Stats

This analysis looks at the underlying trend and volatility of the XLE stock, which is a collection of energy companies. It uses a system called vectorbt to look at the stock's closing prices from 2022 to 2026. The analysis uses six rules to decide when to buy or sell the stock, based on short and long-term trends. It also uses a filter to reduce false signals. The analysis found that the best rule was to buy when the short-term trend was above the long-term trend and the volatility was low. This rule was then tested on data from 2025 to 2026. The test used $10,000 in hypothetical money and accounted for fees and slippage. The results showed a return of -13.59% and a drawdown of -23.39%. For comparison, simply buying and holding the stock would have given a return of 54.91%. The analysis also looked at doubling the position size, which resulted in a return of -15.98% and a drawdown of -24.83%. The results are exploratory and do not establish the effectiveness of the strategy in different markets or over time. A news scan found the source of the data to be https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=XLE&timeframe=1Day&feed=iex&adjustment=all.

A news scan found the following text: Premises: 1) Underlying trend evidence may help understand options, but it cannot determine their expectancy. 2) Parameter choice must be made before testing. 3) Costs and signal delays must be considered. The vectorbt system was used to analyze IEX closing prices for TLT from 2022 to September 4, 2026. The system started using IBIT data on January 2024. Six rules were tested, including fast and slow trends, with and without a calm filter. A long condition was met when the closing price was higher than the fast and slow moving averages, and the calm filter was met when the mean of the 20-day r-squared was less than the mean of the 60-day r-squared. Signals were delayed by one session, and trades were entered and exited at the next closing price. The best-performing rule, 10/50 with a calm filter, was selected based on training data before January 1, 2025, and then frozen. The system was tested from January 2025 to September 4, 2026, starting with no balance. Historical indicators were warmed up using training data. The hypothetical account had $10,000, with no options, multiplier, or 15-minute exits. A base fee of 5 basis points plus slippage of 5 basis points was charged per trade, with an additional stress charge of 10 basis points plus 10 basis points per trade. Dividends were approximated using adjusted total return prices, but actual cash timing was omitted. The results were recorded in observations. The base strategy had a return of -5.9661%, a drawdown of -8.1855%, a win rate of 30.7692%, 13 closed trades, and an average win percentage of 1.1973%. The doubled strategy had a return of -8.3795%, a drawdown of -9.7331%, a win rate of 30.7692%, 13 closed trades, and an average win percentage of 0.9955%. The buy-and-hold strategy had a return of 1.1067%. The results include the terminal open mark where present, and the win rate and average wins and losses only use closed trades. The positive results are exploratory and do not establish trade dependence or post-selection confidence. These rules are distinct from the CLI's crossover baseline. The actual option payoffs and history are not available, and it is not possible to transfer stock payoffs to cheap-leg options or quarter Kelly. The data source is https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=TLT&timeframe=1Day&feed=iex&adjustment=all, and the research was conducted using research_vol/analyze.py and backtest.json.

A news scan found the following text: Premises: 1) Underlying trend evidence may help understand options but cannot determine their expected outcome. 2) Parameter choice must be made before testing. 3) Costs and delayed signals must be considered. The analysis used vectorbt and adjusted IEX closing prices from 2022 to 2026. It considered six rules with different settings, including a 'calm filter' which is a tool to reduce false signals. The rules were: long condition when the short-term trend is above the long-term trend and the calm filter condition is met, which means the average of the last 20 days is less volatile than the average of the last 60 days. The analysis shifted signals by one session and entered or exited at the next close. It selected the rule with the highest return before January 1, 2025, and then froze the '20/100/calm=False' rule. The test started on January 1, 2025, and ended on September 4, 2026. The hypothetical account had $10,000 and no options or multiplier. The base fee was 5 basis points plus slippage of 5 basis points per side, with an additional stress test of 10 plus 10 basis points per side. Dividends were approximated through adjusted total return prices. The results were recorded in observations. The base scenario had a return of 5.55%, a drawdown of -9.35%, a win rate of 41.67%, 24 closed trades, and 1 open trade. The doubled scenario had a return of 0.50%, a drawdown of -12.25%, a win rate of 41.67%, 24 closed trades, and 1 open trade. The buy and hold strategy had a return of 34.00%. The return includes the terminal open mark where present, and the win rate and average wins and losses use only closed trades. The source is https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=SPY&timeframe=1Day&feed=iex&adjustment=all and the research_vol/analyze.py and backtest.json files.

This analysis looks at the underlying trend of the IWM stock, which can help understand options, but does not predict their performance. The analysis uses a method called vectorbt, which is a way to backtest trading strategies, on adjusted closing prices from 2022 to 2026. It uses six rules, including fast and slow moving averages, to decide when to buy or sell. The rules are combined with a calm filter, which is a way to reduce false signals, and the results are shifted by one session to avoid looking at the same day's close. The best rule, 10/50 with a calm filter, was chosen based on the highest return before January 1, 2025, and then frozen for the holdout period from January 2025 to September 4, 2026. The analysis assumes a $10,000 account, with no options or multiplier, and a 1% size or 15-minute exit. It also includes base fees and slippage, which is the difference between the expected and actual price of a trade. The results show a return of -0.0141% for the base strategy, -3.9346% for the doubled strategy, and 35.8769% for a buy and hold strategy. The win rate, which is the percentage of successful trades, was 50% for the base strategy and 45% for the doubled strategy. The average win and loss percentages were also calculated. The analysis is exploratory and does not establish trade dependence or post-selection confidence, and no new fund or stock mandate is inferred. The results are based on data from https://runway.adityaperswal.workers.dev and were analyzed using research_vol/analyze.py and backtest.json.

A news scan found the report of an automated trading fund manager. The report starts with three premises: underlying trend evidence may help understand options but cannot predict their performance, parameter choice must be made before testing, and costs and signal delays must be considered. The manager used data from the IEX exchange, adjusted for dividends and splits, from September 4, 2022, to September 4, 2026, with the IBIT starting on January 2024. The manager tested six rules, combining fast and slow trend lines with a calm filter, which is a way to measure market volatility. The rules were: long condition when the closing price is above the fast trend line and the fast trend line is above the slow trend line, and the calm filter is true when the average volatility over 20 days is less than the average volatility over 60 days. The manager shifted the signals by one session and entered or exited positions at the next closing price. The manager selected the rule with the highest total return before January 1, 2025, which was the 20/100 rule with the calm filter set to false, and then froze it. The manager then tested this rule from January 1, 2025, to September 4, 2026, starting with a flat position and using historical indicators. The manager assumed a hypothetical account with $10,000 in the underlying asset, without options or leverage, and a position size of 1% or 15-minute exits. The manager also assumed a base fee of 5 basis points plus slippage of 5 basis points per side, and stress tests of 10 basis points plus 10 basis points per side. The results were: a return of -9.75% with a drawdown of -15.15%, a win rate of 35.71%, 14 closed trades, 1 open trade, an average win of 1.97%, and an average loss of -2.53% for the base scenario, and a return of -12.33% with a drawdown of -17.12%, a win rate of 28.57%, 14 closed trades, 1 open trade, an average win of 2.21%, and an average loss of -2.46% for the doubled scenario. The buy-and-hold return was -18.32%. The manager noted that the results are exploratory and do not establish trade dependence or post-selection confidence, and do not imply a new fund or stock mandate. The manager also noted that these rules are distinct from the crossover baseline and cannot be transferred to options or quarter Kelly strategies. The data source is https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=IBIT&timeframe=1Day&feed=iex&adjustment=all, and the research was done using the research_vol/analyze.py and backtest.json scripts.

A diagnostic test was run using a command line interface, or CLI, with the following parameters: node /app/agent/dist/backtest.js --symbols UAMY --strategy breakout --lookback 20 --days 1460 --stop-pct 5 --target-pct 10. The data came from https://runway.adityaperswal.workers.dev/data/stock-bars. The test resulted in 1002 bars, 68 trades, a return of 65.4721%, a maximum drawdown of 34.1623%, and a win rate of 39.7059%. The average return per trade was 0.9560%, and a buy-and-hold strategy would have returned 1185.9267%. This test is not the same as a previously registered strategy and does not have a separate model to validate its results. Another test, called vectorbt train/holdout test 543, was used to make a decision and the results of the CLI test were rejected. The good performance of the CLI test should not be confused with a proven advantage in the stock market.

A preregistered note, number 552, was written before a custom test. The stock UAMY was selected after today's filings, so the historical holdout is not pristine or prospective. The data used is from https://runway.adityaperswal.workers.dev/data/stock-bars, which includes SIP-adjusted daily data from January 3, 2022, to September 4, 2026, with 1173 bars and no pagination. The strategy used a fixed close price greater than the prior 20-day close high and a 50-day Simple Moving Average, or SMA, which is a calculation of the average price of a stock over a certain period of time. The next open price was also considered, as well as a frozen prior 20-day Average Daily Range, or ADR, with a stop loss at 1 ADR and a target at 2 ADR, with the fifth session close being the key. The strategy also included a stop loss on the first ambiguous daily bar and a stop gap with a worse open. Fees were 5 basis points, or bp, plus slippage of 10 bp on each side, with a stress test of 20 bp. The vectorbt long-only strategy with two marks per day was used, and all 35 holdout trades were long, with 35 exits, and the compounded entry-fee-inclusive Profit and Loss, or PNL, matched the portfolio with an accuracy of 1e-8. The daily Open, High, Low, and Close, or OHLC, data was used, but not 15-minute trigger or intraday liquidity validation. The Drawdown, or DD, was sampled twice a day and may understate intraday DD. The training period from 2022 to 2024 had 39 trades, with a return of -41.6594% and a DD of 57.9309%, with 30.7692% wins. The holdout period from 2025 to September 4, 2026, had 35 trades, with a return of -7.11606% and a DD of 58.2452%, with 40% wins, an average win of 12.96375%, and an average loss of 8.05117%. The mean return was 0.35480%, and the 95% mean interval was [-3.18592, 4.12585]%. The stress return was -13.39560%, with a DD of 58.8669% and 37.1429% wins, with an average win of 13.72262% and an average loss of 7.86336%. The primary net quarter Kelly point was 0.6842% of capital, and the stress was 0.2811%, but the uncertainty-adjusted approved value was 0 because the mean interval crossed zero, the training failed, and the drawdown was large and the current entry was false. Neighboring sessions were tested for robustness only, with train returns of -54.2387% and -38.8928%, and holdout returns of +17.8054% and +36.7124%, with 41 and 25 sessions, respectively, but the mean intervals crossed zero, so the best holdout was not switched to. This rejects the fixed stock rule, but does not prove negative fundamentals. The numeric edge cannot be imported from two operator-directed tanker wins. The internal default both-direction version was discarded, and the later raw trade return denominator was corrected before analysis. Earlier observation averages were superseded by corrected observations, but the portfolio outcomes were unchanged. There is no trade or paper profit claim.

A required backtest was run using the mean-reversion strategy with 20 and 2 parameters, over 365 days and 366 bars, including a partial live test from September 7, without fees, slippage, or holdout, and without using an exit flag. The results are gross diagnostic figures only and should not be trusted for a proposal. The test was verified using a fixed state implementation from a specific JavaScript file and included daily and hourly tests from 518 to 523, which supersede the decision to use the strategy. The test found that the historical performance of the SOL strategy was insufficient, and the net holdout failed, leading to the conclusion that the allocation should be 0, even when the gross SOL result is positive. The data source for this test was https://runway.adityaperswal.workers.dev/data/crypto-bars.

A required backtest, called CLI, was run on the ETH mean-reversion strategy, which is a way to predict price movements, with settings of z2 and 365 days, including 366 bars of data and a partial day on September 7, without considering fees, slippage, or holdout, and without using an exit flag, which is a signal to stop trading. The results are only for diagnostic purposes and should not be trusted for a proposal. The test was verified using a fixed state implementation from a specific code file, and daily and hourly tests were completed, with costs considered in chronological order, which led to decisions being made based on tests 518-523. The results showed that the historical performance was insufficient, the net holdout, which is a way to evaluate a model, failed, and even though the gross performance was positive for SOL, the conclusion was to allocate 0, meaning no investment. The data source for this test was https://runway.adityaperswal.workers.dev/data/crypto-bars.

A required backtest was run using the mean-reversion strategy with 20 days, 2 standard deviations, and 365 days of data, which includes a partial live test from September 7, and does not account for fees, slippage, or holdout. The results are gross diagnostic figures only and should not be trusted for a proposal. The test was verified using a fixed state implementation from the strategies.js file and completed 518 to 523 chronological tests, which supersede the decision to use. The premise of the test showed insufficient gross historical performance and failed to meet the net holdout requirements, leading to the conclusion that the allocation should be 0, even where the gross result is positive. The data source is https://runway.adityaperswal.workers.dev/data/crypto-bars.

A news scan found the headline 'SOL 1Hour oversold bounce fails chronological holdout after costs'. This refers to a trading strategy for the cryptocurrency SOL. The strategy is based on a premise that when the 20-bar close z-score is less than or equal to -2, it's a signal to buy, and the position is held until this condition is no longer true. The strategy was tested with a virtual capital of $10,000, including fees of 0.25% and a slip of 0.10% on each side, and stress tests of 50% and 20%. The training period was from 2025, and the holdout period was from January 1 to September 5, 2026. The results showed a return of -87.3132% during the training period and -75.1690% during the holdout period. A second premise tested different lookback periods and sensitivity levels, but no winning strategy was found. The data used was a full sample with some missing values, and the strategy was not designed to handle liquidation events. The historical holdout period had been previously used to test other rules, and the results were not promising. The strategy was rejected, and no allocation was recommended.

A news scan found the headline 'SOL 1Day oversold bounce fails chronological holdout after costs'. This refers to a trading strategy for the cryptocurrency SOL. The strategy is based on a premise that when the 20-bar close z-score is less than or equal to -2, it's a signal to buy, and the position is held until this condition is no longer true. The z-score is a measure of how many standard deviations an element is from the mean, and in this case, it's used to identify when the price is unusually low. The strategy was tested with a starting capital of $10,000, fees of 0.25%, and a slip of 0.10% on each side. The results from January 1 to September 5, 2026, showed a loss of 12.3485%. Another premise tested different lookback periods and sensitivities, but none of them resulted in a winning strategy. The data used for this test had 7,832 to 7,843 observations, and the strategy was not tailored to specific events, but rather a generic bounce strategy. The results suggest that this strategy does not work well and should not be allocated any funds.

A news scan found the headline 'ETH 1Hour oversold bounce fails chronological holdout after costs'. A trading strategy, called Premise1, was tested on Ethereum, which is a type of cryptocurrency. This strategy involved buying when the price was very low, and selling when it was no longer low. The strategy was tested using historical data from 2025, and the results were checked from January 1 to September 5, 2026. The strategy made 88.1564% return during training, but only 72.8574% return during the holdout period. The strategy also had a maximum loss of 72.9434%. Another strategy, called Premise2, was also tested with different parameters, but it did not perform well. The data used for testing was incomplete, with some missing values. The strategy was not designed to work with real-time trading, and it did not use stop-loss or target orders. The sources of the data used for testing are https://runway.adityaperswal.workers.dev/data/crypto-bars and https://docs.alpaca.markets/us/docs/crypto-fees.

A news scan found the headline 'ETH 1Day oversold bounce fails chronological holdout after costs'. This refers to a test of a trading strategy for Ethereum, a cryptocurrency. The strategy, called Premise1, is based on a statistical measure called the z-score, which is a way to measure how unusual a price move is. The strategy buys when the z-score is very low, which means the price is unusually low, and sells when the z-score returns to normal. The test used historical data from 2025 for training and from January 1 to September 5, 2026, for testing. The results were: a 3.2049% gain during training, a 7.5962% loss during testing, a maximum drawdown of 20.2821%, and a win rate of 40.000%. Another strategy, Premise2, was also tested with different parameters, but no winning combination was found. The data used for the test was complete, with no missing values, and the test was done without considering any specific events or conditions. The results suggest that this strategy does not work well and should not be used for actual trading. The data for the test came from https://runway.adityaperswal.workers.dev/data/crypto-bars and the trading fees were based on https://docs.alpaca.markets/us/docs/crypto-fees.

A news scan found the headline 'BTC 1Hour oversold bounce fails chronological holdout after costs'. This refers to a trading strategy for Bitcoin that uses a statistical measure called z-score, which is a way to measure how unusual a price move is. The strategy is based on the idea that when the z-score is very low, the price will bounce back. The strategy was tested using historical data from January 1 to September 5, 2026, and the results were: a gain of 85.9068% during the training period and a gain of -74.3840% during the holdout period, with a maximum drawdown of 74.4637%. The strategy also had a win rate of 0.4058% and a loss rate of 0.9172%, with 172 trades. When the costs of trading were taken into account, the gain was -92.3155%. Another test was done with different parameters, but no winning strategy was found. The data used for the test was missing some values, and the strategy was not designed to work with incomplete data. This test was not done to discover a new trading strategy, but rather to test an existing idea. The results suggest that this strategy is not profitable, and it was rejected. The sources of the data used for the test are https://runway.adityaperswal.workers.dev/data/crypto-bars and https://docs.alpaca.markets/us/docs/crypto-fees.

A news scan found the headline 'BTC 1Day oversold bounce fails chronological holdout after costs'. This refers to a trading strategy for Bitcoin that failed to perform well. The strategy was based on a premise that when the 20-bar close z-score is less than or equal to -2, it's a good time to buy, and it would sell when this condition is no longer true. The z-score is a measure of how many standard deviations an element is from the mean, and standard deviation is a measure of how spread out the numbers are. This strategy was tested with a simulated $10,000 investment, including fees of 0.25% and a slip of 0.10% on each side, and stress of 0.50% plus 0.20%. The results from January 1, 2025, to September 5, 2026, showed a 3.8751% return during the training period, but a -15.4031% return during the holdout period, with a maximum drawdown of 18.2248%. The strategy was also tested with different parameters, but none of them performed well. The data used for this test came from https://runway.adityaperswal.workers.dev/data/crypto-bars and the fees were based on https://docs.alpaca.markets/us/docs/crypto-fees.

A test of the QEX90704 system found that it did not work well. The test used frozen diagnostic variants and did not choose a winner. The main test had a return of -1.1431% at 10 basis points per side and -6.7137% at 20 basis points. When the next available bar-open fill was used after a signal, the return got worse to -5.6795% and -10.9945%. When only regular-hours-only closes were used, the return was -6.2437% and -11.5269%. A variant called Narrow0.75ATR allowed 42 later cohorts and had a return of +2.84395% at 10 basis points, but -5.44207% at 20 basis points, and -7.24753% and -11.24016% in training. This variant changed both admissions and exits, so it was not possible to isolate the effect of stop-distance. A variant called Wider1.25ATR did not allow any cohorts because the target exceeded 2 average daily ranges, so no return estimate was possible. All 28 rows of variant, split, and cost data had observations from 7787 to 7814, from the source https://runway.adityaperswal.workers.dev/data/stock-bars. The test was done with parameters that were set before the outcomes were known, but the underlying window and winner had already been looked at. The test was to see if the system was robust, not to select a new system. The conclusion is to reject using the swing system live. The strongest argument against this conclusion is the positive result from the narrow low-cost variant, but it was defeated by the sensitivity to cost, negative training, and uncertainty after selection. This test does not claim that all possible exits will fail. The main sample used same-day exits, which means that any future use must coordinate with the account-wide 3-in-5-day limit. There are no orders now.

A news scan found the notes of an automated trading fund manager. The manager tested a trading strategy on the DBC fund. The strategy had a 12-month adjusted return that was better than the BIL fund. The manager set a stop loss at 1 times the average true range and a target at 1.5 times the average true range. The manager also set a trailing distance of 1 times the average true range. The strategy was tested from 2017 to 2020 and then from 2021 to August 2026. The manager assumed a fee of 20 basis points per side, which is a measure of the cost of trading. The results showed a return of -6.71%, a drawdown of 12.66%, and a win rate of 37.93%. The manager concluded that this strategy does not work for trading because it does not have a positive edge, which is a measure of the strategy's ability to make money. The manager also noted that some months were profitable, but this does not mean the strategy is good. The manager used data from https://runway.adityaperswal.workers.dev/data/stock-bars and was based on the economic hypothesis found in https://pages.stern.nyu.edu/~lpederse/papers/TimeSeriesMomentum.pdf.

A news scan found the notes of an automated trading fund manager. The manager tested a trading strategy on the DBC stock. The strategy had a 12-month adjusted return that was better than the BIL stock. The manager started with a regular open entry and did not reenter the trade within the month. The manager used a 20-session simple Average True Range, or ATR, which is a measure of how much a stock's price moves, to set the initial stop loss at 1 ATR, the target at 1.5 ATR, and the trailing distance at 1 ATR. The manager sampled the raw 15-minute closes at the end of each interval and used the sampled highs to adjust the stop loss. The manager exited the trade at the observed price after costs. The manager skipped targets that were more than 2 ADR, or Average Daily Range. The manager trained the strategy from 2017 to 2020 and tested it from 2021 to August 2026. The manager split the fee and slippage, which is the difference between the expected and actual price of a trade, equally at 10 basis points, or 0.1%, per side. The results showed a return of -1.1431%, a drawdown of 10.0758%, and a win rate of 37.931%. The manager concluded that this strategy did not have a positive edge and should not be used for trading. The manager cited a study by NYU and a data source from Aditya Perswal.

A news scan found the notes of a trading fund manager. The manager tested a trading strategy called QEX90704 on a fund called DBC. DBC is a type of investment fund, and its previous returns were higher than another fund called BIL. The manager started testing the strategy at the beginning of the month, without trying to catch up on any missed trades. The manager used a simple method to set the initial stop loss, target, and trailing distance, which is a way to automatically adjust the stop loss as the price moves. The manager used a 15-minute interval to sample the closing prices and adjust the stop loss. The manager also used a method called ATR, which stands for Average True Range, a measure of price volatility. The manager set the initial stop loss at 1 ATR, the target at 1.5 ATR, and the trailing distance at 1 ATR. The manager also set a timeout, which means the trade would be closed at the next monthly opening. The manager trained the strategy using data from 2017 to 2020 and tested it using data from 2021 to August 2026. The manager assumed a fee of 10 basis points, which is a measure of the cost of trading, per side. The results showed a return of -1.8176%, a drawdown of 7.0494%, and a win rate of 45.4545%. The manager concluded that this strategy does not have a positive edge, which means it is not likely to produce consistent profits. The manager also found that the strategy had a maximum drawdown of 29.7078% and a total return of 104.6964% over the testing period. The manager referenced a study on time series momentum, which is a way to predict price movements based on past trends.

A news scan found the headline about QEX90704 monthly_matched results. The fund manager analyzed the performance of DBC, a ticker symbol, and compared it to BIL, another ticker symbol. DBC's adjusted return for the previous 12 months was higher than BIL's. The manager used a strategy called a swing target, which is a price level that a stock must reach before the manager will sell it, with a gate of 2ADR20, meaning the stock price must not move more than twice the average daily range in the last 20 days. The manager tested this strategy using data from 2017 to 2020 and then from 2021 to August 2026. The results showed a return of 37.36428440482524 percent, a drawdown of 26.470485269856113 percent, and a win rate of 51.724137931034484 percent. The manager also reported that the strategy had an average win of 4.907200275789534 percent and an average loss of 2.7464593861411704 percent. The manager used a tool called Vectorbt, which is a backtesting framework, to simulate the strategy and found that it matched the expected results. The manager concluded that the strategy should be retained for research purposes only, and that monthly profitability does not necessarily mean that the strategy is working. The manager also noted that some months were profitable, but this does not validate the overall strategy. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars and the economic hypothesis is based on a paper by NYU Stern.

A news scan found the notes from a trading fund manager. The fund's previous month return was higher than a benchmark called BIL. The fund started with a regular open entry and did not reenter during the month. The fund held its position for the full month without any restrictions. The fund's performance was tested using data from 2017 to 2020 and then from 2021 to August 2026. The test was done with a 10 basis point fee on each side, meaning 0.1% of the trade value was deducted as a fee. The results showed a return of 104.6964% and a maximum drawdown of 29.7078%. The fund was always long, meaning it only bought assets, and never sold short. The results were calculated using a method called vectorbt, which replays trades to test performance. The drawdown, or maximum loss, was calculated using 15-minute closing prices, including extended hours. The fund's performance was compared to a saved study, which showed a different return. The conclusion is to retain a research watch, meaning the fund's performance is being monitored but not used for actual trading decisions. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars and the economic hypothesis is based on a paper from NYU, https://pages.stern.nyu.edu/~lpederse/papers/TimeSeriesMomentum.pdf.

A news scan found the headline about a preregistered 20-session momentum transport. The report showed a 10/40 sensitivity without selecting a holdout winner. The Vectorbt, a backtesting framework, used a lagged close signal, next positive-volume open, and a persistent desired position, with a flat start for each split and a last open forced exit. The strategy included a 5 basis point fee and 10 basis points of slippage on each side, with a 25 basis point stress test. The training period from November 2024 to December 2025 resulted in a 20-day return of -32.6185%, a drawdown of 35.5869%, 10 trades, a win percentage of 20%, and a win-loss ratio of 6.3481% to 6.0441%. The raw quarter Kelly criterion was -0.140422, which was clipped to 0. The holdout period from January to September 2026 showed a return of +11.7464%, a drawdown of 22.4831%, 7 trades, a win percentage of 28.5714%, and a win-loss ratio of 22.1586% to 4.9969%, with a stress test result of +9.58875%. However, the holdout-only descriptive quarter Kelly criterion of 3.11595% was not adopted because the prior training failed and this strategy holds across events, unlike the Catalyst mandate. The 10/40 day train resulted in -52.0329% and -40.3158%, while the holdout resulted in +32.9979% and +27.4141%, with 10 and 4 trades respectively. All sweep values recorded observations, and unselected positive holdouts were acknowledged. There was no loss bound from daily exits, and IEX first prints were not auction or NBBO, with zero-volume prices remaining an indicator staleness risk. A separate event audit of 500 train and holdout showed no evidence of generic momentum transferring to a three-session pre-PDUFA window, with a reject risk size of 0.

The preregistered note 548 looked at returns before events, entering at the open of the fifth regular session strictly before the PDUFA, and exiting at the close of the second session before. This was done in chronological order, with the first two events used for training and the last one used for a holdout. The dates were found on the issuer's website at https://ir.telixpharma.com/static-files/9959d5f3-16cc-49fc-89eb-4a4af5b57049 and https://telixpharma.com/news-views/telix-reports-204m-revenue-up-63-yoy/. The analysis checked 452 bars on the IEX against SPY session dates, using vectorbt with 5 basis point fees and 10 basis points of slippage on each side, and a 25 basis point stress test. The Gozellix event from March 17 to 20 had an entry price of 17.455 and an exit price of 17.69, resulting in a paper return of 1.042736%, but the exit volume was 0, making it invalid. The Pixclara event from April 21 to 24 had an entry price of 16.75 and an exit price of 19.38, resulting in a paper return of 15.354908%, but the entry volume was 0, making it invalid. The Zircaix event from August 20 to 25 had an entry price of 10.96 and an exit price of 12.18, resulting in a valid positive return of 10.798492%, with a drawdown of 2.706111% and a win rate of 100%. The stress test resulted in a return of 10.466594%, with a drawdown of 2.852198%. This means that there were no eligible training events and one holdout event, with a return of 10.7985% and no observed loss. The calibrated p and average L/Kelly are undefined, with a size of 0. The analysis used the same issuer and non-independent products, with overlapping earnings and an incomplete first news archive, which prevents transport. The early approval of Gozellix also prevents the claim of a fixed deadline exit guaranteeing no news. The current T-5 event on September 3 has already elapsed, and the rule should not be shrunk to fit the September 8 event. The positive lone holdout is reported transparently, but it is not proof that no edge exists. The research and analysis can be found in the final_results.json file, which preserves the figures.

A diagnostic test was run on the TLX symbol using a momentum strategy, which is not a catalyst strategy. The test used a lookback period of 20 days and a total of 650 days. The results showed 444 bars, 18 trades, a return of -4.061632%, a drawdown of 46.722770%, and a win rate of 33.3333%. For comparison, a buy and hold approach would have resulted in a return of -22.753717%. The test was run using a command line interface, which lacks a filter for dated events and does not account for execution costs or holdout periods, making it unsuitable for allocation decisions. The results will trigger a separate audit using vectorbt, which will include costs and be recorded in chronological order. Note that the test used raw data from IEX, which included zero-volume bars that require special handling, and the first print is not equivalent to the opening auction. No trade is proposed based on this test.