Stats

A news scan found the headline 'R91308 fixed prior-close net3R gate: chronological manager-cadence replay'. This is a note from a trading fund manager. The manager uses a system that automatically makes trades. It looks for a 'breakout', which means when a stock's price moves outside a certain range, and also checks the 'SMA50', which is a moving average of the stock's price over 50 days. The system enters a trade when certain conditions are met, including the stock's price being below a certain level, and it only trades during 'RTH', which means regular trading hours on the New York Stock Exchange. The system also has rules for when to exit a trade, including after 40 sessions or when a certain profit is made. The manager tested the system using data from 2020 to 2024 and found that it made a return of 19.190992%, with a maximum drawdown of 49.046502%, and 36.3636% of trades were winners. The system was also stressed, which means it was tested in difficult market conditions, and it made a return of 14.060286%. The manager also tested the system on new data from 2025 and found that it made a return of 1.924385%, with a maximum drawdown of 15.627365%, and 36.3636% of trades were winners. The system made 1682 daily trades and 123209 raw trades, and it was found to be working correctly. The manager used data from a website, https://runway.adityaperswal.workers.dev/data/stock-bars, to test the system.

A test was run using a computer program called backtest.js on the symbols BTC/USD, ETH/USD, and SOL/USD with a breakout strategy, looking back 20 days over 981 days. The results showed a gross return of 46.46% for BTC, 41.34% for ETH, and 72.15% for SOL, with drawdowns of 14.69% for BTC, 24.94% for ETH, and 18.80% for SOL, and win rates of 34.33% for BTC, 32.73% for ETH, and 52.78% for SOL, out of 67 trades for BTC, 55 trades for ETH, and 36 trades for SOL. The test used closing prices to determine when to enter a trade, did not account for fees or holding out, and included partial data from September 13 for 982 daily BTC and ETH trades and 749 SOL trades. This test cannot accurately represent how an actual manager would perform, as it does not account for fees on credited assets, partial exits, or frozen net assets. The idea is that a test result cannot override a negative stress test result. The test compared 60 initial latency cells and 48 managed daily and hourly cells recorded separately, with no option to choose from the command line interface ranking. The data source is https://runway.adityaperswal.workers.dev/data/crypto-bars.

The C91306 actual-cadence extension has 48 predetermined cells. It looks for daily HIGH20 or hourly HIGH480 events that change from false to true, excluding the signal day based on the prior daily ATR20. The first eligible 15-minute open is executed immediately or at the actual manager's UTC schedule, which is daily at 00:02, 04:06, 08, and on weekdays at 12:30, 13:50, 16:30, 19:40, and 21, rounded up to 14:00 and 19:45 when needed. If signals collide, the first event is retained and expires after 2 hours. A delayed price must still exceed the frozen high. The stop is set at 2ATR, and the sell quarter is moved to the entry only at the selected cadence with a +2 gross return. The residual target is solved beforehand for an aggregate net return of 3R, including 25 basis points credited-asset fees and 10/30 basis points slippage. Stops and targets are checked every 15 minutes, and overshoots are charged. A 20-day time exit waits for the selected cadence. Each train from 2024 to 2025 and the known 2026 holdout from January 1 to September 12 starts flat, with the final available close liquidation costed. The vectorbt from_orders ledger cash agrees to less than 1e−5. There are 8 cells with different modes, windows, slippage, return percentages, drawdown percentages, win percentages, and numbers. The results are: immediate 15-minute train with 10 basis points slippage had a 60.82% return, immediate 15-minute holdout with 10 basis points slippage had a 4.74% return, and so on. A news scan found the conclusion that promoting this rule now is rejected due to all daily manager holdouts being negative, and hourly BTC/SOL turning negative at 30 basis points slippage, with ETH manager already negative. No fresh entry exists. The sources are https://runway.adityaperswal.workers.dev/data/crypto-bars, https://docs.alpaca.markets/us/docs/crypto-fees, and the user's UTC schedule.

The C91306 actual-cadence extension has 48 predetermined cells. It looks for daily HIGH20 or hourly HIGH480 events, where the prior daily ATR20 does not include the signal day. The first eligible 15m open is executed immediately or at the actual manager UTC schedule, which is daily at 00:02, 04:06, 08:00, and weekdays at 12:30, 13:50, 16:30, 19:40, and 21:00, rounded up to 14:00 and 19:45 where needed. A stop loss is set at 2 ATR, and a sell order is placed at a 2 gross R quarter, moving the stop to the entry point only at the selected cadence. The residual target is solved ex ante for an aggregate net 3R, including 25bp credited-asset fees and 10/30bp slippage. Stops and targets are checked every 15m open, and overshoots are charged. A 20-day time exit waits for the selected cadence. Each train from 2024-25 and known 2026 Jan 1 - Sep 12 holdout starts flat, with the final available close liquidation costed. The results are as follows: All 8 cells have the following modes, windows, slipbp, return%, DD%, win%, N, and approx half Kelly: immediate15 train with 10bp slip, 46.256440014381425% return, 27.844944199800736% drawdown, 55.55555555555556% win rate, 9 trades, and 1.8893496247368675 approx half Kelly; immediate15 holdout with 10bp slip, -23.58855665335652% return, 36.66451078761456% drawdown, 28.57142857142857% win rate, 7 trades, and 0 approx half Kelly; immediate15 train with 30bp slip, 41.52147769142425% return, 28.203289819312783% drawdown, 55.55555555555556% win rate, 9 trades, and 1.7329408431566498 approx half Kelly; immediate15 holdout with 30bp slip, -24.81196132288119% return, 37.555842224623824% drawdown, 14.285714285714285% win rate, 7 trades, and 0 approx half Kelly; manager train with 10bp slip, 46.51209060420214% return, 27.82868159430437% drawdown, 55.55555555555556% win rate, 9 trades, and 1.8927562089076304 approx half Kelly; manager holdout with 10bp slip, -23.114313182756362% return, 36.66451078761456% drawdown, 28.57142857142857% win rate, 7 trades, and 0 approx half Kelly; manager train with 30bp slip, 41.689738130383304% return, 28.186901760039472% drawdown, 55.55555555555556% win rate, 9 trades, and 1.7353182387132127 approx half Kelly; manager holdout with 30bp slip, -24.441814135943204% return, 37.555842224623824% drawdown, 14.285714285714285% win rate, 7 trades, and 0 approx half Kelly. A news scan found the conclusion that this rule should not be promoted now, as all daily manager holdouts are negative, and hourly BTC/SOL turn negative at 30bp slip, while ETH manager is already negative. No fresh entry exists. The sources of the data are https://runway.adityaperswal.workers.dev/data/crypto-bars, https://docs.alpaca.markets/us/docs/crypto-fees, and the user UTC schedule.

The C91306 actual-cadence extension has 48 predetermined cells. It looks for daily HIGH20 or hourly HIGH480 events that change from false to true, excluding the signal day based on the prior daily ATR20. The first eligible 15m open is executed immediately or at the actual manager UTC schedule, which includes daily times 00, 02, 04, 06, 08, and weekdays 12:30, 13:50, 16:30, 19:40, 21, rounded up to 14:00 or 19:45 if needed. If signals collide, the first event is retained and expires after 2 hours, with the condition that the delayed price must exceed the frozen high. The stop loss is set at 2ATR, and a sell order is placed at a quarter of the gross return, moving the stop to the entry point only at the selected cadence. The residual target is calculated to achieve an aggregate net return of 3R, including 25bp credited-asset fees and 10/30bp slippage. Stops and targets are checked every 15m open, with overshoots charged, and a 20-day time exit waits for the selected cadence. Each training period from 2024 to 2025 and the known holdout period from January 1, 2026, to September 12, 2026, starts flat, with the final available close liquidation costed. The vectorbt from_orders ledger cash agrees with the results to less than 1e−5. The results for all 8 cells, which include mode, window, slippage in basis points, return percentage, drawdown percentage, win percentage, number of trades, and approximate half Kelly, are as follows: immediate15 train with 10bp slippage had a return of 41.81%, immediate15 holdout with 10bp slippage had a return of 14.28%, immediate15 train with 30bp slippage had a return of 39.68%, immediate15 holdout with 30bp slippage had a return of 11.93%, manager train with 10bp slippage had a return of 41.33%, manager holdout with 10bp slippage had a return of -4.76%, manager train with 30bp slippage had a return of 35.67%, and manager holdout with 30bp slippage had a return of -7.18%. Increasing slippage also affects fill-anchored exits and subsequent occupancy, so stress loss is not just subtracting fees. There are missing 15m BTC and ETH slots, as well as SOL slots from August 26, 2024, due to incomplete training. Missing quotes skip action, with no assumed fill. The full sample missed-event counter includes pre-2024 warmup and is not a within-window rate. This is an idealized bar-open execution model, not archived bids or asks or fill proof, and exact run-duration delays and network outages remain unmodeled. The conclusion is to reject promoting this rule now, as all daily manager holdouts are negative, and hourly BTC and SOL turn negative at 30bp slippage, with ETH manager already negative. No fresh entry exists. The sources of the data are https://runway.adityaperswal.workers.dev/data/crypto-bars, https://docs.alpaca.markets/us/docs/crypto-fees, and the user UTC schedule.

The C91306 actual-cadence extension has 48 predetermined cells. It looks for daily HIGH20 or hourly HIGH480 events that change from false to true, excluding the signal day if the daily ATR20 is high. The first eligible 15-minute open is executed immediately or at the actual manager's UTC schedule, which is daily at 00:02, 04:06, 08, and weekdays at 12:30, 13:50, 16:30, 19:40, and 21. If signals collide, the first event is kept and expires after 2 hours. The delayed price must still be higher than the frozen high. The stop loss is set at 2ATR, and the sell quarter is moved to the entry point at the selected cadence with a +2 gross return. The residual target is calculated before the fact, including 25 basis points of credited asset fees and 10/30 basis points of slippage. Stops and targets are checked every 15 minutes, and overshoots are charged. A 20-day time exit waits for the selected cadence. Each training period from 2024 to 2025 and the known holdout period from January 1, 2026, to September 12, 2026, starts flat, and the final available close is liquidated with costs. The vectorbt from_orders ledger cash agrees to less than 1e-5. There are 8 cells with different modes, windows, slippage, returns, drawdowns, win percentages, and numbers. The conclusion is to reject promoting this rule now because all daily manager holdouts are negative, and hourly BTC/SOL turn negative at 30 basis points of slippage, while ETH manager is already negative. No fresh entry exists. The sources of the data are https://runway.adityaperswal.workers.dev/data/crypto-bars, https://docs.alpaca.markets/us/docs/crypto-fees, and the user's UTC schedule. A news scan found the data line with the given information.

A news scan found the headline 'C91306 actual-cadence extension, 48 predetermined cells.' This refers to a set of rules for an automated trading fund manager. The rules are based on daily and hourly high prices, with specific conditions for when to buy or sell. A daily HIGH20 or hourly HIGH480 is a measure of the highest price in a given time period, which is used to trigger events. The actual manager schedule is used to determine when to execute trades, with specific times on weekdays. The first event is retained if multiple signals occur at the same time, and it expires after 2 hours. A stop loss, which is a price level at which to automatically sell, is set at 2 times the average true range, or ATR, which is a measure of price volatility. The target price is calculated to achieve a net return of 3R, which is a measure of profit, including fees and slippage, which is the difference between the expected and actual price of a trade. The results are checked every 15 minutes, and any overshoots are charged. The training data is from 2024-2025 and the holdout data is from January 1, 2026 to September 12, 2026. The vectorbt ledger, which is a record of trades, agrees with the cash ledger to within a very small margin of error, less than 1e−5. The results are presented in a table with 8 cells, showing the mode, window, slippage, return, drawdown, win percentage, number of trades, and approximate half Kelly, which is a measure of the optimal fraction of wealth to invest. The conclusion is that the rule is not promoted due to negative results in the holdout data, and no fresh entry exists. The sources of the data are https://runway.adityaperswal.workers.dev/data/crypto-bars and https://docs.alpaca.markets/us/docs/crypto-fees, and the user's UTC schedule.

The C91306 trading strategy has an actual-cadence extension with 48 predetermined cells. It looks for daily HIGH20 or hourly HIGH480 events where the prior daily ATR20 excludes the signal day. The strategy executes the first eligible 15-minute open immediately or at the actual manager's UTC schedule, which is daily at 00:02, 04:06, 08, and weekdays at 12:30, 13:50, 16:30, 19:40, and 21. If signals collide, the first event is retained and expires after 2 hours. A delayed price must still exceed the frozen high. The strategy uses a stop of 2ATR and sells a quarter of the position with a gross return of +2, moving the stop to the entry point only at the selected cadence. The residual target is solved ex ante for an aggregate net return of 3, including 25 basis points credited-asset fees and 10/30 basis points slippage. Stops and targets are checked every 15 minutes, and overshoots are charged. A 20-day time exit waits for the selected cadence. Each training period from 2024 to 2025 and the known holdout period from January 1, 2026, to September 12, 2026, starts flat, and the final available close is liquidated with costs. The Vectorbt from_orders ledger cash agrees with the results to less than 1e−5. The strategy has 8 cells with different modes, windows, slippage, returns, drawdowns, win percentages, and numbers of trades. The results show that increasing slippage changes fill-anchored exits and subsequent occupancy. The strategy is missing 15-minute data for BTC and ETH, and SOL data from August 26, 2024, is incomplete. The conclusion is to reject promoting this rule now because all daily manager holdouts are negative, and hourly BTC and SOL turn negative at 30 basis points slippage, while ETH manager is already negative. No fresh entry exists. The sources of the data are https://runway.adityaperswal.workers.dev/data/crypto-bars, https://docs.alpaca.markets/us/docs/crypto-fees, and the user's UTC schedule.

A news scan found the text of a trading fund manager's notes. The manager uses a system called cron_manager to make decisions. This system looks at the price of a stock, specifically the L20 prior-close breakout and the SMA50, which is a type of moving average, a way to smooth out price changes over 50 time periods. The manager also looks at the ADR20, which is the average daily range, the difference between the high and low prices, over 20 days. The manager uses this information to decide when to buy or sell a stock. The manager buys when the price breaks out, or moves above, the prior close and the SMA50. The manager sells when the price reaches a target, which is 6 times the average daily range, or when it falls to a stop, which is 2 times the average daily range. The manager only buys during certain times of the day, specifically at 13:50 UTC during the summer and 16:30 UTC during the winter. The manager also sells at specific times, 12:30, 13:50, 16:30, 19:40, and 21 UTC. The manager forces a sale on the 40th session, which is a trading period. The manager uses 15-minute and 5-minute price data to check the trades. The manager does not account for dividend cash, which is money paid to shareholders. The manager pays fees of 1 basis point, a unit of measurement, to buy and sell, and 5 basis points for slipping, or failing to execute a trade. The manager also stresses, or tests, the trades with a 15 basis point change. The manager uses a system called vectorbt to track the trades and calculate the returns. The manager trained the system from 2020 to 2024 and got a return of 36.189046%. The manager also tested the system from 2025 to September 11 and got a return of 17.407005%. The manager notes that the mark, or current price, is not the actual exit price and should be kept separate from the win rate. The manager has an open entry, or trade, on July 24, 2026, with a target price of 48.545919999999995. The manager explains that the difference between the old target and the new target is due to a change in the average daily range. The manager notes that there are missing observations, or data points, and that the daily portfolio size is 100% of the research benchmark, not the account allocation. The manager also notes that the timing changes and recent returns should not be used to choose the control solely for the best holdout. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars.

A news scan found the body of a report from an automated trading fund manager. The report describes a test of a trading strategy called R91306 TEN. The strategy uses a combination of technical indicators, including a prior-close breakout and a 50-period Simple Moving Average, or SMA50, which is a calculation of the average price of a stock over 50 trading days. The strategy also uses an Average Daily Range, or ADR, which is a measure of the average daily price movement of a stock. The test had specific rules, including a stop-loss order at 2 times the ADR and a target price at 6 times the ADR, with a maximum of 40 trading sessions. The test started at a fixed time, 13:50 UTC during summer and 16:30 UTC during winter, which are times when the New York Stock Exchange, or XNYS, is open for trading, known as Regular Trading Hours, or RTH. The test used real-time market data, including 15-minute and 5-minute opening prices, but the data was not always synchronous, meaning it was not always available at the same time. The test also took into account certain market events, such as stock splits, but did not account for dividend payments. The test included fees of 1 basis point, or 0.01%, per trade, and a slippage of 5 basis points, or 0.05%, which is the difference between the expected price of a trade and the actual price. The results of the test showed a return of 10.796047% from 2020 to 2024, with a maximum drawdown, or loss, of 72.455395%, and a win rate of 32.2581%. A separate test, called a holdout, from 2025 to September 11, showed a return of 17.542708%, with a win rate of 45.4545%. The holdout test also showed a stress, or risk, of 38.065430%, and a maximum drawdown of 17.094922%. The report notes that the results are based on a mark, or value, of 47.93 on September 11 at 19:55 UTC, which is not the actual exit price, but rather an estimate of the value of the portfolio at that time. The report also notes that the results are subject to certain limitations, including missing data and assumptions about trading activity. The data for the test came from a source at https://runway.adityaperswal.workers.dev/data/stock-bars.

A trading fund manager tested a strategy called R91306 TEN open_manager. This strategy used a split-unit approach and costed chronological execution. It looked at the stock's prior close breakout and its 50-day simple moving average, or SMA50, which is a measure of a stock's average price over 50 days. The strategy also considered the stock's average daily range, or ADR, over 20 days. The test had a stop loss of 2 times the average daily range and a target of 6 times the average daily range, with a maximum of 40 trading sessions. The test started at 9:30 ET and had specific times to sell or adjust the trade. The test used 15-minute and 5-minute opening prices for checks and did not include dividend cash payments. The test also accounted for fees of 1 basis point per side and a 5 basis point slip, with a stress test of 15 basis points. The results of the test from 2020 to 2024 showed a return of 0.622574% and a maximum drawdown of 66.328898%. A separate test from 2025 to September 11 showed a return of 32.871082%. The current open position has a target of 48.365919999999996 as of July 24, 2026, at 13:30:00 UTC. The data for this test came from https://runway.adityaperswal.workers.dev/data/stock-bars.

This is a test of a trading strategy called R91306 TEN. The strategy uses a combination of rules to decide when to buy or sell stocks. It looks at the stock's price at the end of the previous day and its average price over the last 50 days, which is called a Simple Moving Average or SMA. The strategy also considers the stock's price range over the last 20 days, which is called the Average Daily Range or ADR. The test started with a set of rules, including a stop-loss order at 2 times the ADR and a target price at 6 times the ADR, and a limit of 40 trading sessions. The test began at 9:30 ET and the strategy was set to sell at a 2.25% profit or at a break-even point every 15 minutes, without any human intervention. The test also included a rule to exit the trade on the 40th session. The data used for the test included 15-minute and 5-minute opening prices, which were not necessarily synchronous quotes. The test excluded half-days and after-hours trading sessions and did not account for dividend payments. The test also included fees of 1 basis point per side and a 5 basis point slippage, with a stress test of 15 basis points. The results of the test from 2020 to 2024 showed a return of -5.87%, a maximum drawdown of 66.47%, and a win rate of 41.38%. The results from 2025 to September 11 showed a return of 43.97% and a win rate of 66.67%. The test also included a mark-to-market calculation, which showed a return of 71.81% and a maximum drawdown of 18.83%. The mark is not the actual exit price, but rather an estimate of the value of the position at a given time. The test used data from the XNYS calendar and excluded dividend payments. The data source is https://runway.adityaperswal.workers.dev/data/stock-bars.

A trading fund manager tested a strategy called R91306 TEN. This strategy used a method called a moving average, or SMA50, and a breakout, which is when a stock price moves past a certain level. The test also used something called an ADR, or average daily range, which is the average amount a stock price moves in a day. The test started at 9:30 ET and only automatic exits were allowed. The test used real-time data, but not all of it was synchronized. The test also took into account things like dividends and fees, which are costs associated with buying and selling stocks. The results of the test from 2020 to 2024 showed a return of -7.421094% and a maximum drawdown, or loss, of 72.280462%. A holdout test, which is a test of the strategy on new data, from 2025 to September 11 showed a return of 32.770053%. The current open position has a target price of 48.365919999999996, which was set on July 24, 2026, at 13:30:00 UTC. A news scan found the source of the data to be https://runway.adityaperswal.workers.dev/data/stock-bars.

A news scan found the headline that Stitch Fix announced its fourth quarter and full fiscal year 2026 financial results and conference call date. The automated trading fund manager analyzed historical option trades for SFIX from March 2024 to June 2026. The analysis used 10 consecutive issuer-scheduled AMC earnings, with a daily bars adjustment of 16.578014% from release-close to next-session-close. The manager considered options such as Sep11S2.95 and Oct16C2.5, with a catalog price of $0.50 on Sep 10 and an assumed round-trip cost of $0.05 per share, requiring a 3.389831% return. The manager also looked at C5.05 on Sep 10, which required a 72.881356% return, and P2.5 on Sep 12, which required a -21.016949% return. The manager noted that the mean intrinsic value of the options, excluding residual time, was 1.109296 times the premium, but not the actual option expected value. The manager found that there was no documented 3x expected payoff. The analysis used data from option contracts SFIX, stock bars, and option bars. The manager used a vectorbt system with one whole contract per available event and $1000 initial cash. The manager assumed costs, including a $0.05 per share round-trip cost, and a 0.125% fee and 0.125% price slippage. The manager also ran a stress test with a 0.10 total split, 0.025% fee, and 0.025% slippage per side. The results showed that the train mean return was 4.529617%, but the normalized full reinvestment return was -2.204002%, with a drawdown of 51.121951%. The stress test mean return was -15.218254%, with a normalized return of -76.182143%. The manager concluded that the results did not establish an out-of-sample edge and rejected live adoption. The manager also analyzed the return on variable entry capital and found that it differed from the train mean return. The manager verified endpoint observations and concluded that the option edge was inconclusive.

A news scan found the headline 'C91306 SOL hourly: delayed gross3R breakout control'. This refers to a trading strategy for the cryptocurrency SOL, where a 'gross3R' means a target return of three times the risk, and 'breakout control' is a method to manage risk. The strategy uses a'stop', which is a price level at which to sell a security, and a 'target', which is a price level at which to buy a security. The strategy also uses 'ATR20', which stands for Average True Range over 20 days, a measure of price volatility. Each training period starts with no existing positions, and the strategy uses a'reindex' to adjust the calendar and ensure there are no missing data points. The strategy freezes the ATR calculation, excluding the signal, at the start of each day, and only buys if the price exceeds the previous high. The strategy sells at the start of each hour if the price has exceeded the target, and includes a 'terminal close cost', which is the cost of closing a trade. The strategy also includes a 'coin buy fee' of 25 basis points, and a'sell fee' of 25 basis points plus a'slip' of 10 or 30 basis points, which is the difference between the expected and actual price of a trade. The results are summarized in arrays, including the window of time, delay in hours, slip in basis points, return percentage, drawdown percentage, win percentage, number of trades, and a 'half Kelly' measure, which is a formula to determine the optimal size of a trade. The results are as follows: for the training period, the return was 25.81%, the drawdown was 38.76%, and the win percentage was 54.55%, and for the holdout period, the return was 4.70%, the drawdown was 22.48%, and the win percentage was 50%. 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 'C91306 ETH hourly: delayed gross3R breakout control'. This refers to a trading strategy for Ethereum, a cryptocurrency. The strategy uses a 'gross3R breakout control', which means it aims to make a return of three times the risk taken. The strategy is backtested on historical data from 2024 to 2025, and evaluated on data from 2026. Each test starts with no existing position, and uses a'stop' based on the average true range of the price over the last 20 days. The strategy targets a return of three times the risk taken, and exits positions at the hourly open price if the price overshoots the target. The results are summarized in arrays, which include the window of time, delay hours, slip in basis points, return percentage, drawdown percentage, win percentage, number of trades, and a metric called 'half Kelly'. The arrays show the results for different combinations of these parameters. The strategy is not authorized for live trading, and is compared to a separate 15-minute manager that uses the same policy. The results are also compared to a 'net3R' target, which is not the same as the 'gross3R' target used here. A 'net3R' target would be a return of three times the risk taken, after fees and other costs. 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 Bitcoin trading strategy. The strategy had predeclared HIGH480 false-to-true events, which means it looked for specific price movements, 2 prior daily Average True Range, or ATR, which is a measure of price volatility, 20 stop, gross 3R target, which means the strategy aimed to make three times the risk, and 20-day exit. Each training period from 2024 to 2025, and a known holdout period from January 1, 2026, to September 12, 2026, started with no existing positions, had a warm-up period in 2023, and used a full calendar reindex with no missing bar signals. The strategy used a freeze on Average True Range excluding signal on UTC day, and a delayed open that must exceed the frozen high. It exited at hourly opens with overshoot, and terminal close was costed, meaning it included all fees. The strategy credited 25 basis points for coin buy fees, and 25 basis points for sell, plus 10 or 30 basis points for slippage, which is the difference between expected and actual trade prices. The results were reconciled using vectorbt from orders final cash, which is a method to calculate the final cash balance, and the difference was less than 1e−5, which is a very small number. The summary arrays showed the results for different scenarios, including window, delay hours, slip basis points, return percentage, drawdown percentage, win percentage, number of trades, and half Kelly, which is a measure of the optimal bet size. The results were: [[

Diagnostic C91306 is a test with 60 predeclared cells. It looks for a false-to-true close above prior 20 calendar day highs on a daily basis or 480 hour highs on an hourly basis. The training period is from 2024 to 2025, and the known period is from January 1, 2026, to September 12, 2026, with a warm-up period in 2023. The system freezes the Average True Range (ATR) over 20 periods, excluding signal times on UTC days, and enters at the next open or delays entry. If the delayed open is less than or equal to the frozen high, the entry is cancelled. The stop entry is set at -2 ATR, the target entry at +6 ATR, and the time limit is 20 days. The system checks at each sampled open and charges for gap overshoots and terminal close costs. The gross 3R target is not the same as the net 3R or the mandated quarterly/BE policy, and this is not a live authorization. The system credits 25 basis points for coin buy fees, 25 basis points for sell fees, and 10/30 basis points for slippage, with a custom order ledger that reconciles to less than 1e-5. There is no fill-forward for missing prices in signals, and the calendar is reindexed. The SOL training is incomplete. The results are as follows: there are seven sets of results with different parameters, including return, drawdown, win rate, sample size, mean trade percentage, half Kelly fraction, skipped entries, days, symbol, time frame, delay, slippage in basis points, and window. The models ignore historical quotes, depth, and vintages, and idealize bar-open execution. The data comes from https://runway.adityaperswal.workers.dev/data/crypto-bars and https://docs.alpaca.markets/us/docs/crypto-fees.

A diagnostic report C91306 was generated with 60 predeclared cells. The report indicates a fresh false-to-true close above prior 20 calendar day highs on a daily basis or 480 hour highs on an hourly basis. The training period is from 2024 to 2025, and the known holdout period is from January 1, 2026, to September 12, 2026. Each period starts with a flat position. The report also mentions a warm-up period in 2023. The trading strategy involves freezing the Average True Range (ATR) 20, excluding signal UTC day, and entering at the next open or delaying entry. If the delayed open is less than or equal to the frozen high, the entry is cancelled. The stop entry is set at -2 ATR, and the target entry is set at +6 ATR, with a time limit of 20 days. The report checks for these conditions at each sampled open and charges for gap overshoots and terminal close costs. The gross 3R target is not the same as the net 3R or mandated quarter/BE policy, and this is not a live authorization. The report also mentions a 25 basis point buy fee and a 25 basis point sell fee, with a 10/30 basis point slippage. The custom order ledger reconciles with a final cash value of less than 1e-5. There is no fill-forward missing prices for signals, and the calendar is reindexed. The SOL training is incomplete. The report includes a list of results for different scenarios, including returns, drawdowns, win rates, and mean trade percentages. The results are for the symbol ETH, with a time frame of 1 day, and different delays and slippage values. 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 report, labeled C91306, analyzed 60 cells of data. The report looked for instances where the closing price was higher than the highest price in the previous 20 calendar days or 480 hours. The training period was from 2024 to 2025, and the known period was from January 1, 2026, to September 12, 2026. Each period started with no prior knowledge, and there was a warm-up period in 2023. The report used a method called ATR20, which stands for Average True Range over 20 periods, excluding signal times on UTC days. Entries were made at the next open or delayed, and delayed opens were canceled if they were less than or equal to the frozen high. The stop entry was set at 2 ATRs below the entry point, and the target entry was set at 6 ATRs above the entry point, with a time limit of 20 days. The report checked for these conditions at each sampled open and accounted for gap overshoots and terminal close costs. The target was a gross 3R, which means a three times return, not a net 3R, and it did not follow a mandated quarter or best efforts policy. This was not a live authorization. The report included a 25 basis point buy fee and a 25 basis point sell fee, with a 10 to 30 basis point slippage. The results were as follows: a list of trading results for Bitcoin, including return, drawdown, win rate, sample size, mean trade percentage, half Kelly fraction, skipped entries, days, symbol, time frame, delay, slippage in basis points, and window. The results showed varying returns and drawdowns for different scenarios. The data source was https://runway.adityaperswal.workers.dev/data/crypto-bars, and the documentation for crypto fees was https://docs.alpaca.markets/us/docs/crypto-fees. The models ignored historical quotes, depth, and vintages, and idealized bar-open execution. A separate experiment compared the 15-minute manager.

A diagnostic test was run using the command line interface with the following settings: the symbol TEN, a breakout strategy, a lookback period of 20 days, 1825 days of data, a stop loss of 5%, and a target profit of 15%. The test resulted in 1250 bars and 95 trades, with a return of 60.321944%, a maximum drawdown of 18.279506%, and a win rate of 36.842105%. For comparison, buying and holding the stock would have resulted in a return of 493.742255%, with an average trade duration of 0.467107%. This test was a reproduction and not a held-out validation, meaning it was not an independent test. The test also differed from another strategy that used a simple moving average and other indicators. A separate study using the vectorbt library is pending. The data for the test came from https://runway.adityaperswal.workers.dev/data/stock-bars. No parameters were selected from this diagnostic test, which was only for diagnostic purposes and not for approving trades.

A test was run using a computer program called a node with a script called backtest.js. The test looked at three cryptocurrency pairs: BTC/USD, ETH/USD, and SOL/USD, using a strategy called sma-cross, which is a type of trading strategy that uses moving averages. The test used a fast moving average of 1 day and a slow moving average of 200 days, and it looked at 981 days of data. The results were: for BTC/USD, a gross return of 14.4761, a maximum drawdown of 31.2332, a win rate of 15.3846, and 13 closed trades; for ETH/USD, a gross return of 10.7680, a maximum drawdown of 37.8551, a win rate of 11.1111, and 9 closed trades; and for SOL/USD, a gross return of 8.3228, a maximum drawdown of 32.9545, a win rate of 0, and 3 closed trades. The test used the same closing price for execution and did not include fees or holdouts. The data used for the test included 982 bars for BTC and ETH, and 749 bars for SOL, including a partial bar from September 13, and the open and mark-to-market prices explain the positive total return for SOL despite having no closed winning trades. The test cannot implement a stable coin signal and only uses mandatory command line interface control. The costed daily and hourly data, as well as the holdouts, were recorded separately. There were no orders or choices made from the gross output. The data source is https://runway.adityaperswal.workers.dev/data/crypto-bars.

A news scan found the text describing a trading strategy for the cryptocurrency SOL. The strategy uses a filter based on the growth of the USDC stablecoin supply and the price of SOL. The filter is applied to the SOL price and is used to decide when to buy or sell. The strategy is tested on historical data from 2024 to 2025 and is evaluated on its performance during the period from January 1, 2026, to September 12, 2026. The results show that the strategy had a return of -11.006206% during the training period and 5.433297% during the evaluation period. The strategy also had a maximum drawdown of 34.649737% during the training period and 10.758956% during the evaluation period. The text also mentions that the strategy is not suitable for actual trading due to the lack of a stop-loss or target profit policy. The data used for the strategy is sourced from several websites, including https://runway.adityaperswal.workers.dev/data/crypto-bars, https://stablecoins.llama.fi/stablecoin/2, https://stablecoins.llama.fi/stablecoin/1, and https://docs.alpaca.markets/us/docs/crypto-fees.

A news scan found the headline 'C91304 SOL 1Day: stablecoin filter versus price-only control'. This refers to a test of a trading strategy for the cryptocurrency SOL. The strategy uses a 'frozen baseline' which means it only trades when two conditions are met: the 30-day supply growth of the USDC stablecoin is greater than 0, and the previous day's closing price is higher than the 200-day moving average, a calculation of the average price over the last 200 days. When these conditions are true, the strategy enters a trade at the next available opening price and exits when they are no longer true. The test was run from 2024 to 2025, with a holdout period from January 1 to September 12, 2026, and a warm-up period in 2023. The results showed a return of -11.006206% during the training period, with a maximum drawdown of 33.249904%, and no winning trades. In the holdout period, the return was 5.433297%, with a maximum drawdown of 9.663307%, and 100% winning trades. A control strategy that only looks at price movements had a return of 18.366768% in the holdout period. The test used a 25 basis point fee and a 10 basis point slippage per trade, and assumed a 30 basis point slippage for stress testing. The results are based on data from several sources, including crypto-bars and stablecoin data from llama.fi, and crypto fees from Alpaca markets.

A news scan found the text: Frozen baseline: lag2 calendar-day USDC 30-day supply growth greater than 0 AND previous completed daily close greater than SMA200, which is a moving average of the last 200 days. Enter first available bar open when state becomes true; exit first available open when false. Split starts flat and initializes if state true, terminal liquidation at last close. Training from 2024 to 2025, known holdout from January 1, 2026, to September 12, 2023, warmup, no partial September 13 bar. Vectorbt 100 USD all-in single-symbol independent portfolios, 25 basis points fee plus 10 basis points slip per side; stress 30 basis points slip. Sweeps lag 1, 2, and 7 and combined USDC plus USDT, control price-only; no parameter selection from holdout. Baseline training return 75.031958 percent, drawdown 35.057611 percent, win 33.333 percent, number of trades 9; holdout return 3.666842 percent, drawdown 6.462538 percent, win 100.000 percent, number of trades 2. Price-only holdout 11.301377 percent, number of trades 1. Daily signal mapped to hourly opens; identical fills when 00 UTC exists, hourly marks expose greater intraday drawdown. This is not independent replication or an hourly supply series. Current-vintage supply lacks publication archive; assumed lags cannot remove revisions. Fees modeled in quote currency are a close proxy, not exact venue credited-coin accounting. No stops or targets or plus 2R partial or trailing policy here, so baseline cannot authorize trading; incremental filter is not adopted. SOL has 418 missing calendar days across 2023 to 2024; missing SMA means unavailable, not bearish; trade counts small. Half-Kelly recorded only where both wins and losses exist, uncapped fractions unstable; allocation zero. All 28 cells recorded separately in observations. Sources https://runway.adityaperswal.workers.dev/data/crypto-bars; https://stablecoins.llama.fi/stablecoin/2; https://stablecoins.llama.fi/stablecoin/1; https://docs.alpaca.markets/us/docs/crypto-fees.