ADR, which is a type of stock ratio. The manager used automatic 15-minute and 5-minute signals, with fees of 1 basis point per side and a 5 basis point slip, and a base fee of 15 basis points with stress testing on fixed paths. The training period ended before 2025, and the holdout period started in 2025 with known test entries. The manager marked one open trade that was not closed, and the sampled drawdown was not the worst possible loss. https://runway.adityaperswal.workers.dev/data/stock-bars

A trading fund manager made notes about a trade. The return on the trade was 57.28802654527825 percent. The trade was closed 9 times. The open mark percentage was 20.618200031602196 percent. The mark price was 47.93. The mark time was September 11, 2026, at 19:55. The open entry time was July 24, 2026, at 13:30. The stop price was 36.69136. The target price was 48.365919999999996. The trade was not partially open. The full period sampled drawdown was 18.995177134767115 percent. The unliquidated mark return was 57.54006702150254 percent. A news scan found the headline is not available, but the notes also mentioned that the signal was computed on September 13, 2026, for the period from 2020 to September 11, 2026. The signal was adjusted and split-only, using automatic raw 15-minute and manager 5-minute data, with fees of 1 basis point per side and a 5 basis point slip, and a base fee of 15 basis points with stress on fixed paths. The training data only included exits before 2025, and the test data only included entries from 2025 or later. The trade was marked as open and not closed. The sampled drawdown was not the worst possible loss. https://runway.adityaperswal.workers.dev/data/stock-bars

A trading fund manager made notes about a trade. The return on investment was 60.4654589284134 percent. The trade was closed 9 times. The open mark percentage, which is the percentage of the price movement when the trade was opened, was 20.859678090833157 percent. The mark price, which is the current price of the trade, was 47.93. The mark time, which is the time when the trade was marked, was September 11, 2026, at 19:55:00. The open entry, which is the time when the trade was opened, was July 24, 2026, at 13:30:00. The stop, which is the price at which the trade will be automatically closed to limit losses, was 36.69136. The target, which is the price at which the trade will be automatically closed to lock in profits, was 48.365919999999996. The trade was not partially closed. The full period sampled drawdown, which is the maximum loss during a certain period, was 18.833005195468388 percent. The unliquidated mark return, which is the return on investment of the trade that has not been closed, was 60.561787973107826 percent. A news scan found the headline about a computation on September 13, 2026. The computation was based on data from September 11, 2026, to September 13, 2026, and used a signal called SIP2020 and an adjustment called all-adjusted signal/split-only ADR. The manager used automatic 15-minute and 5-minute signals and fees of 1 basis point per side and 5 basis points per slip, with a base fee of 15 basis points and stress fixed paths. The manager trained the model on data before 2025 and tested it on data from 2025 and later. The manager marked an open trade that was not closed. The sampled drawdown was not the worst quote loss. https://runway.adityaperswal.workers.dev/data/stock-bars

The trading results show a return of 30.401569998614207, a drawdown of 17.362647547615552, and a win rate of 55.55555555555556. The sample size is 9, with a mean percentage of 3.47584479453269 and a median percentage of 4.921927782891999. There were 2 targets, 4 stops, and 0 partials, with 14 missing checks and 3 affected trades. The average duration of trades was 24.444444444444443 days. The data was computed from September 13, 2020, to September 11, 2026, using an adjusted signal and split-only ADR, which is a type of stock price comparison. The system used automatic 15-minute and manager 5-minute settings, with fees of 1 basis point per side and a 5 basis point slip, and a base fee of 15 basis points, with stress testing using fixed paths. The training data only included trades that exited before 2025, while the holdout data included trades that entered on or after 2025, which is a known test period. The results marked one open trade that was not closed. The sampled drawdown was not the worst possible loss, but rather a quote loss, which refers to the loss based on the current market quote. https://runway.adityaperswal.workers.dev/data/stock-bars

The results of a trading strategy are as follows: the return was 32.77%, the drawdown was 17.03%, and the win rate was 55.56%. The sample size was 9, the mean percentage change was 3.68%, and the median percentage change was 5.13%. There were 2 targets, 4 stops, and 0 partials. The strategy had 14 missing checks and affected 3 trades, with an average duration of 24.44 days. This data was computed from September 13, using the SIP2020 signal from September 11, 2026, with all adjustments made for signal and split-only ADR, and using 15-minute raw data and 5-minute manager data. The fees were 1 basis point per side, with a 5 basis point slip and a 15 basis point base stress, using fixed paths. The training exits were before 2025, and the holdout entries were from 2025 onwards, using a known test. One open trade was marked as not closed. The sampled drawdown was not the worst quote loss. https://runway.adityaperswal.workers.dev/data/stock-bars

A news scan found the following data: the return was -12.637925240348268, the drawdown was 72.72044559862007, and the win rate was 31.03448275862069. This data is based on a sample of 29 items, with a mean percentage of 0.657512817393008 and a median percentage of -7.725917816739591. There were 6 targets, 18 stops, and 0 partials, with 151 missing checks and 20 affected trades. The average duration of trades was 16.79310344827586 days. The data was computed from September 13, 2020, to September 11, 2026, using an all-adjusted signal and split-only ADR, which is a type of stock ratio. The data also used automatic raw 15-minute and manager 5-minute signals, with fees of 1 basis point per side and a slip of 5 basis points, and a base stress of 15 basis points with fixed paths. The training data excluded trades that ended before 2025, and the test data included trades that started in or after 2025. The data marked open trades that were not closed, and sampled drawdown, which is a measure of loss, rather than the worst quote loss. https://runway.adityaperswal.workers.dev/data/stock-bars

The results of the open_fixed train for TEN R91306 show a return of -7.421093503784282, a drawdown of 72.28046176718908, and a win rate of 31.03448275862069. This is based on a sample of 29 trades, with a mean percentage change of 0.8590295108587924 and a median percentage change of -7.541184780787347. The strategy used 6 targets and 18 stops, with no partials. There were 151 missing checks and 20 affected trades, with an average duration of 16.79310344827586 days. The data was computed from September 13, 2020, to September 11, 2026, using an all-adjusted signal and split-only ADR, with automatic raw 15-minute and manager 5-minute signals. The fees were 1 basis point per side, with a 5 basis point slip and a base 15 basis point stress, using fixed paths. The train exits before 2025 and the holdout entries are from 2025 onwards, which is a known test period. The results marked one open trade that was not closed. The sampled drawdown is not the worst quote loss. https://runway.adityaperswal.workers.dev/data/stock-bars

The final verification of the trading strategy for SFIX had an exit code of 0, and the selected grid remained unchanged. The baseline total cost per share for a round trip was $0.05, which includes a $0.0125 fee and $0.0125 price slippage on each side. Under stress, the total cost was $0.10, which includes a $0.025 fee and $0.025 slippage per side. This was tested with one whole contract and $1000 in capital, with hypothetical trade fills. The training results showed a return of -0.9% with 2.861371 drawdowns and a 33.3333% win rate for T3, and a return of 2.6% with 0 drawdowns and a 100% win rate for the holdout. The stress test results showed a return of -3.9% and a drawdown of +2.1%. The initial consolidated fee implementation was found to be arithmetically equivalent, and the final explicit costs were verified. Using the command line interface, the strategy was tested with the following parameters: symbol SFIX, momentum strategy, 20-day lookback, and 1000 days. The results showed 685 bars, 32 trades, a return of 41.383224%, a drawdown of 55.802626%, and a win rate of 28.125%. The buy and hold return was -24.777637%. The unsplit stock diagnostic showed no adoption. https://runway.adityaperswal.workers.dev/data/option-bars ; research/study.py ; node /app/agent/dist/backtest.js

A news scan found the headline 'C91306 holdout quarter2R_BE_net3R synchronized Monte Carlo'. The automated trading fund manager used a holdout window, a policy called quarter2R_BE_net3R, and the symbol three29.785probes. The manager ran 10,000 paths, using 254 days of source data. The results showed that 89.355 was deployed, with a mean profit and loss of -0.8122506651302756. The worst profit and loss was -10.852960220554175, and there was a 0 probability of a 3 percent account loss. The manager used a seed of 91306, running 10,000 paths over 480 hours, with 20 synchronized 24-hour blocks sampled by future day of the week, starting on Sunday. The manager trained on data from 2024 to 2025, and tested on a holdout pool from 2026. The manager initially used a spot entry with a 1.001 accounts buy slip, and fees of 25 basis points, plus a 10 basis points sell slip. The manager checked hourly, with partial checks every 2 hours, using a proxy, not actual manager or 15-minute live entry conditioned odds. The expected return was not forecast, with a fixed current average true range, and the historical pool could not model unseen losses. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the headline related to a trading simulation for SOL/USD. The simulation used a method called Monte Carlo, which is a way of making predictions using random samples. It was run on a set of data called the holdout quarter, using a policy called quarter2R_BE_net3R, which is a set of rules for making trades. The symbol being traded was SOL. The simulation ran 10,000 times, using 254 days of data from the past. The results showed a 46.27% chance of stopping a trade, a 1.76% chance of reaching a target, a 51.97% chance of running out of time, a 13.68% chance of making a partial trade, and a 39.51% chance of winning. On average, the trades resulted in a return of -1.22%, with an average win of 12.32% and an average loss of -10.07%. The worst-case scenario was a loss of -15.50%. The simulation also calculated a value called the quarter probe p99 loss, which was $4.22. The target gain was 38.93%. The average time it took to exit a trade was 362 hours. The simulation was run using a seed number of 91306, and it sampled data from future days of the week, starting on Sunday. The data was split into two groups: one for training, from 2024-2025, and one for testing, from 2026. The simulation assumed an initial buy with a 0.1% slip, and a 25 basis point fee, plus a 10 basis point sell slip. It checked the trades every hour, but only updated the partial trades every 2 hours. This was not an actual trading manager, but rather a simulation. The expected return was not forecast, and the simulation used a fixed current average true range, which is a measure of price volatility. The historical data used in the simulation cannot model unseen losses, which are losses that have not happened before. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data: the window is a holdout period, the policy is called quarter2R_BE_net3R, and the symbol is ETH, which is a type of cryptocurrency. The report is based on 10,000 paths, using 254 days of source data. The stop probability is 59.61%, the target probability is 6.92%, and the time probability is 33.47%. The partial probability is 24.81% and the win probability is 36.91%. The mean return is -0.60%, the mean win is 12.13%, and the mean loss is -8.05%. The worst sample loss is -12.22% and the target gain is 28.11%. The report also mentions a seed, 91306, and 10,000 paths over 480 hours, with 20 synchronized 24-hour complete UTC day blocks sampled by future day of week starting Sunday. The training data is from 2024-25 and the holdout data is from 2026, and they are separate. The spot initially has an entry and a 0.1% buy slip, with 25 basis point fees and a 10 basis point sell slip. The report checks hourly, with partial checks every 2 hours, but this is not an actual manager or 15-minute live entry conditioned odds. The expected return is not a forecast, the current average true range is fixed, and the historical pool cannot model unseen losses. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data: the window is a holdout period, the policy is called quarter2R_BE_net3R, and the symbol is BTC. The simulation used 10000 paths and considered 254 days of past data. The stop probability is 56.08%, the target probability is 3.56%, the time probability is 40.36%, the partial probability is 20.53%, and the win probability is 36.71%. The mean return is -0.9019300226542327%, the mean win is 8.636151571050062%, and the mean loss is -6.4342886149260705%. The worst 1% return is -8.782240210887945%, and the worst sample loss is -10.342024717816%. The target gain is 23.340149071215286%. The simulation was run with a seed of 91306, using 10000 paths over 480 hours, with 20 synchronized 24-hour blocks, starting on Sunday. The data was split into training and holdout periods, from 2024-2025 and 2026, respectively. The simulation assumed an initial spot entry with a 0.1% buy slip and 25 basis point fees, plus a 10 basis point sell slip. The simulation checked conditions hourly, with partial checks every 2 hours, but this is not an actual manager or live entry condition. The expected return is not a forecast, and the simulation used a fixed current average true range, which cannot model unseen losses. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the headline about a trading simulation called C91306. This simulation used a method called Monte Carlo to test a trading policy called full_net3R on a symbol named three29.785probes. The simulation ran 10,000 paths, which is a way to test many possible outcomes, over 254 days of data. The results showed a mean profit and loss, or pnl, of -0.8384589462807206. The worst loss was -10.852960220554175 and there was a 0 percent probability of losing 3 percent of the account. The simulation started with a deployed amount of 89.355. It used a technique called synchronized sampling, which means it looked at data from the same time of day, every day, for 480 hours. The data was split into training and testing sets, from 2024 to 2025 and 2026, respectively. The simulation assumed a 0.1 percent fee to buy and a 0.25 percent fee to sell, and it checked the results every hour, but only looked at partial data every 2 hours. The simulation did not try to predict future returns, and it used a fixed measure of market volatility called ATR, which stands for Average True Range, a way to measure how much prices move. The simulation also did not try to model losses that had not been seen before. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data line: a trading simulation for SOL/USD used a holdout window, a full_net3R policy, and the SOL symbol. The simulation had 10,000 paths, used 254 days of source data, and had a 45.7% stop probability, 3.07% target probability, 51.23% time probability, and 0% partial probability. The win probability was 39.19%, the mean return was -1.23%, the mean win was 12.45%, and the mean loss was -10.04%. The worst 1% return was -14.17% and the worst sample loss was -15.49%. The simulation used a seed of 91306, 10,000 paths, and 480 hours, with 20 synchronized 24-hour blocks sampled by future day of the week starting on Sunday. The training data was from 2024-2025 and the holdout data was from 2026. The simulation initially used a spot entry with a 0.1% buy slip and 25 basis point fees plus a 10 basis point sell slip. Hourly checks were used, with partial checks every 2 hours as a proxy, not actual manager or 15-minute live entry conditioned odds. The expected return was not forecast, the current average true range was fixed, and the historical pool cannot model unseen losses. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data line: a holdout window with a policy of full_net3R for the symbol ETH. The data includes 10,000 paths, 254 source days, a stop probability of 57.1%, a target probability of 9.77%, a time probability of 33.13%, and a win probability of 34.51%. The mean return is -0.66%, the mean win is 13.17%, and the mean loss is -7.94%. The worst sample loss is -12.22%. The data also includes a target gain of 24.82% and a mean exit time of 296.05 hours. This data is based on a simulation with a seed of 91306, using 10,000 paths and 480 hours, with 20 synchronized 24-hour complete UTC day blocks sampled by future day of week starting on Sunday. The training data is from 2024-2025 and is compared to a known 2026 holdout pool. The simulation includes a spot initially entry with a 0.1% buy slip and 25 basis point fees plus 10 basis point sell slip. Hourly checks are used, with partial checks every 2 hours as a proxy, but not actual manager or 15-minute live entry conditioned odds. The expected return is not forecasted, with a fixed current average true range, and the historical pool cannot model unseen losses. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data line: a Monte Carlo simulation for BTC/USD with the following parameters: window is holdout, policy is full_net3R, symbol is BTC, 10000 paths, 254 source days, stop probability is 54.72%, target probability is 5.73%, time probability is 39.55%, partial probability is 0%, win probability is 35.37%, mean return is -0.93%, mean win is 9.05%, mean loss is -6.39%, worst 1% return is -8.79%, and worst sample is -10.34%. The simulation used seed 91306, 10000 paths, and 480 hours, with 20 synchronized 24-hour complete UTC day blocks sampled by future day of week starting Sunday. The training data is from 2024-2025 and the holdout data is from 2026. The simulation assumes an initial spot entry with a 0.1% buy slip and 25 basis point fees plus 10 basis point sell slip. Hourly checks are used, with partial only every 2 hours as a proxy, not actual manager or 15-minute live entry conditioned odds. The expected return is not forecasted, with a fixed current average true range, and the historical pool cannot model unseen losses. https://runway.adityaperswal.workers.dev/data/crypto-bars

A train quarter2R_BE_net3R synchronized Monte Carlo simulation was run. The simulation used a window for training, with a policy called quarter2R_BE_net3R, and a symbol called three29.785probes. It used 10000 paths and 488 source days. The simulation was deployed with 89.355, and the mean profit and loss was 0.0392655997184675. The worst profit and loss was -12.093032967877422, and the probability of a 3 percent loss of the account was 0. The seed used was 91306, with 10000 paths and 480 hours. The simulation was run in 20 synchronized 24-hour complete UTC day blocks, starting on Sunday. The simulation was trained on data from 2024-25 and tested on known holdout pools from 2026, separately. The spot initially had an entry and a 1.001 accounts buy slip, with 25 basis point fees and a 10 basis point sell slip. The simulation checked hourly, with partial checks every 2 hours, but this was not an actual manager or a 15-minute live entry conditioned odds. The expected return was not forecast, the current average true range was fixed, and the historical pool cannot model unseen losses. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data line: a train quarter2R_BE_net3R synchronized Monte Carlo simulation for SOL/USD, also known as a computer model that predicts prices, with the following results: window is train, policy is quarter2R_BE_net3R, symbol is SOL, paths are 10000, source days are 488, stop probability percentage is 56.44, target probability percentage is 6.91, time probability percentage is 36.65, partial probability percentage is 24.61, win probability percentage is 39.24, mean return percentage is -0.096, mean win percentage is 16.93, mean loss percentage is -11.09, p01 return percentage is -14.75, worst sample percentage is -16.94, half kelly is 0, quarter probe p99 loss USD is 4.39, target gain percentage is 38.93, mean exit hours is 304.16. The simulation used seed 91306, 10000 paths times 480 hours, 20 synchronized 24 hour complete UTC day blocks sampled by future day of week starting Sunday. The training data is from 2024-25 and the known holdout pools are from 2026, which are separate. The spot initially had an entry of 1.001, with a buy slip, 25 basis points fees, and 10 basis points sell slip. The simulation had hourly checks, partial only every 2 hours proxy, and is not an actual manager or 15 minute live entry conditioned odds. The expected return is not forecast, the current ATR is fixed, and the historical pool cannot model unseen losses. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data: the automated trading fund manager used a synchronized Monte Carlo method for the ETH/USD currency pair. The training window used 488 days of data and 10,000 paths to estimate probabilities. The results showed a 66.06% stop probability, 11.05% target probability, 22.89% time probability, 29.63% partial probability, and 36.08% win probability. The mean return was -0.47%, with a mean win of 14.12% and mean loss of -8.70%. The worst sample showed a -18.65% loss. The target gain was 28.11%. The training data was from 2024-2025 and was compared to a holdout pool from 2026. The simulation used a 25 basis point fee and a 10 basis point sell slip. The results are based on hourly checks and do not reflect actual management or live entry conditions. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data line: a train quarter2R_BE_net3R synchronized Monte Carlo simulation for BTC/USD had the following results. The simulation used 10000 paths and considered 488 days of data. The stop probability was 49.59%, the target probability was 5.1%, and the time probability was 45.31%. The partial probability was 27.59%, and the win probability was 44.81%. The mean return was 0.6948911525055191%, the mean win was 9.444632733281523%, and the mean loss was -6.40922046616766%. The worst sample loss was -10.334308390030266%, and the half Kelly criterion was 0.573979658083215. The quarter probe p99 loss in USD was 2.498397406158138, and the target gain was 23.340149071215286%. The mean exit time was 329.4203 hours. This simulation, known as C91306, used 10000 paths and 480 hours, with 20 synchronized 24-hour complete UTC day blocks sampled by future day of week starting on Sunday. The training data was from 2024-25 and was compared to a known 2026 holdout pool. The spot initially had an entry and 1.001 accounts with a buy slip, with 25 basis point fees and 10 basis point sell slip. Hourly checks were used, with partial only every 2 hours as a proxy, not actual manager or 15-minute live entry conditioned odds. The expected return was not forecasted, with a fixed current average true range, and the historical pool cannot model unseen losses. https://runway.adityaperswal.workers.dev/data/crypto-bars

A train full_net3R synchronized Monte Carlo simulation was run. The simulation used a window for training, a policy called full_net3R, and a symbol called three29.785probes. It generated 10000 paths and used 488 days of source data. The deployed amount was 89.355 and the mean profit and loss was 0.08600822750196574. The worst profit and loss was -12.093032967877422 and the probability of a 3 percent loss in the account was 0. The simulation used a seed of 91306 and ran 10000 paths over 480 hours, with 20 synchronized 24-hour blocks sampled by the day of the week, starting on Sunday. The training data was from 2024 to 2025 and was compared to a holdout pool from 2026. The initial spot entry had a 1.001 accounts buy slip and 25 basis point fees, plus a 10 basis point sell slip. The simulation checked the data hourly, but only used partial data every 2 hours as a proxy, and did not use actual manager or 15-minute live entry conditioned odds. The expected return was not forecasted, the current average true range was fixed, and the historical pool could not model unseen losses. A news scan found the seed and simulation details. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data line: a train full_net3R synchronized Monte Carlo simulation for SOL/USD with the following parameters: window is set to train, policy is full_net3R, symbol is SOL, 10000 paths were used, source days were 488, stop probability percentage is 54.32, target probability percentage is 9.85, time probability percentage is 35.83, partial probability percentage is 0, win probability percentage is 37.63, mean return percentage is -0.0392, mean win percentage is 18.10, mean loss percentage is -10.98, p01 return percentage is -14.75, worst sample percentage is -16.94, half Kelly is 0, quarter probe p99 loss in USD is 4.39, target gain percentage is 34.56, and mean exit hours is 300.83. This simulation used seed 91306, 10000 paths, and 480 hours, with 20 synchronized 24-hour complete UTC day blocks sampled by future day of the week starting Sunday. The training data is from 2024-25 and the known holdout pools are from 2026, separate from the training data. The spot initially had an entry and 1.001 accounts with a buy slip, 25 basis point fees, and 10 basis point sell slip. Hourly checks were used, with partial only every 2 hours as a proxy, not actual manager or 15-minute live entry conditioned odds. The expected return is not forecast, the current ATR is fixed, and the historical pool cannot model unseen losses. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data line: a train full_net3R synchronized Monte Carlo simulation for ETH/USD, also known as Ethereum versus the US dollar, was run with the following parameters: the window was set to train, the policy was full_net3R, and the symbol was ETH. The simulation used 10,000 paths and considered 488 days of historical data. The stop probability percentage was 62.24%, the target probability percentage was 14.51%, the time probability percentage was 23.25%, and the partial probability percentage was 0%. The win probability percentage was 33.13%, the mean return percentage was -0.42%, the mean win percentage was 16.08%, and the mean loss percentage was -8.59%. The worst sample percentage was -18.70% and the worst 1% sample return percentage was -13.80%. The half Kelly criterion was 0, and the quarter probe p99 loss in US dollars was $4.11. The target gain percentage was 24.82%. The simulation was run with a seed of 91306, using 10,000 paths and 480 hours of data, with 20 synchronized 24-hour complete UTC day blocks sampled by future day of the week starting on Sunday. The data was split into training and holdout sets, with the training set covering 2024-2025 and the holdout set covering 2026. The simulation assumed an initial spot entry with a 0.1% buy slip and 25 basis point fees plus 10 basis point sell slip. Hourly checks were used, with partial checks every 2 hours, but this was not an actual manager or 15-minute live entry conditioned odds. The expected return was not forecast, the current average true range was fixed, and the historical pool cannot model unseen losses. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data line: a train full_net3R synchronized Monte Carlo simulation for BTC/USD, also known as a computer model that uses random sampling to estimate outcomes, was run with the following parameters: the window was set to train, the policy was set to full_net3R, a type of decision-making rule, and the symbol was BTC. The simulation used 10000 paths, which are like different possible scenarios, and considered 488 days of past data. The results showed a stop probability of 47.93%, a target probability of 8.53%, a time probability of 43.54%, and a partial probability of 0%. The win probability was 43.68%, the mean return was 0.747636942579093%, the mean win was 9.900968568569386%, and the mean loss was -6.351395824169061%. The worst 1% of returns was -8.393692050829705%, and the worst sample was -10.334308390030266%. The half Kelly criterion, a measure of risk, was 0.5944480354894206. The quarter probe p99 loss in US dollars was 2.5000611773396275, and the target gain was 20.52414716141908%. The mean exit time was 326.0381 hours. This simulation, known as Seed 91306, used 10000 paths and 480 hours of data, with 20 synchronized 24-hour blocks, starting on Sunday. The data was split into training and holdout sets, from 2024-2025 and 2026, respectively. The simulation assumed an initial spot entry with a 0.1% buy slip, 25 basis point fees, and a 10 basis point sell slip. Hourly checks were used, with partial checks every 2 hours, but this was not an actual manager or live entry condition. The expected return was not forecasted, the current average true range, a measure of volatility, was fixed, and the historical pool of data cannot model unseen losses. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data: the symbol is SOL, the time frame is 1 hour, and the window is holdout. The sample size is 6, the mean trade percentage is 1.1647773281099318, and the mean confidence interval 95 percentage is between -5.5853095053415105 and 8.97919902663138. The percentage without the best return is -11.421423222400984. There are 1 target and 2 stops, with a maximum stop overshoot percentage of 0.37118211791479255. The subsets are as follows: a group called weekend with a flag set to false, a sample size of 6, a mean percentage of 1.1647773281099323, and a win percentage of 50; a group called weekend with a flag set to true, a sample size of 0, and no mean or win percentage; a group called trend with a flag set to false, a sample size of 5, a mean percentage of 0.2757357664312579, and a win percentage of 40; and a group called trend with a flag set to true, a sample size of 1, a mean percentage of 5.609985136503304, and a win percentage of 100. This is based on a gross return calculation with a delay of 0 and a slip of 10 basis points, using 10000 resamples with a seed of 913061. The trades are non-overlapping and per coin, but the regimes are not independent, and the confidence interval is conditional and descriptive. The simple moving average of 200 uses the previous completed day, with no missing trend promotion. Removing the best return is a sensitivity test, not an instruction to cut winners. https://runway.adityaperswal.workers.dev/data/crypto-bars