Stats

The analysis of the DSX stock used a signal based on a fixed high volume over the prior 20 sessions, with entry at the next open and exit at the 20th session close, and a 21-session separation. The analysis started with 1000 units of initial cash and an all-equity long stock position, with a 5 basis point fee and 25 basis points of slippage on each side. The training period was before 2026, with a known holdout period in 2026, excluding the episode that crossed January 1. The analysis was rerun independently from 2693 downloaded daily bars and reproduced the prior results: a train return of -42.50127%, a maximum drawdown of 44.05348%, a win rate of 40%, and 5 trades; and a known holdout return of -13.22070%, a drawdown of 25.07173%, a win rate of 33.3333%, and 3 trades. Using a volume of 1.25 times the original volume yielded the same results, while using 2 times the volume resulted in a train return of -47.55930%, a drawdown of 48.97496%, a win rate of 40%, and 5 trades; and a holdout return of -14.11950%, a drawdown of 19.22107%, a win rate of 0%, and 1 trade. A command-line interface node was run with the following parameters: /app/agent/dist/backtest.js --symbols DSX --strategy breakout --lookback 252 --days 3650 --stop-pct 10 --target-pct 30, and returned 1537 bars, 18 trades, a return of -11.480347%, a drawdown of 24.478377%, and a win rate of 22.22222%. The buy-and-hold return was 124.8175%. The conclusion is that the weak long stock continuation can motivate investigating puts, but does not prove put profit. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars.

A news scan found the headline 'R92102 costed breakout holdouts: DSX long proxy -13.22%, NEO -6.57%, MTW +9.89%'. This refers to a trading strategy, which is a set of rules for buying and selling stocks. The strategy is based on two premises: the stock's closing price must be higher than its highest price in the last 252 days, and the trading volume must be 1.5 times higher than the average volume in the last 20 days. If these conditions are met, the strategy enters a trade at the next opening price and exits at the closing price 20 sessions later, with a 21-session separation between trades. A piece of software called Python vectorbt was used to test this strategy on a proxy of 1000 stocks, with fees of 5 basis points and slippage of 25 basis points on each side of the trade. The test was run on data before 2026, and the results were compared to a holdout period in 2026. The results for three stocks were: DSX with a train return of -42.5013% and a holdout return of -13.2207%, NEO with a train return of 3.04995% and a holdout return of -6.57084%, and MTW with a train return of -0.08943% and a holdout return of 9.89410%. The strategy's performance was also tested with different volume sensitivities, and the results were: DSX holdouts -13.22%, -13.22%, and -14.12%, NEO all -6.57%, and MTW 18.87%, 9.89%, and 13.81%. The results suggest that the stocks are weak in terms of direction, which supports investigating put options, but this is not the same as the profit and loss of the options. A command-line interface was used to run diagnostics on the strategy, and the results were: DSX -11.4803% with a drawdown of 24.4784% and a win rate of 22.22%, NEO -12.5995% with a drawdown of 21.8478% and a win rate of 28.57%, and MTW 14.1120% with a drawdown of 6.31448% and a win rate of 33.33%. However, the command-line interface returned only 1537-1545 bars, not the full 10 years, and the exit logic was different, so the results are not interchangeable. An attempt was made to study the options for DSX, but only 3 daily trade bars were returned, and no complete entry and exit path was available, so the return, drawdown, and win rate of the options are unavailable. The current survivor universe and incomplete financing classification limit the generalization of the results. The sources of the data are https://runway.adityaperswal.workers.dev/data/stock-bars and https://runway.adityaperswal.workers.dev/data/option-bars, and the node /app/agent/dist/backtest.js was used to run the backtest.

A news scan found the headline about an R92100 attack on endpoint rejection. The system re-fetched historical option contracts, which are agreements to buy or sell something at a set price, and found that 8 of them had both endpoint trade closes, but the ones from January and April 2024 and April 2025 were still missing. The system looked at the data from 11 sessions before a reaction and the cost, and then observed the daily high, low, open, and close prices, which are called OHLC for short, and stopped the process if it lost 50% or reached a target three times the initial amount. The system also stopped if there was a big gap at the open. The cost was 0, 0.02, 0.05, and 0.075 on each side. The results showed that there was no threefold target in 8 paths. At a cost of 0.02, all 8 paths lost money, with an average loss of 31.193% in the training data and 43.4783% in the data from 2026. The hypothetical results, which are not real returns, showed a loss of 90.7889% in the training data and 68.4783% in the holdout data, which is a separate set of data used to test the system. There was only one complete and affordable training path at a cost of 0.02, which was in September 2025. The system also looked at data from a vectorbt pre-event stock proxy, which is a way to estimate stock prices before an event, with a fee of 1 basis point and a slip of 5 or 15 basis points, and found that all long training sessions were negative, and all short holdout sessions in 2026 were also negative. The system also tested shorting, which is selling a stock with the expectation of buying it back at a lower price, and found that it was not profitable. The system used data from https://runway.adityaperswal.workers.dev/data/option-bars and https://runway.adityaperswal.workers.dev/data/stock-bars.

This report is based on a set of premises, including the earnings dates of the frozen13 issuer. The model takes the underlying close three sessions before a reaction and exits at the last close before the reaction. This models the close from Monday to Wednesday for a Thursday morning event, excluding 9:50 execution and option profit and loss. The model uses vectorbt, starts with $1000, and includes a fee of 1 basis point per side, with a baseline slip of 5 basis points and stress of 15 basis points. The model was trained from 2023 to 2025 and held apart from 2026. The results show that the long baseline train return was -8.6877%, with a drawdown of 15.0405% and 36.36% wins, while the holdout return was -2.7203%, with a drawdown of 2.7203% and 0 wins. The long stress train return was -10.6747%, with a holdout return of -3.1086%. The short diagnostic baseline train return was +5.0551%, with a drawdown of 7.5176% and 63.64% wins, while the holdout return was +2.2679%, with a drawdown of 1.8368% and 2 wins. The short stress train return was +2.7823%, with a holdout return of +1.8680%. However, shorting itself is outside the mandate and positive stock returns do not cover put spread and theta. A CLI command was used to reproduce the full sample, resulting in a return of +114.1606%, with a drawdown of 30.9747% and 34.21% wins, across 685 bars. The CLI aggregate lacks explicit holdout and cost settings and is a separate diagnostic that cannot override failed event gates. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars, with dates ranging from 30869 to 30894. The data selection and history are already known to prior managers, and the chronological holdout is not new prospective evidence. Actual historical option quotes and 15-minute exits remain unvalidated.

A news scan found the headline is not available, but a data line was found with the following information: Ran backtestCLI SFIX/BB breakout20 days1000 stop10%target30%, LWbreakout252 days1600. The results of the CLI aggregate are as follows: SFIX return was -7.1357, drawdown was 38.6560, win was 38.0952, and the number of trades was 42; BB return was 114.1606, drawdown was 30.9747, win was 34.2105, and the number of trades was 38; LW return was -1.9840, drawdown was 2.9359, and the number of trades was 4. The CLI has no explicit cost or holdout flags, it is for diagnostic purposes only. Separately, vectorbt was run from 2022 to 2025 for training and 2026 was held apart, with the following conditions: close price is greater than the prior high 20, 63, or 252 days, plus volume is greater than 1.5 times the prior 20 days; entry is at the next open, exit is at the next open after the close is less than the SMA 50, and force window end. Fees are 1 basis point per side, and slip is 5 basis points for the baseline and 15 basis points for stress. The results for LW are as follows: train 20 to 40 was -40.7232%, train 63 was -21.6027%, and train 252 was 4.1088% with 2 trades; the selected 252 holdout had 0 trades and 0 return, and cannot be validated. SFIX had negative returns for all train periods; the 20 holdout period had a return of -14.2149%, a drawdown of 22.2933, and 1 trade. BB had negative returns for all train periods; the 20 holdout period had a return of 127.4192%, a drawdown of 30.7207, and 1 trade, but this must not be cherry-picked after losing training. Stress test results for the 20 holdout period are as follows: LW had a return of 2.3640, SFIX had a return of -14.3863, and BB had a return of 126.9649%; a total of 36 cells were recorded with observations. The rules differ from CLI intentionally, and this is not claiming replication. There is no current LW or SFIX 52 week breakout. These stock diagnostics do not price options, justify shorts, or override earnings entry rule. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars. No test establishes a new tradable strategy, and known historical holdout is not independent prospective discovery.

This is a pre-event option audit for R92006. The audit had certain premises: it looked at regular sessions before BMO earnings, with an exit before the close of the prior session, and a target relative strike of 40/46.67, rounded to the nearest 2.5. The first standard monthly expiry had to be at least 7 days after the event. The audit requested 11 OCC paths via option bars, but only 9 were available, and only 8 endpoints were available. The audit found that 8 endpoint pairs all lost money after adding 0.05 to the entry and subtracting 0.05 from the exit. Only 2 training tickets fit the $50 criteria, but neither had complete 11 daily bars. The historical open interest and bid asks were not verified. The training results showed a mean loss of 35.5830%, a compounded loss of 93.2563%, and a drawdown of 93.2563%, with 0% win rate. The 2026 holdout results showed a mean loss of 49.9256%, a compounded loss of 75.4464%, and a drawdown of 75.4464%, with 0% win rate. A vectorbt short underlying direction proxy, which is not allowed in actual trades, showed a 7.7635% gain in training and a 12.2338% loss in holdout. The audit concluded that event tail profit cannot be borrowed into mandatory pre-event exit. The data source is https://runway.adityaperswal.workers.dev/data/option-bars and /data/stock-bars, with dates 30507-30517. This is a known historical chronological holdout, not a prospective independent test.