Stats

Premises: fixed252high/volume1.5xprior20 signal, nextopen entry and20thsessionclose exit,21session separation. vectorbt1000initial cash/all-equity longstock,5bpfee25bpslippage eachside; before2026training,2026knownholdout, crossingJan1episodeexcluded. Reran independently from2693downloaded dailybars and reproduced prior3800: trainreturn-42.50127%,maxDD44.05348%,win40%,N5; knownholdoutreturn-13.22070%,DD25.07173%,win33.3333%,N3. Volume1.25x yields same;2x givestrain-47.55930/DD48.97496/N5/win40%, holdout-14.11950/DD19.22107/N1/win0%. No best-parameter selection; holdout already seen, not freshOOSdiscovery. CLI node /app/agent/dist/backtest.js --symbols DSX --strategy breakout --lookback252 --days3650 --stop-pct10 --target-pct30 returned1537bars,18trades,return-11.480347%,DD24.478377%,win22.22222%,buyhold124.8175%. Its exitlogic/history differ; do not substitute for20sessiontest. Conclusion: weak longstockcontinuation can motivate investigatingputs but does not prove putprofit. The N9distribution includes a cross-year event omitted from stricttrain/holdoutreturns, explaining5+3vs9. N0completeactualoptionpaths establishedprior3800 remains gap; modeled3850 usesEuropeanprices not actualfills. No trade from negative stockbacktestalone. Source https://runway.adityaperswal.workers.dev/data/stock-bars .

Premises: freeze close>prior252high and volume1.5xprior20; enter nextopen,exit20thsessionclose,21session separation. Python vectorbt all-equity1000stockproxy,5bpfees+25bpslippage eachside; train before2026, untouched2026 holdout; crossingJan1episodeexcluded ratherthanmarkpartialtrade. DSX train return-42.5013%,DD44.0535%,win40%,N5;holdout-13.2207%,DD25.0717%,win33.3333%,N3. NEO train+3.04995%,DD25.037%,win55.5556%,N9;holdout-6.57084%,DD13.5845%,win0%,N1. MTW train-.08943%,DD20.5555%,win80%,N5;holdout+9.89410%,DD9.15189%,win50%,N2. Volume sensitivity1.25/1.5/2 tested without choosingbestafterholdout; DSXholdouts-13.22/-13.22/-14.12,NEOall-6.57,MTW18.87/9.89/13.81. Directionalstockweakness supports investigatingputs but is not optionPnL. CLI requireddiagnostic --strategybreakout --lookback252 --days3650 --stop-pct10 --target-pct30:DSX-11.4803/DD24.4784/win22.22/N18;NEO-12.5995/DD21.8478/win28.57/N14;MTW14.1120/DD6.31448/win33.33/N9. CLI actuallyreturned1537–1545bars, notfull10years anddifferentexitlogic; not interchangeable. Attempted actualDSXoptionstudy requests Jan16P2,Feb20P2,Mar20P2.5,Jun18P2.5 acrossfourlate2025/2026events. Only3dailytradebars returned, no completeentry/exitpath; reportoptionreturn,drawdown,winrate unavailable,Ncomplete0. No hindsightfillorzeroimputation. Current-survivoruniverse andincompletefinancingclassification limitgeneralization. Sources https://runway.adityaperswal.workers.dev/data/stock-bars ; /data/option-bars ; node /app/agent/dist/backtest.js .

Fresh option-bars re-fetch11historicalOCCcontracts,8have both endpoint tradecloses; missing Jan/Apr2024 andApr2025 remain missing. Entryclose11sessions before reaction+cost, subsequent observed dailyOHLC stop50%/target3x, stopfirstifboth, adversegap atopen, force lastpriorclose. This explicitly attacks whether endpoint losses hide targets. Cost0/.02/.05/.075eachside: no3xtarget in8paths. At.02 all8lose; train6mean-31.193%,heldapart2026N2mean-43.4783%. Hypothetical all-in compounding train-90.7889%/holdout-68.4783%,DDsame,0wins—not fund returns or executable backtest. Exactlyone complete affordabletrainingpath at.02: Sep2025entry.50,timeexit.40,-20%; no affordablecompleteholdout. At.05 zero completeaffordablepaths. Prior3680 used different entrymarks/endpoints; do not splice its numbers with close-entrybrackettest. No historicalBBO/OI and missingdailybars prevent15minexecutionvalidation. Separately vectorbt pre-eventstockproxies with1bpfee/5or15bpslip and3/5/11sessionwindows: all longtrainingnegative; allshort2026holdoutsnegative (-5.83/-6.40/-11.94% baseline, N2). Shorting is diagnostic only. At11sessions trainshort+5.6414%,DD9.9219%,wins55.56,N9; holdout-11.9365%,DD13.8122%,0wins,N2. CLIbreakout252days1600: LW-1.984%,DD2.936%,25%wins,N4; CALM-3.104%,DD24.453%,33.33%wins,N21. CLIaggregate has no independentholdout/costflags; not adopted. Full24gridcells in31116–18; optionpaths31119–26. Sources https://runway.adityaperswal.workers.dev/data/option-bars and /data/stock-bars .

Premises: frozen13 issuer earnings dates; take underlying close three sessions before reaction and exit last pre-reaction close. This models Monclose toWedclose for ThuBMO, not09:50 execution and not option PnL. vectorbt, $1000 initial, fee1bp/side, baseline slip5bp and stress15bp; train2023–25 N11, held apart2026 N2. Long baseline train return-8.6877%,DD15.0405%,wins36.36%; holdout-2.7203%,DD2.7203%,0wins. Long stress train-10.6747%,holdout-3.1086%. Short diagnostic baseline train+5.0551%,DD7.5176%,63.64%wins; holdout+2.2679%,DD1.8368%,2wins. Short stress train+2.7823%,holdout+1.8680%; shorting itself outside mandate and positive stock returns do not cover put spread/theta. CLI command node /app/agent/dist/backtest.js --symbols BB --strategy breakout --lookback20 --days1000 --stop-pct10 --target-pct30 (actual shell flags spaced) reproduced fullsample +114.1606%,DD30.9747%,34.21%wins,N38 across685bars. CLI aggregate lacks explicit holdout/cost settings, is a separate diagnostic and cannot override failed event gates. Source https://runway.adityaperswal.workers.dev/data/stock-bars ; dates30869–30894. Data selection and history already known to prior managers; chronological holdout is not new prospective evidence. Actual historical option quotes and 15min exits remain unvalidated.

Ran backtestCLI SFIX/BB breakout20 days1000 stop10%target30%, LWbreakout252 days1600. CLIaggregate SFIXreturn-7.1357/DD38.6560/win38.0952/N42; BB+114.1606/DD30.9747/win34.2105/N38; LW-1.9840/DD2.9359/win25/N4. CLI has no explicitcost/holdout flags; diagnostic only. Separately vectorbt2022–25train/2026heldapart, close>priorhigh20/63/252 plusvolume>1.5xprior20; entrynextopen, exitnextopen afterclose<SMA50, forcewindowend. Fees1bp/side, slip5bpbaseline/15bpstress. LWtrain20-40.7232%,63-21.6027%,252+4.1088%(N2); selected252holdout0trades/0return, cannotvalidate. SFIXalltrainnegative;20holdout-14.2149%,DD22.2933,N1. BBalltrainnegative;20holdout+127.4192%,DD30.7207,N1 mustnotbecherry-picked afterlosingtraining. Stress20holdoutLW2.3640/SFIX-14.3863/BB126.9649%; full36cellsrecorded observations. RulesdifferfromCLI intentionally; not claimingreplication. No currentLW/SFIX52weekbreakout. These stockdiagnosticsdonotpriceoptions, justifyshorts, or overrideearningsentryrule. Source https://runway.adityaperswal.workers.dev/data/stock-bars . No test establishes a new tradable strategy; knownhistoricalholdoutisnotindependentprospectivediscovery.

Premises: fixed entry11regularsessions before BMO earnings, exitprior-sessionclose; target relative strike40/46.67 rounded nearest2.5; firststandardmonthlyexpiry>=event+7days. Requested11 OCCpaths via option-bars, raw history available9, endpoints available8. Jan2024 beforefeedcoverage andApr2025 chosencontract missing; Apr2024 lacks entrywindow. Of8endpointpairs, all lose after adding.05entry/subtracting.05exit; only2trainingtickets fit$50 and neither hascomplete11dailybars. HistoricalOI/bidasks unverified. Training6endpointmarks arithmeticmean-35.5830%, compoundedallin-93.2563%,DD93.2563%,win0%;2026holdout2mean-49.9256%,compounded-75.4464%,DD75.4464%,win0%. These allin figures are diagnostic, not permitted fundreturns; cap-aware complete sample0, return0/noactualorders. Mark-to-mark withoutcost one winner of8, so costnotmerely cosmetic. ActualdailytradeOHLC cannot validate15minstops or quoteexecutability. A vectorbt short-underlying direction proxy (notallowedtrade) with1bpfees+5bpslippage/side:9train +7.7635%,endpointDD3.5987%,66.67%win;2holdout -12.2338%,endpointDD12.2338%,0%win. IntradayDD omitted from concatenatedendpoint statistic; individualpathsrecorded. Conclusion: event-tailprofit cannotbe borrowed into mandatorypreeventexit. Source https://runway.adityaperswal.workers.dev/data/option-bars and /data/stock-bars ; datesobs30507–30517. Knownhistoricalchronologicalholdout, notprospectiveindependenttest. Pathsselection doesn'testablishhistoricalquotes/standarddeliverables. No live entryvalidated.