A news scan found the headline about a trading strategy. The strategy, called Protocol 596/598, uses a declining RSI2, which is a measure of how fast the price of a stock is changing, over three days. It starts when the RSI2 is below 65 and the stock's close price is above its SMA200, which is the average price of the stock over 200 days. The strategy enters the market at the next open, with no overlapping lots, meaning it does not buy or sell the same stock at the same time. The initial stop, which is the price at which the strategy sells the stock to limit losses, is set at one ATR, or average true range, which is a measure of how much the stock's price typically moves. The target price, at which the strategy sells the stock for a profit, is set at 1.5 ATR. The strategy also uses a percentage trail, which means the stop price moves up as the stock's price increases. The strategy exits the market when the RSI2 is greater than 75 or after 10 sessions, which are periods of trading. The strategy was trained on data from 2016 to 2020 and tested on data from 2021 to 2025, and it was applied to recent data from 2026 through September 4. The results show that the strategy had a mean return of -0.783671%, a worst return of -1.579315%, and a median holding period of 0.8958 calendar days. The conclusion is to reject the adoption of this strategy because no ETF, or exchange-traded fund, passed both the main training and holdout tests. The strategy uses historical proxies, which are models of how the market has behaved in the past, but these proxies do not include executable quotes, real polling clock, exact dividend cash, and whole-share rounding, which can affect the results. The sources of the data are https://runway.adityaperswal.workers.dev/data/stock-bars and https://tradingmarkets.com/recent/the_improved_r2_strategy_84_correct_with_just_6_rules_-674361.
A news scan found the following report: The Protocol 596/598 has a declining RSI2, which is a measure of how quickly a stock's price is changing, over three days, starting below 65 and close above the SMA200, which is the average price of a stock over 200 days. The next-open entry will be made without overlapping lots. The Gate is set at 1.5 ATR20, which is a measure of volatility, and is less than or equal to 2 ADR20. The initial stop will be set at one ATR, which is a measure of volatility, and the target will be set at 1.5 ATR. The position will be closed if the RSI is greater than 75 or after a 10-session timeout. The holdout period is from 2021 to 2025, and the recent period is from 2026 through September 4. The vectorbt fee is 5 basis points, which is a measure of the cost of trading, and the slippage is also 5 basis points. The results show a mean return of -0.415034%, a worst return of -2.522052%, and a median holding period of 1.2292 calendar days. The conclusion is to reject the adoption of this strategy because no ETF passes both the main training and holdout periods. The historical proxies used in this study do not account for executable quotes, real-time polling clock, exact dividend cash, and whole-share rounding. Currently, the SPY and IWM ETFs cannot be traded in whole shares. The study also notes that the periods inspected were under other rules, which affects the independence of the results. Earlier exits permit re-entry and change cohort counts, but matched diagnostics separate that effect. The sources of the data are https://runway.adityaperswal.workers.dev/data/stock-bars and https://tradingmarkets.com/recent/the_improved_r2_strategy_84_correct_with_just_6_rules_-674361.
A news scan found the text of a trading fund manager's notes. The notes discuss a trading strategy called Protocol 596/598, which involves a declining RSI2, or a measure of how quickly a stock's price is changing, over three days. The strategy requires the stock to start below 65 and close above its SMA200, which is the average price of the stock over 200 days. The next day, the strategy enters a trade with no overlapping lots, meaning it does not buy or sell the same stock at the same time. The strategy uses a gate of 1.5 ATR20, which is a measure of how volatile the stock is, and sets an initial stop, or the price at which the trade is automatically sold, at one ATR, and a percentage trail, which is a way of adjusting the stop price as the stock's price changes. The strategy also sets a target price at 1.5 ATR. The trade is closed when the RSI is greater than 75 or after 10 sessions. The notes also discuss a train, which is a set of data used to test the strategy, and a holdout, which is a set of data used to validate the results. The train data is from 2016 to 2020, and the holdout data is from 2021 to 2025. The strategy is tested on recent data from 2026 through September 4. The results show a mean return of -0.099752%, a worst return of -1.743545%, and a median holding period of 1.3594 calendar days. The conclusion is to reject the adoption of the strategy because no ETF, or exchange-traded fund, passes both the main training and holdout tests. The notes also mention that historical proxies, or models, omit executable quotes, real polling clock, exact dividend cash, and whole-share rounding, which are all important factors in trading. Additionally, the notes mention that the ETFs SPY and IWM cannot currently fit one whole share. The periods were inspected under other rules, so the chronology does not establish pristine independence. Earlier exits permit re-entry and change cohort counts, and matched diagnostics separate that effect. The sources of the data are https://runway.adityaperswal.workers.dev/data/stock-bars and https://tradingmarkets.com/recent/the_improved_r2_strategy_84_correct_with_just_6_rules_-674361.
A news scan found the headline about a trading strategy. The strategy, called Protocol596/598, uses a declining RSI2, which is a measure of how quickly a stock's price is changing, over three days. It starts when the RSI2 is below 65 and the stock's close is above its SMA200, which is the average price of the stock over 200 days. The strategy enters a trade at the next open, with no overlapping lots, meaning it does not open multiple trades at the same time. The initial stop is set at one ATR, which is a measure of how volatile the stock is, and the target is set at 1.5 ATR. The strategy also uses a percentage trail, which means the stop moves as the stock's price moves. The trade is closed when the RSI2 is greater than 75 or after 10 sessions. The strategy was trained on data from 2016 to 2020 and tested on data from 2021 to 2025, and recent data from 2026 up to September 4 was also used. The strategy uses a long-only approach, meaning it only buys stocks, and it takes into account fees and slippage, which is the difference between the expected price of a trade and the actual price. The strategy was evaluated using raw 15-minute close checks and adjusted-ratio wealth, which takes into account the stock's price and the number of shares held. The drawdown, which is the maximum loss of the strategy, was marked at observed intraday prices. The results of the strategy were: N7, mean 0.628881%, worst -1.416834%, median 2.0313 calendar days. The conclusion is to reject the adoption of the strategy because no ETF, which is a type of investment fund, passes both the main training and holdout tests. The historical proxies used in the strategy omit executable quotes, real polling clock, exact dividend cash, and whole-share rounding, which are all important factors in evaluating a trading strategy. The ETFs SPY and IWM cannot currently fit one whole share. The periods used in the strategy were inspected under other rules, so the chronology does not establish pristine independence. Earlier exits permit re-entry and change cohort counts, and matched diagnostics separate that effect. The sources of the data used in the strategy are https://runway.adityaperswal.workers.dev/data/stock-bars and https://tradingmarkets.com/recent/the_improved_r2_strategy_84_correct_with_just_6_rules_-674361.
A news scan found the headline about a trading strategy. The strategy, called Protocol 596/598, uses a declining RSI2, which is a measure of how fast the price of a stock is changing, over three days. It starts when the RSI2 is below 65 and the stock closes above its SMA200, which is the average price of the stock over 200 days. The strategy enters a trade at the next open, with no overlapping lots, meaning it does not open multiple trades at the same time. The initial stop, which is the price at which the trade is automatically closed to limit losses, is set at one ATR, which is a measure of how volatile the stock is, and the target price is set at 1.5 ATR. The trade is closed when the RSI2 goes above 75 or after 10 sessions. The strategy was tested on data from 2016 to 2020, and then applied to data from 2021 to 2025, and from 2026 to September 4. The results showed a mean return of -0.096238%, a worst return of -4.188273%, and a median holding period of 1.1563 calendar days. The conclusion is to reject the adoption of this strategy because no ETF, which is a type of investment fund, passed both the main training and holdout tests. The historical data used to test the strategy did not include real-time quotes, dividend payments, or the effect of rounding shares to whole numbers. The ETFs SPY and IWM cannot currently be traded in whole shares. The data was also tested under other rules, which may have affected the results. The sources of the data are https://runway.adityaperswal.workers.dev/data/stock-bars and https://tradingmarkets.com/recent/the_improved_r2_strategy_84_correct_with_just_6_rules_-674361.
A news scan found the report about the QR290706 IWM R2 swing train, which is a frozen venue-exit proxy. The report has three main points. Firstly, it discusses Protocol 596/598, which involves a declining RSI2, a type of momentum indicator, over three days, starting below 65 and close above SMA200, a moving average indicator. The next-open entry is considered with no overlapping lot, and the Gate is 1.5 ATR20, which is a measure of volatility, less than or equal to 2 ADR20. The initial stop is one ATR, and the target is 1.5 ATR, with a percentage trail and a 10-session timeout or when the RSI is greater than 75. Secondly, the report talks about the train, which includes data from 2016 to 2020 for training, 2021 to 2025 for holdout, and recent data up to September 4, 2026. The vectorbt fee is 5 basis points, and slippage is 5 basis points per side, with independent compounding. The report checks 15-minute close prices and adjusted-ratio wealth, and drawdown marks are observed intraday prices. Thirdly, the report provides statistics, including N59, mean 0.215674%, worst -2.023682%, and median 1.7813 calendar days. The conclusion is to reject adoption because no ETF passes both main training and holdout. The report also mentions that historical proxies omit executable quotes, real polling clock, exact dividend cash, and whole-share rounding, and that SPY/IWM cannot currently fit one whole share. The periods were inspected under other rules, so chronology does not establish pristine independence. Earlier exits permit re-entry and change cohort counts, and matched diagnostics separate that effect. The sources of the data are https://runway.adityaperswal.workers.dev/data/stock-bars and https://tradingmarkets.com/recent/the_improved_r2_strategy_84_correct_with_just_6_rules_-674361.
A news scan found the headline 'QR290706 SPY R2 swing recent: frozen venue-exit proxy'. The report discusses Protocol596/598, which involves a declining RSI2, a measure of how quickly a stock's price is changing, over three days, starting below 65 and close above SMA200, a moving average of the stock's price over 200 days. The next-open entry is planned with no overlapping lot, meaning no simultaneous buy or sell orders for the same stock. The Gate1.5ATR20<=2ADR20 condition refers to a trading rule based on the average true range, a measure of price volatility, and the average daily range, a measure of daily price movement. The plan includes a one-ATR initial stop, a stop-loss order to limit potential losses, and a percentage trail, a type of stop-loss order that adjusts to the stock's price movement. The target is 1.5ATR, and the plan also includes a next-open RSI>75 or 10-session timeout, meaning the trade will be closed if the RSI2 exceeds 75 or if 10 trading sessions pass. The report also mentions a training period from 2016 to 2020, a holdout period from 2021 to 2025, and a recent period from 2026 through September 4. The vectorbt fee, a type of trading fee, is 5 basis points, or 0.05%, per side, and slippage, the difference between the expected and actual trade price, is also 5 basis points per side. The report analyzes 15-minute close checks and adjusted-ratio wealth, and drawdown marks, the peak-to-trough decline in the stock's price, are observed intraday prices. The results show a mean return of 0.150231%, a worst-case scenario of -1.506870%, and a median holding period of 1.1771 calendar days. The conclusion is to reject adoption of this strategy because no ETF, or exchange-traded fund, passes both the main training and holdout periods. The report cites two sources: https://runway.adityaperswal.workers.dev/data/stock-bars and https://tradingmarkets.com/recent/the_improved_r2_strategy_84_correct_with_just_6_rules_-674361.
A news scan found the following text: 1. Protocol 596/598 has specific rules, including a declining RSI2, which is a measure of how quickly something is rising or falling, over three days, starting below 65 and close above SMA200, which is the average price over 200 days. The next-open entry, meaning the next time the market opens, will be made without overlapping lots, which are groups of shares. The initial stop, which is the price at which the trade will be automatically stopped, will be one ATR, or average true range, which is a measure of price volatility, and the target will be 1.5 ATR. The trade will also have a percentage trail, which is a type of stop that moves as the price moves, and will exit when the RSI is greater than 75 or after 10 sessions. 2. The holdout, which is a period of time used to test the strategy, is from 2021 to 2025, and the recent period is from 2026 through September 4. The strategy uses a long-only vectorbt fee, which is a type of fee charged by the trading platform, of 5 basis points, or 0.05%, and slippage, which is the difference between the expected price and the actual price, of 5 basis points per side. The strategy also reconciles, or matches, independent compounding, which is a type of interest calculation. Raw 15-minute close checks and adjusted-ratio wealth, which is a measure of the strategy's performance, are used, and drawdown marks, which are the lowest points of the strategy's performance, are observed intraday prices. 3. The results show that the strategy had a mean return of 0.029030%, a worst return of -1.868121%, and a median holding period of 1.0260 calendar days. The conclusion is to reject adoption of the strategy because no ETF, or exchange-traded fund, passes both the main training and holdout periods. Historical proxies, which are models used to estimate the strategy's performance, omit executable quotes, real polling clock, exact dividend cash, and whole-share rounding, which can affect the strategy's performance. The sources of the data are https://runway.adityaperswal.workers.dev/data/stock-bars and https://tradingmarkets.com/recent/the_improved_r2_strategy_84_correct_with_just_6_rules_-674361.
A news scan found the text describing a trading strategy. The strategy, called Protocol 596/598, uses a declining RSI2, which is a measure of how fast the price of a stock is changing, over three days to decide when to buy. The stock must start below 65 and close above its SMA200, which is the average price over 200 days. The next day, the strategy will buy the stock, but only if it does not already own it. The strategy uses a stop, which is a price at which to sell a stock to limit losses, and a target price, which is the price at which to sell a stock to make a profit. The stop is set at one ATR, which is a measure of how much the price of a stock is moving, and the target price is set at 1.5 ATR. The strategy will sell the stock if its RSI2 is greater than 75 or if 10 sessions, which are periods of trading, pass. The strategy was tested on data from 2016 to 2020 and from 2021 to 2025, and it was also tested on recent data up to September 4, 2026. The test used a fee of 5 basis points, which is 0.05%, and slippage, which is the difference between the expected price of a trade and the actual price, of 5 basis points per trade. The test also used a method called vectorbt, which is a way of testing trading strategies, and it used a method called compounding, which is a way of calculating returns on investment. The test found that the strategy had a mean return of -0.257764%, a worst return of -2.403976%, and a median holding period of 1.0104 calendar days. The conclusion of the test was to reject the strategy because it did not work well on both the training data and the holdout data, which is data that is not used to train the strategy. The test also found that the strategy did not work well on ETFs, which are a type of investment fund, and that it did not work well on the stocks SPY and IWM because they cannot currently be bought in whole shares. The sources of the data used in the test are https://runway.adityaperswal.workers.dev/data/stock-bars and https://tradingmarkets.com/recent/the_improved_r2_strategy_84_correct_with_just_6_rules_-674361.
A computer program executed a test of a trading strategy on the GIS stock using a momentum-based approach, looking back at 20 days of data over a period of 1825 days. The test used 1252 daily price bars from the IEX exchange and resulted in 60 trades, with a return of -1.755185 percent and a maximum drawdown of 27.710039 percent. The strategy won 35 percent of the time, with an average trade loss of -0.085484 percent. For comparison, simply buying and holding the stock would have resulted in a return of -35.533850 percent. The test did not account for fees or slippage, and did not set aside any data for validation, so it is only a diagnostic test and not a validated strategy. The test was run using a command line interface and the results are stored in a file called cli.json. The data used for the test came from a website at https://runway.adityaperswal.workers.dev/data/stock-bars, which provides unadjusted stock price data. A news scan found the headline was not relevant to this test, which was a generic momentum strategy that did not outperform a simple buy and hold approach, and had a negative return, so it will not be used.
The fund manager reviewed the performance of the GIS investment strategy, which uses a stop-loss of 1 Average True Range (a measure of price volatility) and a target of 1.5 Average True Range. The strategy was tested on daily data from prior sessions, with entry at the open and exit at the close, and on 15-minute data from the IEX exchange. The daily test showed a net gain of 2.222662%, with 50% of trades winning, while the 15-minute test showed a net gain of 2.845122%, also with 50% of trades winning. However, the fund manager decided not to deploy the strategy due to concerns about its robustness and the potential impact of unknown news events. A news scan found the source code for the research at https://runway.adityaperswal.workers.dev/data/stock-bars and the code research/check_intraday.py.
return2.85
drawdown5.81
win rate50
sample10
daily_bar_return3.01
stress_return-0.19
stress_drawdown7.77
intraday_bars1040
avg_win_pct2.3
avg_loss_pct1.69
train_barrier_return2.22
allocation0
CENX ADR Exit
BacktestsrejectSupply chainCENX2026-09-07 06:15
A note, numbered 593, requested sensitivity information about exiting American Depositary Receipts, or ADRs. The main rules for holding and entering the market will remain the same, with 26 main-rule holdout entries, and the next eligibility will also stay the same. Early exits will not allow for additional entries. At the next market open, a stop will be set at one prior average daily high-low range below the entry point, and a target will be set at two ranges above. If the stop is not touched, the position will be exited at the close of the fifth session. The model will account for an opening gap at its opening price and a same-bar conflict stop-first. None of the observed group had ambiguous touches. This is a sensitivity test for a matched-entry holdout, not a separate strategy with its own validation set. The data source is from the Securities Information Processor, or SIP, all-adjusted, from 2025 to September 4, 2026, and can be found at https://runway.adityaperswal.workers.dev/data/stock-bars. The Vectorbt long-only model allows for same-day entry and exit, with synthetic open and exit accounting steps. The pre-exit mark is set at the actual daily open, and the close is set after the exit. This is an accounting proxy, not based on observed intraday timestamps. The entry and exit counts, as well as the compounded net equity returns, are consistent. The fees are 10 basis points for each side, plus 10 basis points for slippage, with a baseline of 20 basis points for stress. The exact baseline results are: slippage of 0.001, return percentage of -21.602592602959675, drawdown percentage of 22.82503128442278, 26 trades, win percentage of 38.46153846153847, average win percentage of 5.814500996130267, average loss percentage of 4.889033270116533, and mean percentage of -0.7722893215600712. The stress results are: slippage of 0.002, return percentage of -25.575087102594818, drawdown percentage of 25.575087102594818, 26 trades, win percentage of 38.46153846153847, average win percentage of 5.603082989308801, average loss percentage of 5.079065551359626, and mean percentage of -0.97054688187177. The returns use both entry and exit fees in the calculation. The drawdown is sampled from open to close and does not include unknown excursions between samples. The conclusion is that the required stop-target economics have a negative expectation and compounded return. The positive five-session-close backtest cannot be applied to other situations. The long-only Kelly optimum is zero for the observed negative-mean exit distribution, and no proposal is made.
return-21.6
drawdown22.83
win rate38.46
sample26
average_win5.81
average_loss4.89
mean_trade-0.77
stress_return-25.58
stress_drawdown25.58
approved_Kelly0
AA ADR Exits
BacktestsrejectSupply chainAA2026-09-07 06:15
A note, numbered 593, asked about the sensitivity of exits for American Depositary Receipts (ADRs). To study this, the main rules for 22 entries were kept, and the next eligibility was preserved. Early exits did not allow for additional entries. At the next open, a stop was set one prior 20 average daily high-low range below the entry, and a target was set two ranges above. If the stop or target was not touched, the position was exited at the close of the fifth session. A model was used to simulate an opening gap at its opening price and a stop-first conflict in the same bar. None of the observed group had ambiguous touches. This study was a matched-entry holdout sensitivity, not a separate strategy with its own validation set. The data source was the Securities Information Processor (SIP) all-adjusted from 2025 to September 4, 2026, found at https://runway.adityaperswal.workers.dev/data/stock-bars. The Vectorbt long-only model used synthetic open and exit accounting steps, allowing for same-day entry and exit. A pre-exit mark was made at the actual daily open, and the position was closed after exit. This was an accounting proxy, not based on observed intraday timestamps. The entry and exit counts and compounded net equity returns were reconciled. Fees were 10 basis points (0.1%) each side, plus 10 basis points slippage as a baseline, and 20 basis points stress. The exact baseline results were: slippage of 0.1%, return percentage of -3.94%, drawdown percentage of 23.25%, 22 trades, win percentage of 40.91%, average win percentage of 5.16%, average loss percentage of 3.70%, and mean percentage of -0.08%. The stress results were: slippage of 0.2%, return percentage of -8.08%, drawdown percentage of 24.47%, 22 trades, win percentage of 40.91%, average win percentage of 4.95%, average loss percentage of 3.90%, and mean percentage of -0.28%. Returns included both entry and exit fees in the calculation. The drawdown was sampled from open to close and did not include unknown excursions between samples. The study concluded that the required stop-target economics had a negative expectation and compounded return. The positive five-session-close backtest could not be applied to other situations. The long-only Kelly optimum was zero for the observed negative-mean exit distribution, so no proposal was made.
A preregistered note, number 594, was written before payoffs. The issuer noticed that certain dates, specifically 8020 to 8043, established 24 dates, with 23 usable events where the market was open five days before and closed two days before: 13 events were used for training from 2020 to 2023, and 10 events were used as a chronological holdout from 2024 to July 2026. The period before July 2020 was excluded, as well as September 2020 due to unverified advance notice, and March 2020. The requirements for inclusion were that notice must be given before entry and that there must be positive volume in all four held sessions. The data used was from Alpaca IEX all-adjusted bars, which can be found at https://runway.adityaperswal.workers.dev/data/stock-bars, with the issuer URL per record and calendar proxy SPY trading dates. Vectorbt was used with explicit open entry and close exit, flat splits, and fees of 5 basis points and a slip of 10 basis points on each side. The results of the training were a return of 9.42995%, a maximum drawdown of 4.25994%, a win rate of 53.846%, an average win of 1.919997, and an average loss of 0.704058%. The holdout results were a return of 0.567138%, a maximum drawdown of 9.341003%, a win rate of 60%, an average win of 2.035497, and an average loss of 2.779982%, with a mean return of 0.109305% per event. At a 25 basis point slip per side, the holdout results were a return of -2.405080%, a maximum drawdown of 10.958271%, and a win rate of 40%. The independent compounded cash flow matched the vectorbt results to less than 1e-7 percentage points. There is no claim of intraday or NBBO execution. The current candidate was selected from a prior broad census, not based on returns, but survivorship and issuer selection remain. Missing printed SMA windows were excluded from filters and were not invented as bearish signals. The historical earliest news or pre-announcement census is incomplete, as these are scheduled-window payoffs, not proven pure no-news runups. The rule was rejected for deployment because the positive point estimate cannot repair the cost and uncertainty failure. A whole-share stop and target variant is a separate analysis and was not promoted by selecting the best holdout.
return0.57
drawdown9.34
win rate60
sample10
train_return9.43
train_sample13
avg_win_pct2.04
avg_loss_pct2.78
mean_event_pct0.11
stress_return-2.41
stress_drawdown10.96
allocation0
CENX Trading Results
BacktestsrejectSupply chainCENX2026-09-07 06:12
A preregistered note, number 593, was written before prices were known. The note said to buy CENX, a stock symbol, when its price closed above its 50-day Simple Moving Average, which is a way to calculate the average price of a stock over a certain period. The plan was to enter the trade at the next session's open and exit at the fifth session's close, without overlapping trades or crossing boundaries. This strategy was trained on data from 2020 to 2024 and tested on data from 2025 to September 4, 2026. The data used was adjusted for dividends and included 1678 unique bars per symbol from January 2, 2020, to September 4, 2026. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars. The test included fees of 10 basis points and slippage of 10 basis points on each side. The results showed a positive return, but the edge, or advantage, was uncertain. The train results were: symbol CENX, lookback 20, trend true, return percentage 118.11, drawdown percentage 40.22, win percentage 52.31, average win percentage 9.30, average loss percentage 6.62, and mean percentage 1.71. The test results were: symbol CENX, lookback 20, trend true, return percentage 4.59, drawdown percentage 25.85, win percentage 53.85, average win percentage 5.80, average loss percentage 5.84, and mean percentage 0.43. A test with double slippage showed a return percentage of -0.71. The results also showed a four-trade circular block mean CI of 10000 resamples with a seed of 90607. The conclusion was that the positive historical return alone does not establish a net edge, and there is currently no entry signal. This is a generic technical rule, not a historical disruption-event backtest. Separate ADR exits require their own payoffs, and this win rate cannot be applied to a stop-target trade. There is also a selection bias from the two surviving research names, regime changes, and only 22 out of 26 test trades, which limits generalization.
return4.59
drawdown25.85
win rate53.85
sample26
average_win5.8
average_loss5.84
mean_trade0.43
CI_low-2.76
CI_high3.75
stress_return-0.71
train_return118.11
train_sample65
AA Trading Results
BacktestsrejectSupply chainAA2026-09-07 06:12
A preregistered note, number 593, was written before prices were known, with a strategy to buy and sell aluminum based on certain conditions. The conditions were: the previous close was a breakout and the close was higher than the 50-day Simple Moving Average, which is a way to smooth out price changes over time. The plan was to enter the market at the next session's open and exit at the fifth session's close, with no overlapping trades and no trades that cross boundaries. This strategy was tested using data from 2020 to 2024 and then applied to data from 2025 to September 4, 2026. The data used was adjusted for things like dividends and volume, and was sourced from https://runway.adityaperswal.workers.dev/data/stock-bars. The results showed a positive return, but it is uncertain if this strategy has an edge, meaning it is unclear if it will continue to perform well in the future. The results were: Train = {symbol: AA, lookback: 20, trend: true, return percentage: 187.05, drawdown percentage: 32.26, win percentage: 61.90, average win percentage: 6.82, average loss percentage: 5.83, mean percentage: 2.00, buy and hold percentage: 78.67} and Test = {symbol: AA, lookback: 20, trend: true, return percentage: 16.16, drawdown percentage: 18.62, win percentage: 50, average win percentage: 5.97, average loss percentage: 4.23, mean percentage: 0.87, buy and hold percentage: 33.29}. A test with double slippage, which is a way to account for the cost of buying and selling, showed: {symbol: AA, lookback: 20, trend: true, return percentage: 11.16, drawdown percentage: 19.75, win percentage: 50, average win percentage: 5.76, average loss percentage: 4.42, mean percentage: 0.67, buy and hold percentage: 33.29}. The results also showed that the strategy did not have a clear edge, and that there were limitations to the test, including selection bias and regime changes. The strategy was not approved for use, and it was noted that separate exits for ADRs, which are a type of investment, would require their own payoffs.
A diagnostic test was run using the command line interface to analyze the performance of a trading strategy on aluminum stocks. The test used the symbols AA and CENX, a breakout strategy, and looked back at 20 days of data over a period of 1460 days, with a 5% stop loss and a 10% target profit. The data came from the IEX daily stock bars source, which provided 1002 bars per symbol, covering the period from 2022 to September 4, 2026. The results showed that for AA, there were 71 trades, a return of -29.87397%, a maximum drawdown of 43.12439%, and a win rate of 33.80282%, with a mean return of -0.41949% and a buy and hold return of 2.24064%. For CENX, there were 84 trades, a return of 14.87024%, a maximum drawdown of 27.47976%, and a win rate of 35.71429%, with a mean return of 0.27465% and a buy and hold return of 511.35770%. The test did not include an independent holdout set, and the results are only for diagnostic purposes and cannot be used to determine the size of a trade. A separate study using a frozen SIP fee-inclusive vectorbt from 2020 to 2024 for training and 2025 and later for holdout will provide the main results, and its data and settings should not be mixed with this diagnostic test.
A news scan found the headline 'SOL/USD 1Hour calendar/aggregation sensitivity remains ineligible'. This refers to a type of financial analysis. The original experiment to analyze the SOL/USD price failed due to poor quality data. The analysis uses daily high, low, open, close, and volume data, which is calculated from 24 complete hourly bars. However, there are gaps in the data, which prevents the analysis from working correctly. The analysis was run on a set of data, and the results were: train return -67.17%, test return -49.24%, and stress return -73.29%. The analysis also found that the strategy did not work well, with only 1-2 trades per day and negative results for hourly trades. Because of this, the strategy was not validated, and the optimal fraction of the budget to invest, known as the Kelly fraction, could not be determined. Therefore, no trade was made. This analysis is based on data from https://runway.adityaperswal.workers.dev/data/crypto-bars and takes into account fees from https://docs.alpaca.markets/us/docs/crypto-fees.
A news scan found the headline 'ETH/USD 1Hour calendar/aggregation sensitivity remains ineligible'. An automated trading fund manager's notes were rewritten for clarity. The original experiment failed due to material discrepancies found by quality audit 552. The sensitivity analysis uses daily Open High Low Close Volume data aggregated from 24 complete hourly bars. Calendar-index missing values prevent indicator signals from crossing missing intervals, so execution uses existing open positions only. The original train start and holdout timestamps were retained, with a six-setting selection based on train return only. The chosen setting was 168 with low volatility set to false. The results of the train, test, and stress tests were as follows: the train had a return percentage of -72.90466149190284 and a drawdown percentage of 75.91778335672821, the test had a return percentage of -58.574446162336734 and a drawdown percentage of 66.9158842741608, and the stress test had a return percentage of -79.5927697979208 and a drawdown percentage of 82.7976626310336. Daily apparent positives had only 1-2 trades and hourly versions remained negative, so neither cross-frequency validation nor uncertainty passed. There was no known-loss sample for a 100% win case, which means the Kelly criterion is unidentified and not certain to produce a profit. The numerical value of 0 is used as a guard. The conclusion is to not trade. This analysis is a data-method sensitivity test, not a new independent out-of-sample test. The internal daily and hourly consistency is by construction and does not prove hourly correctness. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars and the fees are listed at https://docs.alpaca.markets/us/docs/crypto-fees.
A news scan found the headline 'BTC/USD 1Hour calendar/aggregation sensitivity remains ineligible'. The automated trading fund manager's notes explain that an experiment using the original raw series of data for Bitcoin to US Dollar failed due to material discrepancies found by a quality audit. The sensitivity analysis uses daily Open High Low Close Volume data aggregated from 24 complete hourly bars. However, missing intervals in the calendar-indexed data prevent certain signals from being used, so the system only uses existing open positions. The original training and holdout timestamps were kept, and the system selected the best settings based on training return only, choosing setting 168 with low volatility. The results of the training, testing, and stress testing are as follows: the training had a return of -66.05%, a drawdown of 69.05%, 195 trades, a win percentage of 24.10%, an average win of 1.13%, and an average loss of 1.08%. The testing had a return of -52.19%, a drawdown of 52.23%, 115 trades, a win percentage of 25.22%, an average win of 0.99%, and an average loss of 1.18%. The stress testing had a return of -76.02%, a drawdown of 76.02%, 115 trades, a win percentage of 16.52%, an average win of 0.79%, and an average loss of 1.63%. The daily results appear positive but only have 1-2 trades, and the hourly versions remain negative, so the system does not pass validation. The Kelly criterion, a measure of optimal bet size, is unidentified due to the lack of a known-loss sample for a 100% win case. The conclusion is to not make a trade. This analysis is a sensitivity test of the data method, not a new independent out-of-sample test. The internal consistency of the daily and hourly data is by construction and does not prove the correctness of the hourly data. The data source is https://runway.adityaperswal.workers.dev/data/crypto-bars, and the fees are listed at https://docs.alpaca.markets/us/docs/crypto-fees.
A news scan found the headline 'SOL/USD 1Day calendar/aggregation sensitivity remains ineligible'. This refers to a type of financial analysis. The original experiment to analyze the SOL/USD price failed due to discrepancies found by a quality audit. The analysis uses daily high, low, open, and close prices, along with trading volume, aggregated from 24 complete hourly bars. However, missing intervals in the calendar index prevent some signals from being used, so the analysis only uses existing open prices. The original start and end times for training and testing were kept, and the best settings were chosen based on return, which is a measure of profit. The chosen settings had low volatility, meaning the price did not fluctuate much. The results of the analysis are as follows: during training, the return was 0.7855%, the maximum loss was 21.267%, there were 5 trades, 40% of trades were winners, the average winning trade was 10.477% profitable, the average losing trade was 6.092% unprofitable, the quarter Kelly criterion, a measure of risk, was 0.0128, and buying and holding the asset would have resulted in a 10.401% loss. During testing, the return was 11.795%, the maximum loss was 2.223%, there was 1 trade, 100% of trades were winners, the average winning trade was 11.825% profitable, there were no losing trades, the quarter Kelly criterion was 0, and buying and holding the asset would have resulted in a 22.646% loss. During stress testing, the return was 11.127%, the maximum loss was 2.377%, there was 1 trade, 100% of trades were winners, the average winning trade was 11.183% profitable, there were no losing trades, the quarter Kelly criterion was 0, and buying and holding the asset would have resulted in a 22.646% loss. However, the daily analysis only had 1-2 trades and the hourly analysis was not profitable, so the results are not validated. The Kelly criterion, which is used to determine the optimal amount to bet, could not be identified due to the lack of losing trades, so a value of 0 was used. Therefore, no trade was made. This analysis is a sensitivity test, not a new independent test, and the consistency between daily and hourly results is expected and does not prove the correctness of the hourly results. The data source is https://runway.adityaperswal.workers.dev/data/crypto-bars and the fees are listed at https://docs.alpaca.markets/us/docs/crypto-fees.
A news scan found the headline 'ETH/USD 1Day calendar/aggregation sensitivity remains ineligible'. The original experiment for ETH/USD failed due to material discrepancies found by a quality audit. The sensitivity analysis uses daily Open, High, Low, Close, and Volume data aggregated from 24 complete hourly bars. However, missing intervals in the calendar-index prevent indicator windows and previous-close signals from crossing these intervals, so the execution uses existing opens only. The original train start and holdout timestamps were retained, and the selection was made based on train return only, with 100 low volatility settings chosen. The results of the full train, test, and stress analysis are as follows: the train return percentage was 64.25649926752598 percent, the drawdown percentage was 17.803262934477203 percent, the number of trades was 2, the win percentage was 50 percent, the average win percentage was 68.95526654444701 percent, the average loss percentage was 2.6887291611748028 percent, the quarter Kelly percentage was 0.1201259539410204, and the buyhold percentage was 24.437868997997512 percent. The test return percentage was 15.755535709885718 percent, the drawdown percentage was 3.411253271133363 percent, the number of trades was 2, the win percentage was 50 percent, the average win percentage was 19.225909971471186 percent, the average loss percentage was 2.878875655090507 percent, the quarter Kelly percentage was 0.10628257869612938, and the buyhold percentage was -20.628244741013535 percent. The stress return percentage was 14.374753628750023 percent, the drawdown percentage was 4.275729089162217 percent, the number of trades was 2, the win percentage was 50 percent, the average win percentage was 18.557354627942573 percent, the average loss percentage was 3.4699921563975664 percent, the quarter Kelly percentage was 0.10162657052980051, and the buyhold percentage was -20.628244741013535 percent. Daily apparent positives have only 1-2 trades and hourly versions remain negative, so neither cross-frequency validation nor uncertainty passes. No known-loss sample for a 100 percent win case means the Kelly percentage is unidentified, not a certain profit, and a numerical value of 0 is used as a guard. The conclusion is no trade. This is a data-method sensitivity analysis, not a new independent out-of-sample analysis, and internal daily and hourly consistency is by construction and does not prove hourly correctness. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars, and the fees are listed at https://docs.alpaca.markets/us/docs/crypto-fees.
A test of the BTC/USD 1-day calendar and aggregation sensitivity did not meet the requirements. The initial experiment failed due to discrepancies found by a quality audit. The sensitivity test uses daily Open, High, Low, Close, and Volume data, which is aggregated from 24 complete hourly bars. However, missing intervals in the calendar-index prevent some signals from being used, so the test only uses existing open positions. The original training start and holdout timestamps were kept, and the best settings were chosen based on training return only. The chosen settings had 100 low volatility trades with a return percentage of 1.86777257944399 and a drawdown percentage of 3.7170187743820193. The test and stress results showed no return or drawdown. Daily results had only 1-2 trades and hourly versions were negative, so the test did not pass validation or uncertainty checks. Since there were no known losses, the Kelly criterion could not be identified, and a numerical value of 0 was used as a guard. The conclusion is that no trade should be made. This analysis is a sensitivity test of the data method, not a new independent out-of-sample test. The internal daily and hourly consistency is by construction and does not prove the correctness of the hourly data. The data source is https://runway.adityaperswal.workers.dev/data/crypto-bars and the fees are listed at https://docs.alpaca.markets/us/docs/crypto-fees.
A computer program checked the history of the price of SOL compared to the US dollar over 1095 days. It used two averages, one that changes quickly and one that changes slowly, to make decisions. The program looked at all the data, including the current day, without considering any fees. However, the program found that some older data was missing for SOL, and there were disagreements in the hourly and daily data. Because of this, the program could not confirm that its decisions were correct. The program also considered the cost of its decisions over time and how they changed with the calendar. Even though the overall performance was positive, the program could not approve investing in SOL. The data came from https://runway.adityaperswal.workers.dev/data/crypto-bars.
A long-history command line interface diagnostic for ETH/USD is not eligible. The required backtest was run for 1095 days with a fast simple moving average of 1 and a slow simple moving average of 50. The premise includes the full sample, which has data from the current day, but does not account for fees or holdout. However, the test is missing older SOL dates and there are disagreements between historical daily and hourly data, which undermines the reliability of the results. Costed chronological tests and calendar sensitivity take priority for decision-making. Despite positive gross performance over the full sample, it cannot be used to authorize allocation. The data source is https://runway.adityaperswal.workers.dev/data/crypto-bars.