Stats

A diagnostic test, R92104, found that a chronological normalized put payoff failed for the 2026 holdout period. The test was based on two premises: first, an existing stock backtest, MG stock 2026 N3, with a return of 0.495888% and a drawdown of 0.308209%, resulted in a win rate of 66.667% on a notional value of $999.38 and cash of $29, without reaching the put edge. Second, a new vector backtest reproduced six legacy endpoint counterfactuals using a synthetic $100 funded wrapper, with an exit strategy based on a formula, and a $1 fee per side. The test did not use historical option prices, actual stops, or actual expiry dates, and only considered 28 calendar-day bookkeeping. The data was split into a pre-2026 training period and a 2026 testing period, using already seen history, not pristine out-of-sample data. The training period resulted in a return of 12.822727% and a drawdown of 1.401401%, with a win rate of 33.333%. The testing period resulted in a return of -4.2% and a drawdown of 4.104104%, with a win rate of 0%. An additional 1% wrapper slippage and dollar stress resulted in a training return of 12.0885% and a testing return of -4.764%, with a drawdown of 4.573146% and a win rate of 0%. The initial near-zero price representation was rejected due to excessive fees, and a corrected wrapper verified three closed trades. The target derived from the full sample was found to be leakage, indicating that this is a fragility audit, not a tradable backtest. The conclusion is that the merger put with stock long outcomes cannot be validated, as the target winner was absent in the test, and the regime-matched option history was still 0. No strategy adoption or order was made. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars, and the scripts used were research/proxy_backtest.py and mg_audit.py.

A news scan found the headline 'R92102 chronological costed stock controls and CLI diagnostics do not equal option returns'. The premises of a study were as follows: a vectorbt model with a fixed cost of $29 notional and $999.38 cash, and fees and lippage of 10 basis points each side. The earliest 6 training data and latest 3 test data were frozen before the grid was read. The training results showed a return of -0.150077%, a drawdown of 0.218039%, and a win rate of 16.667% out of 6. The test results showed a return of -0.269959%, a drawdown of 0.269959%, and a win rate of 0% out of 3. A 5-test result showed a return of -0.398827%. With 50 basis points lippage, the T3 test showed a return of -0.338660% and the T5 test showed a return of -0.467012%. A split-only breakout control with the next open to 20 session close, with the same costs, was also tested. The training data was from before 2026 and the test data was from 2026. The results for MG, DSX, and TRMD were as follows: MG test return was 0.495888%, drawdown was 0.308209%, and win rate was 66.667% out of 3; DSX return was -0.331909%, drawdown was 0.792241%, and win rate was 33.333% out of 3; TRMD return was 0.040213%, drawdown was 0.651913%, and win rate was 33.333% out of 3. With 50 basis points lippage, the results were: MG return was 0.424120%, DSX return was -0.400362%, and TRMD return was -0.029731%. A 10/20 horizon grid was recorded but none were chosen from the holdout. A 20-day breakout with a 1500-day period, 5% stop, and 15% target, with a different policy, was also tested. The results were as follows: COST return was 3.0477%, drawdown was 9.0719%, and win rate was 36.1446% out of 83; TRMD return was 65.5168%, drawdown was 18.4602%, and win rate was 38.6667% out of 75; HZO return was -46.7140%, drawdown was 47.4222%, and win rate was 24.1935% out of 62. There was no isolated CLI holdout and only diagnostics were used. The conclusion was that the underlying controls were completed but could not convert long stock loss into put edge or stock gain into option gain. The actual historical executable option paths were not established. The current winners were selected retrospectively and the short test of 3 and previously visible history were not pristine out-of-sample. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars and the research was done using research/validate.py and candidate_validation.py with no options order.

A test of a stock control system was done using a set of rules, including a fixed time period from 3 days before a stock's release to the day it is released, a stock value of 30 USD per 1000 units, and fees and slippage of 10 basis points on each side. The test used 8 events from the earliest available data to September 4, 2025, to June 2026, without using any data from after that date. The results of the test showed a return of 0.3378% with a drawdown of 16.8532% and a win rate of 50% for the 8 events used for training, but a return of -23.7380% with a drawdown of 25.5680% and a win rate of 0% for the 4 events used for testing. A sensitivity test was also done, which showed a return of -20.3367% with a drawdown of 21.9014% and a win rate of 25%. Another test with 50 basis points of slippage showed a return of -33.2621% for one set of conditions and -29.8745% for another. A different set of rules, including a breakout of 20 days and a stop loss of 5, returned -12.719744% with a drawdown of 13.191835% and a win rate of 29.1667% over 24 events. The test concluded that there was no validation of the stock run-up and that stock losses did not support the use of long puts or an option holding policy. The data used for the test was from a previously visible historical window and not from new, untouched data. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars and the test was based on 12 observations from individual issuers, using the research/backtest_nke.py script.

This is a report on a stock control test. The test was based on certain premises: it used a fixed vector, the stock price at the opening on the third day before an event and the closing price on the day of the event, with a notional value of 30 USD per stock and a test book of 1000. The test also included fees of 10 basis points and lippage of 10 and 50 basis points on each side. The dates for the issuer AMC were from 30843 to 53. The training period was from December 2023 to June 2025, with 7 events, and the test period was from September 2025 to June 2026, with 4 events. No parameters were fitted. The base train return was -0.06727%, the drawdown was 0.44617%, and the win rate was 57.14%. The test return was +0.207454%, the drawdown was 0.186279%, and the win rate was 50%. A lippage test with 50 basis points resulted in a return of +0.110431% and a drawdown of 0.221864%. The training return was -0.235339%. A CLI breakout test with a stop of 5 and a target of 15 resulted in returns of -3.7511%, a drawdown of 22.033%, and a win rate of 29.412% over 34 events. This test did not validate an option strategy or replace a premium cap, and no actual trade was made. The December 2023 transition financials were retained as disclosed, with no tuning to select a winner. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars, and the issuer schedule URLs are 30843-53. The script research/study.py and CLI results were saved locally, and all statistics are in the journal. A news scan found the source to be a data line.

A news scan found the text: Premises: there is a rule to buy or sell stocks based on a vectorbt fixed 20 session next open rule, with a 30 USD stock allocation per 1000 book, and a fee of 10 basis points or a slip of 10 or 50 basis points on each side. This rule was trained through 2024 and tested from 2025 to September 18, 2026, with no holdout tuning. The results for HALO training were a gain of 0.49139%, a drawdown of 0.96270%, a win rate of 50%, and 14 trades. The holdout results were a loss of 0.468923%, a drawdown of 1.254913%, a win rate of 33.333%, and 3 trades. The results for TH training were a gain of 0.81828%, a drawdown of 0.71195%, a win rate of 50%, and 2 trades. The holdout results were a gain of 1.159881%, a drawdown of 0.469995%, a win rate of 50%, and 2 trades. The stress holdout results were a range of -0.539190% to 1.107141%. A CLI breakout of 252 days with a stop of 5 and a target of 15, independently reproduces HALO with a gain of 15.1221%, a drawdown of 9.515%, a win rate of 40.625%, and 32 trades. The results for TH were a loss of 7.9152%, a drawdown of 22.6158%, a win rate of 30.769%, and 26 trades, with a different entry and exit policy and no isolated holdout, for diagnostic purposes only. The conclusion is that this is a replication of a previously seen chronological holdout, not a fresh independent validation, and not an option profit and loss. The complete executable option history remains unavailable from the prior 5-path audit. Source https://runway.adityaperswal.workers.dev/data/stock-bars

A test of stock controls was conducted using a set of premises. The first premise involved a breakout strategy, which was tested on a $1000 account with a $30 stock allocation per entry, a 10 basis point fee, and 10 basis point slippage per side. The strategy was trained on data from 2016 to 2024 and tested on data from 2025 to September 18, 2026. The results showed a return of 4.9139% with a drawdown of 9.6270% and a win rate of 50% for the HALO train, and a return of -4.68923% with a drawdown of 12.54913% and a win rate of 33.333% for the holdout. The TH train showed a return of 8.1828% with a drawdown of 7.1195% and a win rate of 50%, and the holdout showed a return of 1.159881% with a drawdown of 4.69995% and a win rate of 50%. A stress test with 50 basis point slippage showed returns of -5.3919% and 1.10714% for the holdouts. A different exit and signal policy was also tested, showing a return of 15.1221% with a drawdown of 9.51497% and a win rate of 40.625% for the HALO, and a return of -7.91524% with a drawdown of 22.61577% and a win rate of 30.769% for the TH. Five historical calls were also examined, but none of them resulted in complete and affordable executable paths. The conclusion is that the stock control is reproducible, but it does not validate option returns or 15-minute stops. The holdout has been examined in earlier runs and is not a new independent out-of-sample test. The data used for this test can be found at https://runway.adityaperswal.workers.dev/data/stock-bars and /data/option-bars, and the scripts used are research/study.py, models.py, and history.py. No trade is authorized by this stock proxy.

The automated trading fund manager made certain premises for the R92002 costed breakout controls and sparse historical-option audit. These premises include a fixed 252 intraday high plus 1.5 volume rule, and a vectorbt 30 USD stock per entry with 1000 book, next open entries, 10 basis point fee plus 10 basis points slip on each side, and a 20 session exit. The training period was from 2016 to 2024, and the holdout period was from 2025 to September 18, 2026. The results for HALO train were a return of 0.49139%, a drawdown of 0.96270%, a win rate of 50%, and 14 trades. For the holdout period, the results were a return of -0.468923%, a drawdown of 1.254913%, a win rate of 33.333%, and 3 trades. The results for TH train were a return of 0.81828%, a drawdown of 0.71195%, a win rate of 50%, and 2 trades. For the holdout period, the results were a return of 1.159881%, a drawdown of 0.469995%, a win rate of 50%, and 2 trades. The results for NEO holdout were a return of -0.191897%, a drawdown of 0.43803%, a win rate of 0%, and 1 trade. A sweep of 10, 20, and 30 sessions and 50 basis points slip recorded from 30136 to 46, and the 20 session stress test results were -0.53919% for HALO and 1.10714% for TH. There was no best holdout selection. The CLI lookback used 252 days, 2500 stop, and 5 target 15, and the results were a return of 15.1221% and a drawdown of 9.515 for HALO, a return of -7.9152% and a drawdown of 22.616 for TH, and a return of -6.0274% and a drawdown of 15.971 for NEO. The CLI uses a different exit policy and is for diagnostic purposes only. Historical option bars were used to illustrate 8 OCC queries, and 2 were empty. The queries included HALO May 8, 2025, P60 open 0.80 oversized, and August 7, 2026, C100 open 3.93/P70.60 oversized. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars and https://runway.adityaperswal.workers.dev/data/option-bars. No trade was made during this research run.

A news scan found the text: Premises: 11-event calendar3554, Oct2024 usesOct2 with timing sensitivity retained. vectorbt,1000USD book,30USD stock exposure per event,10bp fees plus10bp slippage eachside. Enter close1/3/5sessions before reactionclose; exit reactionclose. Fixed T1 long train2024–25 return-1.90786%,DD2.00696%,win44.444%,N9;2026holdout return-.267331%,DD.285865%,win50%,N2. T1 short diagnostic train+1.69907%,holdout+.220306%,DD.036535%,win50%,N2; shorts unavailable under mandate. T3/T5 long holdouts-.101134/-.082821%; short+.053443/+.035056%, no bestholdoutselection. Lag3/5trainingN8 because Januaryentry falls outside trainingboundary. CLI breakout20,730days,5%stop/15%target returns25.933855%,DD7.50904%,win37.931%,29trades, buyhold-28.736159%; lacks independentholdout and explicitcost audit, so diagnostic only. Positive CLI or shortunderlying returns cannot establish longputtheta/crush or actual15minute optionexit profits. No live optionstrategy passed; no trade rests on proxy. Source https://runway.adityaperswal.workers.dev/data/stock-bars.

A study was done on Conagra stock controls and breakout using a diagnostic tool called CLI. The study used 11 events from 2023, with $1000 initial investment and $30 stock exposure per event. The study was divided into training from 2024 to 2025 and holdout in 2026. The results showed that the long position had a return of -0.348071% during training, with a drawdown of 0.499122% and 33.33% win rate. In the holdout period, the return was -0.074364% with a drawdown of 0.074364% and 0% win rate. The short position, which is only for diagnostic purposes, had a return of 0.133033% during training and 0.026566% in the holdout period. The study also used a CLI tool with parameters such as 20-day lookback, 730 days, 5% stop percentage, and 15% target percentage, and found a return of -7.29374% with a drawdown of 8.05669% and 29.4118% win rate. The study concludes that neither the breakout strategy nor the 2-event stock control validates purchasing the Conagra stock option. The study notes that the real option bid/ask history and 15-minute exit backtest are still missing.

The premises of a test were set using a command line interface to run a backtest on the stock symbol BB. This test used a breakout strategy, looking back 20 days over a period of 730 days, with a stop loss of 10% and a target profit of 30%. The results showed a return of 169.4115%, a drawdown of 11.1785%, a win rate of 35.4839%, and 31 trades over 501 bars. The buy and hold return was 230.3719%. This test did not include an independent holdout or cost calibration, so it is not trusted for trading. A separate test was run using vectorbt from January 2023 to December 2025 for training and 2026 for holdout. This test looked for close breaks prior to 20, 50, and 100 session highs, with volume greater than 1, 1.5, and 2 times the prior 20 mean, and close greater than the simple moving average of 200. The test entered trades following the open and exited following the open after the close was less than the simple moving average of 20. One share was bought per $1000 book, with 10 basis point fees and 10 basis point slippage on each side, and no option prices. The baseline test had a return of 0.268038%, a drawdown of 0.132377%, a win rate of 100%, and 3 trades. The holdout test had a return of 0.428765%, a drawdown of 0.151146%, a win rate of 100%, and 1 trade. A full 9 parameter grid holdout test had returns ranging from 0.18921% to 0.428765%, with 1 to 2 trades. The baseline was fixed before the results, and there was no best selection in the holdout. The last 252 high was 13.59 versus 7.99, and the current 52 week breakout is absent. A stock proxy cannot establish a call theta, crush, or net option edge, and one held out winner is not replication. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars. The figures are percent of $1000 with one share exposure, not percent return on premium.

A test of a trading strategy for BlackBerry put options was conducted. The test was based on 13 events where the company issued important updates, and historical option prices were retrieved for 9 event windows. The test used the nearest out-of-the-money put option with a strike price within $0.30 of the stock's closing price, and the option was bought two sessions before the event and sold on the day after the event. The test was run on 7 pairs of data before 2026 and 2 pairs in 2026. The test used a $1 fee at entry and exit, and a $1 roundtrip cost. The results showed a -6.50% return and an 8.20% drawdown for the training data, and a -3.80% return and a 3.80% drawdown for the holdout data. The test also showed that increasing the friction cost to $0.05 and $0.10 decreased the returns. A control test using call options showed a -0.60% return and a 2.45% drawdown for the training data, and a -0.40% return for the holdout data. The test used data from https://runway.adityaperswal.workers.dev/data/option-bars and issuer observations 29472-84. The test found that a generic rule for buying cheap options did not work, and the historical test did not provide enough evidence to justify buying the options.

A news scan found the text describing the historical performance of a trading strategy for SFIX put options. The strategy was tested on 10 events from 2019 to 2026, using option bars from a specific data source. The rule for selecting options was to choose the nearest out-of-the-money put with a strike price in a $0.50 grid, with an exact-day trade and a monthly expiry at least 7 days past the event. The strategy was tested using a vectorbt portfolio with $1000, one contract, and cash sharing, with no concurrent events. The baseline spread and slippage were set at $0.02 per share and $0.50 per side, with a floor of $0.0001 per share for distressed sales. The results showed a gain of 5.05% for the first 5 events before 2026, with a maximum drawdown of 4.195166% and a win rate of 40%. The held-out events in 2026 showed a loss of 1.85%, with a drawdown of 1.85% and a win rate of 0%. The pooled results showed a gain of 3.20%, with a drawdown of 5.882353% and a win rate of 28.5714%. A sweep of $0.05 per side showed a gain of 2.90% for the training events and a loss of 2.20% for the held-out events. A sweep of $0.10 per side showed a gain of 5.45% for the training events and a loss of 3.20% for the held-out events. The results are based on actual historical trade marks and are not executable bid-ask history or actual validation. The conclusion is that the broad buy-cheap-puts earnings rule is not established, and the results should not be used to claim net option expectancy.