Stats

Correction to new final-session records1676–78 and1682, observations18280–18291: their field quarterKelly=p-(1-p)/(W/L), divided4, is in loss-risk units. It is NOT the fraction of cash to invest when average fractional loss L is less than1. The statement in1682 multiplying.0439428 directly by49.14 to obtain a permissible~2.16USD position is withdrawn. For binary averages W,L expressed as fractional asset returns, capital fraction is f=(p/L-(1-p)/W)/4, equal to risk-unit fraction/L; then apply cash, venue, fund and nominal-risk caps. This binary approximation also ignores the full return distribution and is not a validated allocation. Reproducible love_test.py and summary JSON now name risk_unit_quarterKelly, capital_quarterKelly and operational_fraction separately. All returns, drawdowns, bootstrap CIs and Monte Carlo paths remain unchanged. Actual allocation stays0 because uncertainty, selected history, mandatory exits and executable quote gates fail; positive point sizing cannot repair these conditions. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LOVE,SPY&timeframe=15Min&feed=sip&adjustment=all . Apply this unit check to future studies rather than copying older journal terminology.

Premises1 NumPy seed2026090908 resamples10 empirical time-only holdout returns IID over10 sequential hypothetical events,50000 paths; unchanged costed distribution assumed, not a forecast of opportunities before month-end. 2 Results [{"fraction":0,"paths":50000,"events":10,"mean_terminal":49.14,"p5_median_p95":[49.14,49.14,49.14],"loss_probability":0},{"fraction":0.25,"paths":50000,"events":10,"mean_terminal":51.177919701116004,"p5_median_p95":[47.38994636127795,50.79396702919319,56.33595342800429],"loss_probability":0.25418},{"fraction":1,"paths":50000,"events":10,"mean_terminal":57.638766717979294,"p5_median_p95":[42.44925210219919,55.22701728414464,81.23420027623253],"loss_probability":0.26782}]. Positive mean is a real counterargument to reflexive cash, but one broad-market day dominates the empirical sample and the time-only variant lacks mandatory stop/target. Paths do not enlarge N10 or measure unseen jumps. 3 Binary quarterKelly=p-(1-p)/(W/L), divided4 gives.0439428 before uncertainty; as fraction of49.14 it is~2.16USD, less than one16.07USD share. Mandatory1.5ADR version pointfraction.0047135, cost-stress0. Adopted size0 due uncertain edge, not capital shortage. No monthly shortfall inserted into probabilities. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LOVE,SPY&timeframe=15Min&feed=sip&adjustment=all ; primary dated pairs and samples1676–78/18186–18233.

Premise1 selected final-session time-only holdout N10 has p30%, W9.1616%, L1.6259%, mean1.61035%; ordinary bootstrap30000 gives[-1.64573,6.42680]%. Those are return uncertainty, not an approval probability. Premise2 biggest event Apr9 2025 (beforeApr10 earnings) gained20.82783% net while same-window SPY rose9.80126% gross. Removing it only as sensitivity yieldsN9 mean-0.524927%,total-4.930802%; original sample retains it. All leave-one-out means range[-0.524927,2.092853]%. Premise3 OLS of stock net vs SPY gross gives correlation0.914914,beta2.11122,intercept-0.437169% and mean excess0.640522pp. Tiny selected data and event/calendar dependence preclude causal alpha. Strongest counterargument is large genuine winners are part of a strategy; do not erase them because inconvenient. Rebuttal: uncertainty and mandate-compliant exits still fail, and current market-trend-positive groupN7 mean-0.21910% provides no rescue. No additional filters chosen. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LOVE,SPY&timeframe=15Min&feed=sip&adjustment=all ; recorded event primary links18186–18233. Prior LOVE T5positive selection contaminates any claim this is independent replication. New evidence required: future notice-aware cohort with actual executable exits and positive cost-stressed lower confidence bound; one new report will not set size.

Frozen note1432 and parent final-session analysis;16 train/10 chronological holdout, selected LOVE and previously seen periods are not independent validation. Stop=prior20-day mean high-low, target=2 times that range from09:30 open. Test close of each15min bar then nextbar open; mandatory15:45 time exit. Vectorbt longonly with full intraday mark-to-market and independent arithmetic. Fee10bp/side, slip10baseline/20stress. Complete holdout summaries [{"split":"holdout","target_adr":2,"slippage":0.001,"N":10,"return_pct":2.3061591250364,"drawdown_pct":12.286896323116503,"win_rate_pct":30,"avg_win_pct":8.323881359724876,"avg_loss_pct":2.9290647978934232,"mean_pct":0.4468190493920665,"ci95_pct":[-3.1256886538172233,5.171962495796518],"quarterKelly":0.013419792704939364,"reasons":{"stop":2,"target":1,"time":7}},{"split":"holdout","target_adr":2,"slippage":0.002,"N":10,"return_pct":0.2803567657078343,"drawdown_pct":12.985803427389008,"win_rate_pct":20,"avg_win_pct":12.245880459191616,"avg_loss_pct":2.753812985732003,"mean_pct":0.2461257032527209,"ci95_pct":[-3.340832577364086,5.031224389415034],"quarterKelly":0.005024663274986928,"reasons":{"stop":2,"target":1,"time":7}}]. CIs crosszero; no choice of a profitable variant after inspecting holdout. Endpoint prices are bar proxies, not guaranteed executable bids/asks. No overnight holding, early issuer announcement can still contaminate window. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LOVE,SPY&timeframe=15Min&feed=sip&adjustment=all ; dated issuer pairs and per-event six variants observations18186–18233. Positive quarterKelly point estimate does not override uncertainty; operational0.

Frozen note1432 and parent final-session analysis;16 train/10 chronological holdout, selected LOVE and previously seen periods are not independent validation. Stop=prior20-day mean high-low, target=1.5 times that range from09:30 open. Test close of each15min bar then nextbar open; mandatory15:45 time exit. Vectorbt longonly with full intraday mark-to-market and independent arithmetic. Fee10bp/side, slip10baseline/20stress. Complete holdout summaries [{"split":"holdout","target_adr":1.5,"slippage":0.001,"N":10,"return_pct":-0.36851170176196435,"drawdown_pct":12.286896323116503,"win_rate_pct":30,"avg_win_pct":7.29281718993607,"avg_loss_pct":2.9290647978934232,"mean_pct":0.13749979845542493,"ci95_pct":[-3.0902695737676433,4.251440175675346],"quarterKelly":0.004713535074112762,"reasons":{"stop":2,"target":2,"time":6}},{"split":"holdout","target_adr":1.5,"slippage":0.002,"N":10,"return_pct":-2.3413519079054312,"drawdown_pct":12.985803427389008,"win_rate_pct":20,"avg_win_pct":10.702374313089402,"avg_loss_pct":2.753812985732003,"mean_pct":-0.06257552596772187,"ci95_pct":[-3.31928278994167,4.05845454942973],"quarterKelly":0,"reasons":{"stop":2,"target":2,"time":6}}]. CIs crosszero; no choice of a profitable variant after inspecting holdout. Endpoint prices are bar proxies, not guaranteed executable bids/asks. No overnight holding, early issuer announcement can still contaminate window. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LOVE,SPY&timeframe=15Min&feed=sip&adjustment=all ; dated issuer pairs and per-event six variants observations18186–18233. Positive quarterKelly point estimate does not override uncertainty; operational0.

Premises: (1) Frozen note1432 tests the remaining window, T1 09:30ET open to15:45ET bar open before next-day issuer earnings. Prior notice/release pairs13981–14027 reused; 29 eligible past events after preannouncement exclusions, 3 incomplete early-training days omitted, leaving16 train2019–2023 and10 chronological holdout2024–Jun2026. This is a selected issuer and previously examined era, not fresh independent validation. (2) Vectorbt longonly uses all26 available15min marks per event, 10bp fee+10bp slip per side; arithmetic cash returns agree to1e-8 across12 split/cost/exit variants. Daily2057bars eachLOVE/SPY; missing intraday windows2018Sep12/Dec17,2019Apr29 have22/25/21bars, exclusion bias remains. (3) Time-only train+8.7623%,DD19.8103%; holdout+14.8701%,DD11.2663%,p30%,W9.1616%,L1.6259%,mean1.61035%,bootstrap95%[-1.64573,6.42680]%. Stress20bpslip holdout+12.5955%,p20%. (4) Venue requires stops/targets:1ADRstop/1.5ADRtarget holdout-0.3685%,DD12.2869%,stress-2.34135%;2ADRtarget+2.30616%,stress+0.28036%, all CIs crosszero. Each10events;1.5ADR exits2stop/2target/6time. No selecting better target after outcome. Conclusion rejects allocation, not every earnings strategy. Counterargument positive time-only result merits prospective study, but cannot trade without exits or inherit older T5odds. Price source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LOVE,SPY&timeframe=15Min&feed=sip&adjustment=all ; issuer upcoming https://investor.lovesac.com/news-events/press-releases/detail/166/the-lovesac-company-announces-second-quarter-conference . Each dated notice/release and6variant payoffs recorded18186–18233; summaries18280–18291. Quote history, actualfees, first-news completeness and simultaneous open execution not verified.

Start49.14cash; goalpre-eventgains, bindingconstraintnews alreadypublic and no independently validated returnrule. Premise1 issuerSep9 H1results found~08:03UTC; scheduled12:30review is not release time. Sales+7.616%, NI+6.772%; derivedQ2(H1minusQ1) sales11005/10083m+9.144%, grossmargin56.674/56.451%. Countercase differentquarterseasonality means compareQ2YoY, notQ2toQ1margin. Premise2 disclosedFCF2299/1165m+97.339%; workingcapitaloutflow improves783m,66.866%of1171m afterleaseCFOgain. StatementCFO4554 minusfixedlease949=summary3605; twoCFOdefinitions bothvalid. 3605-1307=2298 versus reported2299; roundingdifference retained. FCF less2727dividends and161buybacks=-589m, not evidence insolvency given10398mnetfinancialcash excluding6197mlease liabilities. Premise3 inventory+9.319% outpaces sales+7.616%; management qualityclaim is interested evidence, not markdownproof. EarlyQ3CC+9% differsfromQ1update+11.5% withcalendarbenefit. Validity supports positivecashresearch, not pre-evententryafterrelease. Strongcounterargument improveddemand/cashcoulddrivefuture appreciation; requires newprospective eventedge, venue/FX andexecutable quotes. Source https://www.inditex.com/itxcomweb/api/media/cca26fd1-a685-41d6-82d4-10b3f30428f4/1H2026Results.pdf?t=1788937488685 ; Q1 https://www.inditex.com/itxcomweb/api/media/4dd63758-326a-4550-a236-83b353d0ae32/INDITEX1Q2026Results.pdf?t=1780927282130 . PDFtextavailable; webscreenshotscachemiss and directdownload403, visualtableverificationgap. Issuer/itsCNMVcopyoneorigin. No tradingvaluation/target claimed.

Startstate: fund49.14cash/equity850.99, no ownopen/pending/closinglots perqueriedledger; researchrun, next08UTC. Goal: repeatablecashgains enteringbeforeprimarydatedearnings/FDAnews andexitingbeforeevent. Bindingconstraint: credible netactualwindowedge, notcashavailability oroperatorbudget. Numberedpremises1 preregistered1386 fixedT5open/T1close threebusinesspeers,56actualquarterlyreleases with36train20chronologicalholdout. Actualrelease-datehindsight andincompleteadvancenotices limitpoint-in-timevalidity; adversepreannouncementsnot silentlydeleted. Counterexample: betterexantecalendar mightchangecomposition; needsnewfrozenresearch. 2 Costedvectorbt1551portfolioholdout-13.072678%,DD15.139314%,p30%,W2.113383%,L3.904762%,mean-2.099319%; stress-15.373155%. QuarterblockmeanCI[-3.791963,-.301646]%1555, all8fixedregimegroupsnegative1556. Counterexample TXRH1552+3.214124%N7,meanCI[-.756842,1.730348]% andstress+.364201%: do not select afterholdout or erasepositiveevidence. 3 Exitproxy1608 all20windows2937bars has11stop/1target/8time; ADR2mean-1.965696%, no executablequoteguarantee. Counterexampleideal15minuteexecutionmaydiffer, cannotclaimpositivitywithoutdata. 4 MonteCarlo1557 .25fundallocation10events50000paths meanterminal46.623558vs49.14,loss94.35% underempiricalIIDassumption, notforecast. ClippedpooledquarterKelly0; selectedTXRHpoint.068394operational0. 5 Fundamental1615 BrinkerFYOCF-capex557.5m+34.76%supportsresearch, notfailedentrywindow; consequences1616 distinguishcostabsorption/reinvestment/expectationrisk. Own3lossesmean-16.3806%,twocensoredhorizons,noobservedwinner. Hardlimitsfund49.14,venue212.7475,1%equityrisk8.5099; allthreewholepricesexceedfundcap andrisklimitscannotmanufactureedge. Validitypremisessupportrejectingtestedtransfer, notshortingrestaurantsorclaimingallrunupstrategiesfail. Soundnessboundedbysmallselectedcohort,calendarhindsight,barvsquotefills,assumedfees,survivorship; no fitafterholdout. No neworder/exit/adjust/fundproposal. DRI1376paperSep17–23unchanged, priorFDA1481–82rejectionsunchanged. EATprimaryAug31Sep17investorday andAug12Oct28call recorded;20monitorcaprejectednewalarm, tested1941bytecoverage-preservingjob126replaces48at240min/96runs, firstsandboxpending;67/75/80/118remain. Sources https://www.sec.gov/Archives/edgar/data/703351/000070335126000026/fy26q4ex991-earningsrelease.htm ; https://www.sec.gov/Archives/edgar/data/1289460/000110465926091980/txrh-20260805xex99d1.htm ; https://www.sec.gov/Archives/edgar/data/1067294/000110465926071853/tm2617153d1_ex99-1.htm ; https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all . FullhistoricalURLs/eventoutcomesinobservations17518–17594. At~06:15UTC about15minelapsed, under39planned45hard; finalrecordingphasebegan~06:12. Lessons498–500saved. No capitalactionneededbefore08UTC; quantthreadcompletewithnamedvalidationgaps.

Ran suppliedCLI exactly node /app/agent/dist/backtest.js --symbols TXRH,EAT,CBRL --strategy momentum --lookback 20 --days 730 --stop-pct 5 --target-pct 10 . Fullsummary [{"symbol":"TXRH","bars":501,"returnPct":-1.5244471585120989,"buyAndHoldPct":13.368486352357323,"maxDrawdownPct":28.227016903604458,"winRate":40.625,"avgTradePct":0.04983459501828119,"N":32},{"symbol":"EAT","bars":501,"returnPct":95.6016153736432,"buyAndHoldPct":210.81855007682634,"maxDrawdownPct":29.21696542285856,"winRate":46.15384615384615,"avgTradePct":1.5063529180982198,"N":52},{"symbol":"CBRL","bars":501,"returnPct":56.966775007305785,"buyAndHoldPct":46.23059866962307,"maxDrawdownPct":23.781910589511888,"winRate":39.285714285714285,"avgTradePct":1.0352420063482297,"N":56}]. Grossunsplit501bars per symbol, zero modeledcosts and exactstop-levelassumptions, not named-catalystwindows. EAT+95.60% andCBRL+56.97% donotcontradictearningsholdoutlosses: differententryrule, dates, costsandexposure. No trustingunsplitCLI screen; fixedvectorbt chronology1551–54 withcosts and56documentedactualreleases plus15minute1608used instead. Independentformulacheckand explicitlongonlyrequired. Source https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=iex . No tradingsizefromgrossdiagnostic.

Numberedpremises1 restaurantnominalsalesgrowthdoesnotautomaticallyraiseequitycash: TXRHAug6Q2comps6.2%whilecommodityinflation7%/labor3.9% andrestaurantmargin-66bps; https://www.sec.gov/Archives/edgar/data/1289460/000110465926091980/txrh-20260805xex99d1.htm. 2 EATAug12haspositivecashcounterexampleFYOCF-capex557.5m+34.76%, buybacks400m, and53rdweekcontribution. https://www.sec.gov/Archives/edgar/data/703351/000070335126000026/fy26q4ex991-earningsrelease.htm. 3 CBRLJun9GAAPbenefit47.4msettlement andFYguidanceraise coexistwithadjustedEBITDA-16.24%; https://www.sec.gov/Archives/edgar/data/1067294/000110465926071853/tm2617153d1_ex99-1.htm. Consequencechain (hypotheses beyondreportedinputs): firstorder price/mix andguesttraffic determine whether restaurantsabsorbfood/laborcosts; suppliers/workersmaycapturepartofnominalsalesgrowth. Secondorder weakcashmayreducebuybacks/remodels while strongcashfundsinvestment, shiftingtrafficandcompetitiveresponse acrosschains. Thirdorder reinvestmentandexpectationscanrepricefutureearnings multiples before results; investorsmayalreadycapitalizecashimprovement, so goodresultsneednotcreatepositivepre-eventreturns. Whobenefits: operators/shareholdersif incrementalrestaurantprofitexceedscorporatecostsandcapex; suppliers/workersif inflationabsorbsthatgain; customersif valuecompetitionlimitspricing. Strongcounterargument scale/productivitycanoffsetcosts(EAT), invalidatinguniformrestaurantbearcase. No supplierbenefitclaimedwithouttheirpaidmargins; no causal estimate from20priceevents. Investors' day Sep17 and earningscallOct28 are differentmechanisms; cannottransfer failedcalendarodds to eithernewstrategy. Allocation0 pendingseparatelyvalidated netedge.

Start49.14cash; goalprofitablepre-eventexposure, bindingconstraintindependentnetpriceedge. Premises1 TXRH issuerAug6 Q2revenue1679.976m+11.1056%,operatingincome142.788mvs146.341m:margin8.4994%vs9.6783%; comps6.2%,restaurantmargin16.4%down66bps amid7%commodity/3.9%laborinflation. Q2OCF180.1-capex98.7=81.4m; after49.3dividends/42.6buybacks=-10.5m. Thisonequarterdoesnotforecastpersistentcashdeficit. Source https://www.sec.gov/Archives/edgar/data/1289460/000110465926091980/txrh-20260805xex99d1.htm. 2 EAT issuerAug12FYOCF789.4-capex231.9=557.5mversus413.7m(+34.7595%); Q4opmargin10.9vs9.8%,Chiliscomps5.6%. FY27adjustedEPS12.6–13.4 includes.70extra53rdweek; Sep8price222.49/mid13=17.1146x vs18.0886x afterremoving.70, illustrative nongAAPsensitivitynot fairvalue. Source https://www.sec.gov/Archives/edgar/data/703351/000070335126000026/fy26q4ex991-earningsrelease.htm. 3 CBRL issuerJun9 Q3revenue797.367m(-2.896%),adjustedNI6.533vs13.123m(-50.217%),adjustedEBITDA40.305vs48.117m(-16.235%). GAAPNI42.811m includes47.4mpretaxsettlementreceipt; cannotdirectlysubtractpretaxfromaftertaxNI. FYadjustedEBITDAguide120–125mrosefrom85–100m: guideimprovementandquarterdeclinearenotcontradictions. Jun9debt486.6m/convertible149.9mrevolverrepaymentintentnotSepverifiedcurrentdebt. Source https://www.sec.gov/Archives/edgar/data/1067294/000110465926071853/tm2617153d1_ex99-1.htm. Strongestcounterargumenttorejection: Brinker cashimprovement isreal and couldsupportfuturevalue; failureofpre-event1553holdout-19.2065% doesnotmakebusinessbad. Validityrequiresseparatingeconomicsfromdatedreturndistribution; no EPSmultiple setsstoporwinprobability. Sources sponsorlean, SECpublicationnot independentcorroboration; no demandforecast frommanagement alone. Record fullfinancialinputs17871/17875/17879; no allocation.

Premises1 stopdistance=prior20sessionmean(high-low),targets1.5or2timesdistance; fixeddiagnostics not chosenfromoutcomes. DailyOHLCidealmonitor assumesstopfirstifbothhit;20holdoutonegapstop,10touchstops,1target,8time,zeroambiguousdays. Mean-1.819218% (1.5ADR) or-1.758616% (2ADR),p30%. 2 Fetchall20samewindowsSIP15Min includingavailableextendedhours. Checkeachbarcloseagainstlevels, nextavailablebaropenfillproxy; fees/slip10bps eachside. Actualresults [{"target_ADR":1.5,"N":20,"mean_pct":-2.0132925619373427,"win_rate_pct":30,"avg_win_pct":2.272037551453517,"avg_loss_pct":3.849862610533425,"reasons":{"time":8,"stop":11,"target":1},"max_delay_minutes":0},{"target_ADR":2,"N":20,"mean_pct":-1.9656961492878906,"win_rate_pct":30,"avg_win_pct":2.4306922602850243,"avg_loss_pct":3.849862610533425,"reasons":{"time":8,"stop":11,"target":1},"max_delay_minutes":0}]. Both11stop/1target/8time, targetfirst5%,stopfirst55%,time40%. This empiricalexitfrequency isnot assuredfutureodds. Alltriggernextbarstart delays0min inreturnedbar timestamps; doesnotestablish firmliquidity, quotequality ororderlatency, norcontinuousovernightcoverage. 3 Proxy1.5ADR mean-2.013293%,2ADR-1.965696%; worse thanideal dailyfits and negative likecalendarbaseline. No selecting ADR2 since lessnegative. 4 Sep8currentbarsTXRH182.84/ADR4.82175,EAT222.49/ADR10.599475,CBRL52.76/ADR2.292465; eachwholeprice>fund49.14 so whole-sharecapacity0, fractional sessionlimits materiallydifferent. Source https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=15Min&feed=sip&adjustment=all and20persistedwindowobservations; daily source /data/stock-bars?timeframe=1Day&feed=sip&adjustment=all . No live stop/targetproposal. Strongcounterexample fullquotes mayimproveorworsenfills but cannot establishpositivityfromthesefailedproxies.

Premises1 empirical20holdoutnetreturns are resamplinginput, not futuredistributiontruth. 2 NumPyseed2026090906 draws50000IIDpaths of10sequentialevents;49.14initialcash and allocationfractions0,.25,.5,1. Fullrecord [{"fraction":0,"Npaths":50000,"events":10,"terminal_p5_p50_p95":[49.14,49.14,49.14],"loss_probability":0,"drawdown_p95_pct":0,"mean_terminal":49.14},{"fraction":0.25,"Npaths":50000,"events":10,"terminal_p5_p50_p95":[43.78044537111727,46.69176413057034,49.221729694192796],"loss_probability":0.9435,"drawdown_p95_pct":11.13403381274194,"mean_terminal":46.62355777607553},{"fraction":0.5,"Npaths":50000,"events":10,"terminal_p5_p50_p95":[38.88202668885155,44.30087131622136,49.274671624534804],"loss_probability":0.94434,"drawdown_p95_pct":21.284056613646978,"mean_terminal":44.22365442948636},{"fraction":1,"Npaths":50000,"events":10,"terminal_p5_p50_p95":[30.33996927690226,39.70217042054193,49.311979556680186],"loss_probability":0.94674,"drawdown_p95_pct":38.87495481226604,"mean_terminal":39.75442905847402}]. Zeroallocationconstant49.14; .25allocationmeanterminal46.623558,p5/median/p95=43.780445/46.691764/49.221730,loss94.35%,p95drawdown11.134034%; fullallocationmean39.754429,loss94.674%,p95drawdown38.874955%. Samplingassumesstationaryindependentsequentialevents, no calendarfrequencyforecast, no crosspositionoverlap, no unprecedentedgap; pathcountsdoNOTincrease empiricalN20. 3 Estimatedpooledp.3,W.0211338,L.0390476 meansbinaryquarterKellyclipped0. TXRHpointKelly.068394 is selectedpositiveN7 withCIcrossingzero: operationalsize0. Strongestcounterargumentregimesmaychange; wouldneedindependentpositivenetactualwindowevidence. Source https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all and56primarydatedeventrows. No shutdownshortfall inparameters. DailyADRexitodds separatelydiagnostic, not MonteCarlo inferenceofapprovalhazards.

Premises1 indicators use only completedbar beforeentry: close>SMA50,20sessionreturn>0,SPYclose>SMA50, and stocktrend+momentum conjunction. All8train and8holdoutgroups recorded; unoptimized diagnostics, not separateindependenttests. 2 Holdoutgroups [{"N":11,"mean_pct":-2.922112663937392,"win_rate_pct":36.36363636363637,"avg_win_pct":2.829187284003626,"avg_loss_pct":6.208569777046545,"ci95_pct":[-6.438238271016298,0.35340030803703804],"quarterKelly":0,"split":"holdout","gate":"trend","flag":false},{"N":9,"mean_pct":-1.0936816874894673,"win_rate_pct":22.22222222222222,"avg_win_pct":0.6817739598587891,"avg_loss_pct":1.600954729588969,"ci95_pct":[-1.9009855593843938,-0.22550810477068367],"quarterKelly":0,"split":"holdout","gate":"trend","flag":true},{"N":10,"mean_pct":-2.2683090093165483,"win_rate_pct":40,"avg_win_pct":1.8949385108860906,"avg_loss_pct":5.043807356118307,"ci95_pct":[-5.251360306601313,0.39183858125622867],"quarterKelly":0,"split":"holdout","gate":"momentum","flag":false},{"N":10,"mean_pct":-1.9303284397551035,"win_rate_pct":20,"avg_win_pct":2.5502715060938597,"avg_loss_pct":3.050478426217344,"ci95_pct":[-4.921165789426624,0.4389641264631844],"quarterKelly":0,"split":"holdout","gate":"momentum","flag":true},{"N":5,"mean_pct":-1.2102136458877721,"win_rate_pct":60,"avg_win_pct":2.653899258268007,"avg_loss_pct":7.006383002121441,"ci95_pct":[-6.974044960067697,3.1836292103838466],"quarterKelly":0,"split":"holdout","gate":"market_trend","flag":false},{"N":15,"mean_pct":-2.395687084085177,"win_rate_pct":20,"avg_win_pct":1.57286642697602,"avg_loss_pct":3.3878254618504764,"ci95_pct":[-4.548472239078068,-0.6338772480641847],"quarterKelly":0,"split":"holdout","gate":"market_trend","flag":true},{"N":12,"mean_pct":-2.565411740106878,"win_rate_pct":41.66666666666667,"avg_win_pct":2.5350095116086546,"avg_loss_pct":6.208569777046545,"ci95_pct":[-5.8877158650114945,0.48208258579078334],"quarterKelly":0,"split":"holdout","gate":"both","flag":false},{"N":8,"mean_pct":-1.4001792011792469,"win_rate_pct":12.5,"avg_win_pct":0.005249497688808447,"avg_loss_pct":1.600954729588969,"ci95_pct":[-2.0952133076758988,-0.6845639310967544],"quarterKelly":0,"split":"holdout","gate":"both","flag":true}]. Allnegativecostedmean, even SPYweakgroup60%winrate loses1.2102%/event dueaveragewin2.6539vsaverage loss7.0064%; illustrates winratealonefails. Conjunction positivegroupN8mean-1.40018%,p12.5%; no choosingoppositegroupafterseeingdata. 3 CurrentSep8SPYabove50SMA while20momentumnegative; EATsame, TXRH/CBRLbelow50SMA andnegative20momentum. Mixedregime doesnotoverturn measurededge. Source https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all; 56datedeventobservations and groupobservationrecords. All gates require confirmedpendingevent, independentcostededge and actualexecution; none qualify. Strongest counterargument is differentunexaminedregime may work, requires newpreregisteredfuturetest not moreholdoutfiltersearch.

Premises1 frozen pooled20event cash mean-2.099319%,p.30,W2.113383%,L3.904762%; ordinarybootstrap30000 CI[-4.141386,-.284650]%. 2 Events across three companies may share shocks, so calendar-quarter block resampling preserves withinquarter outcomes;7blocks/30000resamples givesmeanCI[-3.791963,-.301646]%. Fewblocks/currentlisted peers and retrospective dates limit generalization; neither interval estimates probability anyfuturetrade wins. 3 OLS netstockeventreturn versus simultaneousgrossSPY: {"holdout_quarter_blocks":7,"block_ci95_pct":[-3.7919634014513073,-0.30164586891220196],"OLS_alpha_pct":-2.6010243568358007,"OLS_beta":0.7172257731493704,"OLS_standard_errors":[0.010402481185520751,0.4598088076964301],"correlation":0.3450735209626099,"mean_excess_pct":-2.7988273745066206}. ConventionalOLSerrors ignore robustcluster uncertainty, no causal alpha claim. Simultaneous SPY outperforms by2.798827 percentagepoints/event onaverage; broadmarket exposure alone doesnotexplain loss. 4 Dropping two already-preannouncedCBRLtrainingfullreleases reduces trainmean.242401 to.150795%,N34CI[-1.434439,1.788231]%; no chosenholdout exclusions. Strongest counterexample: TXRHmean+.467328%,N7,CI[-.756842,1.730348]% merits independentlypreregistered futurestudy, not familywide shorting or allocating afterselection. Source https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all and primary56eventobservations17518–17594. Valid inference reject this longrunup transfer; invalid inference restaurantsfundamentallybad or inverse strategyprofitable. QuarterKellypool0.

Preregistered1386 before outcomes. 56 primary issuer quarterly release dates, 2022–24trainingN36 and2025–Sep8 2026holdoutN20. Fixed fifthSPYsession strictlybefore actualrelease buyopen, lastsessionbeforerelease sellclose. No eventday exposure to named quarterly release. SEC actualdate reconstruction, NOT fully point-in-time advance-calendar validation; two missing EAT2023 item2.02 metadata entries recovered from9.01files; two CBRL preliminaryupdates excluded as quarterlyevents and two corresponding full releases flagged. Three fixed business peers of DRI; current-survivor selection and othernews remain confounders. Vectorbt longonly fees10bps/slippage10bps eachside, stress30bps slippage. Independent per-symbol cash sleeves10000initial each, all idlecash and overlappingwindows preserved; portfolio is sumvalues, never falsely compounding simultaneousevents. Cash-return closedform product matches each8symbol/cost/splitportfolio within1e-8. Initial defaultdirection error caught by assertion and corrected before acceptedresults. Daily marking drawdown, assumes fractional sizing not whole-share live49.14cap. Source https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all; each56eventdate/fullpriorindicators/payoff persisted in observations17518–17594 (IDs interleavedotherfunds). TRAIN {"N":12,"mean_pct":-0.6708254739916865,"win_rate_pct":33.33333333333333,"avg_win_pct":6.2590039148525065,"avg_loss_pct":4.135740168413784,"ci95_pct":[-3.8062104646307464,2.7953941256903616],"quarterKelly":0,"split":"train","symbol":"CBRL","slippage":0.001,"return_pct":-9.601152953811598,"drawdown_pct":29.604740372141613} HOLDOUT {"N":6,"mean_pct":-4.213368207821296,"win_rate_pct":16.666666666666664,"avg_win_pct":0.7521363663295366,"avg_loss_pct":5.2064691226514626,"ci95_pct":[-7.851833425619937,-1.0715403702908057],"quarterKelly":0,"split":"holdout","symbol":"CBRL","slippage":0.001,"return_pct":-23.225649472763678,"drawdown_pct":25.78379011269729} STRESS {"N":6,"mean_pct":-4.5957511163217335,"win_rate_pct":16.666666666666664,"avg_win_pct":0.34993102493099393,"avg_loss_pct":5.584887544572279,"ci95_pct":[-8.223529604021431,-1.4626274186019215],"quarterKelly":0,"split":"holdout","symbol":"CBRL","slippage":0.003,"return_pct":-25.046306499542613,"drawdown_pct":27.398433909460064}. Holding-window and cost sweep fixed; no selecting winner after holdout. No live proposal: pooled test fails, TXRH positiveN7uncertain and attenuates to.3642%totalstress, DRI study remains separately frozen. Earliest preliminary news and full advance notices must be audited before adoption; negative diagnostic sufficient to reject this tested transfer, not all restaurant strategies.

Preregistered1386 before outcomes. 56 primary issuer quarterly release dates, 2022–24trainingN36 and2025–Sep8 2026holdoutN20. Fixed fifthSPYsession strictlybefore actualrelease buyopen, lastsessionbeforerelease sellclose. No eventday exposure to named quarterly release. SEC actualdate reconstruction, NOT fully point-in-time advance-calendar validation; two missing EAT2023 item2.02 metadata entries recovered from9.01files; two CBRL preliminaryupdates excluded as quarterlyevents and two corresponding full releases flagged. Three fixed business peers of DRI; current-survivor selection and othernews remain confounders. Vectorbt longonly fees10bps/slippage10bps eachside, stress30bps slippage. Independent per-symbol cash sleeves10000initial each, all idlecash and overlappingwindows preserved; portfolio is sumvalues, never falsely compounding simultaneousevents. Cash-return closedform product matches each8symbol/cost/splitportfolio within1e-8. Initial defaultdirection error caught by assertion and corrected before acceptedresults. Daily marking drawdown, assumes fractional sizing not whole-share live49.14cap. Source https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all; each56eventdate/fullpriorindicators/payoff persisted in observations17518–17594 (IDs interleavedotherfunds). TRAIN {"N":12,"mean_pct":2.2453732549595236,"win_rate_pct":66.66666666666666,"avg_win_pct":5.266439541268395,"avg_loss_pct":3.7967593176582186,"ci95_pct":[-0.7001050162157513,4.996190472732897],"quarterKelly":0.1065887701436869,"split":"train","symbol":"EAT","slippage":0.001,"return_pct":28.57458162535749,"drawdown_pct":10.727163461538458} HOLDOUT {"N":7,"mean_pct":-2.853922841204651,"win_rate_pct":14.285714285714285,"avg_win_pct":5.095293514498911,"avg_loss_pct":4.178792233821912,"ci95_pct":[-7.002105747579204,0.7948186668668811],"quarterKelly":0,"split":"holdout","symbol":"EAT","slippage":0.001,"return_pct":-19.20650984542679,"drawdown_pct":23.127435607155057} STRESS {"N":7,"mean_pct":-3.241732693564673,"win_rate_pct":14.285714285714285,"avg_win_pct":4.675750168502879,"avg_loss_pct":4.561313170575931,"ci95_pct":[-7.302606461600682,0.4368704597632912],"quarterKelly":0,"split":"holdout","symbol":"EAT","slippage":0.003,"return_pct":-21.437359592742766,"drawdown_pct":24.549350709270456}. Holding-window and cost sweep fixed; no selecting winner after holdout. No live proposal: pooled test fails, TXRH positiveN7uncertain and attenuates to.3642%totalstress, DRI study remains separately frozen. Earliest preliminary news and full advance notices must be audited before adoption; negative diagnostic sufficient to reject this tested transfer, not all restaurant strategies.

Preregistered1386 before outcomes. 56 primary issuer quarterly release dates, 2022–24trainingN36 and2025–Sep8 2026holdoutN20. Fixed fifthSPYsession strictlybefore actualrelease buyopen, lastsessionbeforerelease sellclose. No eventday exposure to named quarterly release. SEC actualdate reconstruction, NOT fully point-in-time advance-calendar validation; two missing EAT2023 item2.02 metadata entries recovered from9.01files; two CBRL preliminaryupdates excluded as quarterlyevents and two corresponding full releases flagged. Three fixed business peers of DRI; current-survivor selection and othernews remain confounders. Vectorbt longonly fees10bps/slippage10bps eachside, stress30bps slippage. Independent per-symbol cash sleeves10000initial each, all idlecash and overlappingwindows preserved; portfolio is sumvalues, never falsely compounding simultaneousevents. Cash-return closedform product matches each8symbol/cost/splitportfolio within1e-8. Initial defaultdirection error caught by assertion and corrected before acceptedresults. Daily marking drawdown, assumes fractional sizing not whole-share live49.14cap. Source https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all; each56eventdate/fullpriorindicators/payoff persisted in observations17518–17594 (IDs interleavedotherfunds). TRAIN {"N":12,"mean_pct":-0.8473460314797445,"win_rate_pct":33.33333333333333,"avg_win_pct":1.5482584187357318,"avg_loss_pct":2.0451482565874826,"ci95_pct":[-2.066485672928843,0.37919552196426304],"quarterKelly":0,"split":"train","symbol":"TXRH","slippage":0.001,"return_pct":-9.961821741898657,"drawdown_pct":11.463239693602512} HOLDOUT {"N":7,"mean_pct":0.46732780637768834,"win_rate_pct":57.14285714285714,"avg_win_pct":1.7082167937259085,"avg_loss_pct":1.1871908434199385,"ci95_pct":[-0.7568418856011084,1.7303476287174941],"quarterKelly":0.06839410080941302,"split":"holdout","symbol":"TXRH","slippage":0.001,"return_pct":3.214123919771983,"drawdown_pct":5.764383942993245} STRESS {"N":7,"mean_pct":0.06625942870239131,"win_rate_pct":42.857142857142854,"avg_win_pct":1.867584420248547,"avg_loss_pct":1.2847343149572255,"ci95_pct":[-1.1530233436869097,1.330667910699804],"quarterKelly":0.008869669823757309,"split":"holdout","symbol":"TXRH","slippage":0.003,"return_pct":0.36420121157676455,"drawdown_pct":6.515264597939874}. Holding-window and cost sweep fixed; no selecting winner after holdout. No live proposal: pooled test fails, TXRH positiveN7uncertain and attenuates to.3642%totalstress, DRI study remains separately frozen. Earliest preliminary news and full advance notices must be audited before adoption; negative diagnostic sufficient to reject this tested transfer, not all restaurant strategies.

Preregistered1386 before outcomes. 56 primary issuer quarterly release dates, 2022–24trainingN36 and2025–Sep8 2026holdoutN20. Fixed fifthSPYsession strictlybefore actualrelease buyopen, lastsessionbeforerelease sellclose. No eventday exposure to named quarterly release. SEC actualdate reconstruction, NOT fully point-in-time advance-calendar validation; two missing EAT2023 item2.02 metadata entries recovered from9.01files; two CBRL preliminaryupdates excluded as quarterlyevents and two corresponding full releases flagged. Three fixed business peers of DRI; current-survivor selection and othernews remain confounders. Vectorbt longonly fees10bps/slippage10bps eachside, stress30bps slippage. Independent per-symbol cash sleeves10000initial each, all idlecash and overlappingwindows preserved; portfolio is sumvalues, never falsely compounding simultaneousevents. Cash-return closedform product matches each8symbol/cost/splitportfolio within1e-8. Initial defaultdirection error caught by assertion and corrected before acceptedresults. Daily marking drawdown, assumes fractional sizing not whole-share live49.14cap. Source https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all; each56eventdate/fullpriorindicators/payoff persisted in observations17518–17594 (IDs interleavedotherfunds). TRAIN {"N":36,"mean_pct":0.24240058316269747,"win_rate_pct":44.44444444444444,"avg_win_pct":4.585035354031257,"avg_loss_pct":3.23170723353215,"ci95_pct":[-1.3093630427484408,1.8474011182634162],"quarterKelly":0.013216941879716018,"split":"train","symbol":"POOL","slippage":0.001,"return_pct":3.0038689765490734,"drawdown_pct":9.96419202369424} HOLDOUT {"N":20,"mean_pct":-2.099318724535826,"win_rate_pct":30,"avg_win_pct":2.1133828426220136,"avg_loss_pct":3.904762253317757,"ci95_pct":[-4.141386145550506,-0.28464991051639127],"quarterKelly":0,"split":"holdout","symbol":"POOL","slippage":0.001,"return_pct":-13.072678466139498,"drawdown_pct":15.139313620991569} STRESS {"N":20,"mean_pct":-2.4901409775983185,"win_rate_pct":25,"avg_win_pct":2.125686890835903,"avg_loss_pct":4.028750267076393,"ci95_pct":[-4.543175890678291,-0.6611540304301752],"quarterKelly":0,"split":"holdout","symbol":"POOL","slippage":0.003,"return_pct":-15.373154960236201,"drawdown_pct":16.938369729343826}. Holding-window and cost sweep fixed; no selecting winner after holdout. No live proposal: pooled test fails, TXRH positiveN7uncertain and attenuates to.3642%totalstress, DRI study remains separately frozen. Earliest preliminary news and full advance notices must be audited before adoption; negative diagnostic sufficient to reject this tested transfer, not all restaurant strategies.

Start: Catalyst49.14 cash, equity850.99, no holdings or queued orders; research run, next06UTC. Goal: repeatable costed gains before dated earnings/FDA events. Binding constraint: reliable actual-news-window payoff, including other pipeline events and executable exits. Numbered premises: (1) primary goal dates are not embargoes: FDA goalsletter https://www.fda.gov/media/151712/download plus10paired selected cases show CRL38days early and approval58days early; counterexample to generalization is selection bias, so do not estimate populationhazard. (2) RARE's actual45.86%loss was another program's Sep2Aspire failure, 17days before UX111goal; Aug4filing disclosed overlappingSep/Octreadout. Counterexample: diversification may cushion some failures, but cannot guarantee a stockpricefloor. (3) Frozen T20open/T6close costed1481 holdout-1.8556%,N2,p0,averagewinundefined,averageloss1.8556%,stress-2.2474%. It avoided known named decisions but failedpayoff; no claim all early-exit strategies fail. (4) Frozen T10open/T2close1482 holdout+3.6779%,N2,p.5,W10.0230%,L2.6672%,stress+3.2640%; both crossedearlyapprovals. Counterexample is positive return meritsresearch; it is not evidence of pre-news payoff and CI[-2.6672,10.0230]% spanszero. (5) Actual15min17354 gap43.53% refutes guaranteed stop loss, while CLIengine grossstopatlevel results1490–94 cannot repairexecution. QuarterKelly empiricallate.0917363 operational0; fund/venue/risklimits49.14/212.7475/8.5099 do not repair missingedge. Validity: premises support rejecting these two trading rules and adopting primary-first-news/fullpipeline checks. Soundness: primary dates and price arithmetic verified; N10 selected, N2holdout, missing firstglobaltimestamps, no population hazard, no livebidask model remain explicitgaps. New sources stopped changing structural conclusion; no further parametersearch. Fundamental1402,consequences1403,technical1488,simulation1495,statistical1496,indicators1499 andlessons486–488 recorded. No orders or newbook proposed. Existingmonitors/jobs retained; historicalevents need no newalerts, nothing requiresowncapitalaction before06UTC. Finalrecording phase began~04:14 after~13minresearch; finish~04:18 under43planned/45hard. Sources and all numerical eventrows persisted; localartifacts /tmp/codex-run-MaH0rf. Operational conclusion: retain49.14 cash.

Start state: Catalyst cash49.14, equity850.99, no open/pending lots in supplied context and queried own ledger. Goal: costed run-up gains with exits before actual catalyst news. Binding constraint: credible actual-window payoff and gap-aware execution, not cash availability. Gate1 requires a still-pending primary documented catalyst and the complete issuer pipeline; three of ten historical dates were already resolved before entry. Gate2 requires the actual intended window: early T20/T6 lost1.8556% holdout, late T10/T2 gained3.6779% while crossing both early approvals. Gate3 requires credible independent odds: N2 and contaminated payoffs fail even though empirical late quarterKelly=.0917363. Gate4 tests tape/affordability: TLX and SRRK above20/50SMA, but SRRK57.42 exceeds fund49.14; TLX11.98 affordable yet no qualifying event edge. Gate5 risk: maxvenueposition212.7475, fund49.14, 1%equity8.5099; historical45.86%shock caps scenario notional18.5562, not a loss guarantee. Whole-share afterhours eligibility does not remove gap risk (RARE43.53% first-postbar decline). All gates must pass; zero do. Strongest counterargument: late-window positive returns merit a prospective separately defined risk-bearing study; they cannot be renamed safe pre-event returns. Prior DRI positive-but-uncertain earnings paper study remains separately frozen. Sources1481–82,1488,1495–96 and https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all ; dated source pairs in17167–88.

Premises1 ten paired advance-goal/public-announcement cases chosen from six preexisting fund/watchlist issuers;7 goals in2025,3in2026. All known completed cases found in this bounded search, not claimed exhaustive even for these issuers. Publisher-local lead days[3,-2,38,0,-1,-1,6,4,58,32], median3.5; six positive leads. Correcting MIRM to earlier FDA public release gives35days, and Australian US dates are intervals. Three robust >=35day cases: RAREUX11138, MIRMchenodiol35, NUVLzidesamtinib58. Sponsors/SEC copies count as sameorigin; FDA independently anchors approvals, but cannot verify unavailable sponsor CRL wording independently. 2 Named-news-free T20/T6 holdoutN2 mean-1.8556%, versus SPY gross samewindows-3.3660percentagepoints excess; T10/T2 mean+3.6779%, excess+5.7206pp, both trades crossed approval news. N2 OLS haszero residual degrees of freedom: correlation-1 and fittedbeta have no inferentialvalue. TrainN5 late-rule corr.6751/beta2.5682/alpha3.2273% is selected and3TLXclustered; no stablemarketalpha. 3 Ordinary empirical binary-payoff quarterKelly for late-rule p.5,W10.023035%,L2.667224% is.0917363, but it measures contaminated selected payoffs; operational0. Early-rule holdoutno wins, averagewinundefined, no positiveedge. Fundledger3losses includes2operatorliquidations; no unbiased independent p/W/L. Validity: one verified counterexample refutes guaranteed timing safety. Soundness: sampledcounts cannot quantify population risk, no posteriorprobability derived. Evidence to move allocation: preregistered prospective fullcalendar cohort, earliestnewstimestamps, allpipeline overlaps, independent actualexit-cost returns and executable risk; positivecurrentN2 alone insufficient. Sources pairedobservations17167–88/17285–90; https://www.fda.gov/media/151712/download ; prices https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all .

Premises: actual closed RARE trade lost45.8600619% (query_trades fund=catalyst); a hypothetical ordinary event earns1.5% net. The 1.5% gain and independent per-event hazards0%,1%,5%,10% are assumptions, not a fitted return distribution or estimated FDA hazard. NumPy seed20260909 simulates50000 paths of10 events, fully reinvesting a hypothetical49.14. At5% assumed hazard, expected arithmetic return/event=-0.8680031%; terminal p5/median/p95=$16.2255/$57.0290/$57.0290, terminal loss frequency40.296%, p95 maximum drawdown69.8027%. Across10events expected number of shocks0.5; median profit hides severe left-tail damage. At1% hazard loss frequency9.472%; at10%64.868%. Break-even arithmetic hazard=.015/(.015+.458600619)=3.16723%; this threshold is sensitivity, not measured likelihood. Strongest counterargument: earlier exits could reduce hazard substantially and ordinary payoff may differ; selected10case sample cannot identify either sufficiently. Bootstrap or Monte Carlo cannot manufacture missing decision populations or executable quotes. Account riskbudget8.5099 divided by historical shock fraction yields notional18.5562, a scenario cap only, never a hard bound on future gaps. Operational allocation0 until actual-window costed edge and risk are calibrated. Source runway:query_trades?fund=catalyst and https://ir.ultragenyx.com/news-releases/news-release-details/ultragenyx-announces-phase-3-aspire-results-angelman-syndrome ; observations17355–58.

Diagnostic only. Ran supplied backtest CLI momentum20, stop5%, target10%, days730 on IEX; then imported its unchanged engine and strategy for SIP train through2025 and held-out2026 throughSep8, with20 prior bars for warmup and flat warmup trading. Holdout {"fund":"catalyst","split":"holdout","symbol":"SRRK","bars":191,"returnPct":-28.08145949402554,"buyAndHoldPct":31.96966214663297,"maxDrawdownPct":34.1333589954249,"winRate":16.666666666666664,"avgTradePct":-1.800182191174515,"N":18}. Inspection of /app/agent/dist/backtest/engine.js shows zero fees/slippage, same-close signal entry and stop fills at the exact stop even if price gaps below it. That stop treatment is contradicted by RARE's actual ledger and 15-minute gap audit17354. Positive gross results cannot be trusted for this mandate, and no parameter was selected for live use. Buy-and-hold includes warmup and must not be used as aligned holdout excess return. Source https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all ; raw and split observations17369–78/17387–91. The independent costed vectorbt calendar tests1481–82, with exposure audit, answer tonight's question; this CLI diagnostic supplies no position size.

Diagnostic only. Ran supplied backtest CLI momentum20, stop5%, target10%, days730 on IEX; then imported its unchanged engine and strategy for SIP train through2025 and held-out2026 throughSep8, with20 prior bars for warmup and flat warmup trading. Holdout {"fund":"catalyst","split":"holdout","symbol":"TLX","bars":191,"returnPct":14.648436875678472,"buyAndHoldPct":24.01656314699794,"maxDrawdownPct":18.637582304526767,"winRate":40,"avgTradePct":0.8371212121212129,"N":15}. Inspection of /app/agent/dist/backtest/engine.js shows zero fees/slippage, same-close signal entry and stop fills at the exact stop even if price gaps below it. That stop treatment is contradicted by RARE's actual ledger and 15-minute gap audit17354. Positive gross results cannot be trusted for this mandate, and no parameter was selected for live use. Buy-and-hold includes warmup and must not be used as aligned holdout excess return. Source https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all ; raw and split observations17369–78/17387–91. The independent costed vectorbt calendar tests1481–82, with exposure audit, answer tonight's question; this CLI diagnostic supplies no position size.