Stats

Premises:1.Freshledger has3 closed trades, all operator liquidations, and0open/pending/closing. They censor intended4–6week outcomes, so cannot estimate intended stop/target probabilities or execution slippage.2.Saved2255/2208/2209 costed gap holdoutN2151 return-26.212876%,DD27.351241%,winrate43.979544398%,72monthlyclusters meanCI[-.165928,-.052452]%,stressedreturn-56.878654%. Reused evidence, not a rerun/newsample.3.Python recomputation p=.43979544398,W=.617581664%,L=.682768517% yields-.1108804318176%mean; halfKelly=max(0,.5*(p-(1-p)/(W/L)))=0. Conclusion no entry under currenthalfKelly even though oldernotes usedquarterKelly. Validity correct conditional on imported outcomes; soundness limited by vendorhistory/cost assumptions and selection. Counterexample sustainednewregime might reverseexpectancy, but intradayrally/onepaperwin cannot independently demonstrate it. No newparameter sweep justified in tradingreview. Source obs22873; https://runway.adityaperswal.workers.dev/data/stock-bars ; saved2255.

Premises:1. Re-read2183-85: supplied hourly holdouts BTC-14.9605%,ETH-15.5393%,SOL-6.6888%,daily-8.6832/-9.5346/-3.5641%; all36 hourly variants lose. 2025training and REUSED2026Jan1-Sep9holdout,25bpfees/side10bpslippage20stress. 2. Python recomputed pW-(1-p)L and requested halfKelly .5*(p-(1-p)/(W/L)), observation22873 (verify returned ID). 3. One forced liquidation cannot calibrate setup probability. Conclusion zeroallocation; audit not new experiment. Validity holds under positive-expectancy requirement; soundness limited by smallsample/reusedholdout/history. Counterexample another strategy can profit; price rally alone not new evidence for this rule. Frozen1.5R target/old+1Rexit discussion does not satisfy current>=3R/+2Rquarter policy. New compliant strategy requires independent daily/hourly costed15minexecution test; do not simply triple historicaltargets or retune reusedwindow. Sources https://runway.adityaperswal.workers.dev/data/crypto-bars and stored2183-2185.

This is a review of recorded tests, not a new backtest or an independent holdout. Premise 1: EPM backtest1928 had all12 training cells negative; preferred training -6.5511%, six-event holdout +3.8230%, win probability4/6, average win2.6432%, loss3.2626%, drawdown3.9565%, bootstrap mean CI[-1.77895,3.06512]%. Raw quarter-Kelly .0638059 is descriptive, approved allocation0. Premise2: LOVE1677 stressed holdout N10, p20%, W10.7024%/L2.7538%, return-2.34135%, DD12.9858%, CI crosses0. Premise3: TLX2167 generic momentum training-25.273%, holdoutN19+11.0352%, p9/19,W6.5854%/L4.4908%,DD23.9874%, CI[-2.0924,3.5902]%; not FDA-event validation. Premise4: current trade query contains3 losses,2 operator liquidations; this tiny censored sample cannot estimate new-event win odds or routine slippage. Conclusion: EPM's partial-day paper gain supplies no completed additional outcome; retain failed gates without retuning. A favorable new regime is the strongest countercase, but must be tested prospectively with costs and event timing. Sources runway:query_research?fund=catalyst (1928,1677,2167,2247); runway:query_trades?fund=catalyst ; https://runway.adityaperswal.workers.dev/data/stock-bars .

Premises:1. PrimarySep10 weeklyCSV hasSep4 andAug28 columns https://ir.eia.gov/wpsr/table7.csv ;2. Pythonobs22864 rate*7 givescrude23.919m bbl(-7.462m),products59.157(+7.266),total83.076(-.196); WoW-23.7787/+14.0024/-.23537%. Roundedestimates, pipeline/NGL inclusion, no vessel/distanceidentification. 3. JulyCensus https://www.census.gov/wholesale/pdf/mwts/currentwhl.pdf electricalstock+2.216bn vsmonthly sales+.100bn; ratio.89813 vs.87904. Survey~4200,54.8%response; manufacturerbranches andbrokers excluded, missingresponsesweighted/imputed. Conclusion newmix andnominalstockintensity, not provenrent. Validity: arithmetic coherent; soundness boundedrounding/sampling/seasonality andnominalvaluation. Counterexamples: productgrowthcanbenefitgascarriers ratherthanSTNG; electricalstockcanbe deliberatecustomerbuffer ratherthanunsoldgoods. Oneobservation cannotcalibrate stockreturns; no correlation/regressionorcausalprobability claimed. Cashconversion and independentcostedpriceedge remainbinding.

Premises:1.Freshquantledger0open/pending/closing,3closedoperatorliquidations; Pythonexactexit/entry return arithmetic matches prior sample. Interventions truncate4–6weektheses; cannot infer intendedstophit/targethit/slippage from these outcomes.2.Saved2222 p=.43979544398,W=.617581664%,L=.682768517% gives pW-(1-p)L=-.1108804318%/trade;max(0,.25*(p-(1-p)/(W/L)))=0.3.Carry2208/2209OOSN2151return-26.212876%,DD27.351241%,stress-56.878654%,72monthlyclusters/meanCI[-.165928,-.052452]%. These are saved tests, no new backtest or independent sample. Latestneighbor2216recent+.03478% disappears atstress2217-2.18831%;alreadyinspectedneighborcannotrescuefixedrule. Conclusionzero sizebeforefundcap187.22/positioncap212.7475/riskcap8.5099. Assumedcostsnotmeasuredcurrentquotes; absentpositiveedge quoteswonotchangerejection. Source query_trades(fund=quant), obs22847; https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all&timeframe=15Min ; https://personal.lse.ac.uk/polk/research/TugOfWar.pdf .

Premises:1. Re-read backtests2183–85,2025training/2026Jan1–Sep9 REUSEDholdout,25bpfees/side,10bpslippage20stress,nextopen. All36hourlyholdout variants negative; supplieddaily negative too. 2. Python independently recomputed pW-(1-p)L and .25*(p-(1-p)/(W/L)); observation22838 exact figures, all negative and clamped0. 3. Own BTC10 sole trade was operatorliquidation; insufficient setup win/loss or slippage estimation. Conclusion retain disabled compression; this is audit, not new backtest. Validity: failed required positiveedge logically giveszeroallocation; soundness limited by smallsamples/reusedholdout/quote-basedhistory rather than globaltrades. Counterexample independentnewrule couldwork; current data doesnotestablish it. Require unseen positive costed daily/hourly and15min/+1Rexit evidence before size changes; macro shifts affect size by0 until tested. Source https://runway.adityaperswal.workers.dev/data/crypto-bars ; backtests2183–85.

Reused frozenCLI/vectorbt backtests,notnewtests. Premise1 EPM1928all12trainingcellsnegative;preferredtrain-6.551%,holdout6+3.823%,p4/6,W2.6432%,L3.2626%,sampledDD3.9565%,bootstrapCI[-1.77895,3.06512]%. RawquarterKelly.063806descriptiveonly.2 LOVE1677stressN10p20%,W10.7024/L2.7538%,return-2.34135%,DD12.9858%,CIcross0 andeventalreadyreleased.3 TLX2167training-25.273%,holdout19+11.0352%,p9/19,W6.5854/L4.4908%,DD23.9874%,CI[-2.0924,3.5902]%,genericmomentumnotFDAcohort.4 Ownquerytrades3losses,2operatorliquidationscensored; cannot equatep0tofutureodds. Conclusion approvedquarterKelly0; oneprospectiveEPMentrynotfinishedoutcomeorindependentvalidation. Countercase regimechangecouldyieldpositiveedge, requiresnewprospectiveevidence,notholdoutretune. Sources runway:query_research?fund=catalyst (1928,1940,2234) ; runway:query_trades?fund=catalyst ; https://runway.adityaperswal.workers.dev/data/stock-bars .

1.Freshfundledger has0open/pending/closing and3closedoperatorliquidations. Pythonexactexit/entry arithmetic givesp1/3,W7.93994%,L5.26143%,mean-.860974%;notunbiasedcalibrationofintended4–6weekPEADexits. Arrivalquotes/orderbook, naturalstop/targetoutcomes absent: cannotestimate slippage or planned exit odds. 2.Recomputed saved2222 p=.43979544398,W=.617581664%,L=.682768517% yields-.1108804318%pertrade, quarterKelly=max(0,.25*(p-(1-p)/(W/L)))=0. Carry2208/2209 holdoutN2151,return-26.212876%,DD27.351241%,stress-56.878654%,72monthlyclusters,meanCI[-.165928,-.052452]%. No new backtest or independent sample. 3.Saved2221 five-session-block20,000paths/15-session horizon remainsconditional historical simulation; no rerun or revisedexitoddsclaimed. Zero allocation precedes187.22fundcap,212.7475positioncap and8.5099riskcap. New primary headlines cannot alter these payoff estimates without independent validation. Source query_trades(fund=quant); https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all&timeframe=15Min . Obs22816.

Premises:1. Read frozen vectorbt backtests2183–85:2025train,2026Jan1–Sep9reusedholdout; received-asset fees25bp/side,slip10bp base20stress; daily/hourly nextopen simulation. All36 hourly diagnostics lose; daily supplied BTC-8.6832%,ETH-9.5346%,SOL-3.5641%. 2. Python recalculates hourly pW-(1-p)L:BTC-1.273793%,ETH-1.771007%,SOL-.737221%,N12/9/8. QuarterKellyraw-.117421/-.176456/-.047961,clamped0. 3. OwnBTCtrade10 operatorliquidation -.0809902% over47.9278min is not independent strategy calibration. Conclusion: keep compression disabled; no new test or refit this tradingrun. Validity: negative estimated expectancy warrants no long allocation under mandate. Soundness: small samples and reconstructed daily differences could conceal real edge, but cannot establish positive independent executable edge; original confidence intervals cross0 and histories/15min/+1R remain unresolved. Data sampled from venue quote bars, not global executedmarket; missing history and repeatedholdout selection preclude independent inference. To change size require unseen positive costed daily/hourly results and venueexit validation; a news headline changes size by0. Source https://runway.adityaperswal.workers.dev/data/crypto-bars ; prior analyses2183–86; observation22825.

Trading run reuses prior tests, not a new experiment. Premise1 TLX2167 stressholdoutN19,p9/19,W6.5854%,L4.4908%,return11.0352%,DD23.9874%,CI[-2.0924,3.5902]%; training-25.273%,genericmomentum notFDAcohort. Premise2 LOVE1677stressN10,p20%,W10.7024%,L2.7538%,return-2.34135%,DD12.9858%,CI[-3.3193,4.0585]%; actualrelease endsrunup. Premise3 EPM1928N6,p66.6667%,W2.64321%,L3.26259%,holdout+3.823%,training-6.551%,CI[-1.77895,3.06512]%; all12traincellsnegative. Ownrecord3losses,2operatorcensored fromquery_trades, notfuturep=0. Conclusion approvedquarterKelly0; unknown independent liveedge, no averaging incompatiblecohorts. Strongcountercase changedregimesandpositivehistoricalcells possible; needsprospective evidence notreuseofalreadyobservedholdout. CLI/vectorbt/costs/windowrecords retained1677/1688/1928/1940/2167; no rerun or new backtest claim. Source https://runway.adityaperswal.workers.dev/data/stock-bars and runway:query_research?fund=catalyst .

1. Fresh query_trades fundquant yields3closedalloperatorliquidations. Python exit/entry-1 givesSNOW-5.39236448%,DELL7.93994131%,GTLB-5.13049905%;p1/3,W7.93994%,L5.26143%,mean-.860974%,clippedquarterKelly0. Censored intended4–6week exits andN3 cannot calibrate PEAD,slippage,stophitfrequency or merger completion. 2. Recompute saved2222 expectancy .43979544398*.617581664-(1-.43979544398)*.682768517=-.1108804318%pertrade. Saved2151holdout return-26.212876%,DD27.351241%,stress-56.878654%;72monthlyclusters,meanCI[-.165928,-.052452]%. This is arithmetic validation of journal figures,not new backtest or independent data. 3. Prior2211/2221 simulations remain historical conditional resamples,notnewlycalibrated future odds; no rerun needed for unchanged gate. AVAVN9CIcross0 remainsinconclusive; observed premarket news cannot supply pendingSep10close EAR. Allocation0 beforecashcap187.22,positioncap212.7475 orrisk8.5099 bind. Sources query_trades(fund=quant),saved2208/2209/2218/2221/2222; https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all&timeframe=15Min .

1. The fixed 2020–25 holdout has 2,151 trades across 887 signal dates and 72 monthly clusters. Mean net return is −0.110880% with 95% month-cluster bootstrap CI [−0.165928%, −0.052452%]; stressed mean −0.310459%, CI [−0.365700%, −0.251403%]. Resampling 20,000 times does not create new observations. 2. Pooled OLS on 11,968 ETF-days, with symbol effects, lagged RV20 and SMA200 status, gives gap coefficient −0.009604 percentage points per ADR, day-cluster SE 0.024997, t=−0.384, R²=0.005536. This is association, not evidence of causal order flow. 3. Strong counterargument: gross expectancy is slightly positive. Removing the constant assumed cost factor transforms the mean CI to [−0.046055%, +0.067557%], still spanning zero. All six portfolio holdout cells lose; QQQ's one-ADR neighbor gains 5.281% at base cost but loses 4.928% stressed (N=51). Conclusion: reject this adaptation, not all overnight/intraday anomalies. Historical eras were used in earlier fund research; current ETF survivors and assumed costs limit external validity. Source: https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all&timeframe=15Min .

Premises: 1. On 64 weekdays in the last90days, 08UTC-open to12UTC-open gross mean returns were BTC+.1194%,ETH+.1451%,SOL+.0684%; gross mean bootstrap95% intervals [-.0653,.3068],[-.0910,.3761],[-.2033,.3634]% all span zero. 2. With25bp received-asset fees/side and10bp slippage, means become -.5796/-.5541/-.6302%. 3. Last60completedUTCdays provide1439 alignedhourlylogreturns: BTC-ETH correlation.8470,BTC-SOL.7557,ETH-SOL.7590; ETH/SOL regressions onBTC give betas1.1536/1.1072. Conclusion: pre-open activity alone does not justify entry; three sleeves share substantial market risk. Sampling: fixedmajors, recentregime, missing other venues and events; iidCI/naiveOLSSE are descriptive, not serial-dependence-adjusted inference. This is a distribution diagnostic, not a traded time-of-day strategy. Ownrecord N1 operator-liquidatedBTC cannot calibrate setup probabilities. Source https://runway.adityaperswal.workers.dev/data/crypto-bars ; stats.py; observations21555–61.

1. SIPdaily2025Jan1–Sep9 422barseach;source https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all . 2. Last252simple dailyreturns obtained via exp(logreturn)-1: TSM5/50/95percentiles -4.1321/.2943/4.6872%; FTK-6.5474/.0803/8.9716%. Lag1logreturncorrelations -.15913/-.06152. Current20sessionreturn percentile among402overlappingwindows TSM45.27,FTK1.99. 3. Descriptive historicalrank and weakautocorrelation are not an eventstrategy: overlappingwindows, selectednames, regimechange and no matchednewscontrols. Conclusion FTKaboveSMA50 doesnoterase left-tail20sessionstate; tails make daily/reference-stop fills uncertain. No confidence level for positive nextreturn inferred; approved0. Distribution data feeds riskdiagnostics2180/2181; simulations reuse248blocks and are notnewindependentobservations.

P1 query_trades(fund=quant)08:02UTCreturned3closedonly;SNOW373.046->352.93,DELL474.336->511.998,GTLB51.418->48.78. Returns-5.39236448,+7.93994131,-5.13049905%;win1/3,avgwin7.93994131%,avgloss5.26143176%,mean-.86097407%. P2 Alloperatorliquidationsbefore4–6weekhorizons,notstrategyexits;entryarrivalquotes/feebreakdownunavailable,soexecution slippageandplannedstop/targetfrequencyunmeasured. P3 Currentcontextnoopen/queuedlots,187.22fundcash;prior06auditagrees. ConclusionpointquarterKellyclipped0;N3PEADlotoutcomescannotvalidatenewETFdayruleorjustifyrecoverysize. Original ledger sourcequery_trades(fund=quant);providerreturnmathnotcausalattribution.

Premises1 directquery_trades?fund=catalyst returns3closed,0wins,avgprice loss16.38060%,PnL-31.93726; MIRM/IONS operatorliquidation censorintendedhorizons. RAREactualprice loss45.86006%=6.44035listedstopR; stopnotcashlossbound.2 Noarrivalquotes/ordercomparison,so marketversuslimitslippagenotestimable; cannotimputezero.3 TLX60dailyOLSbetaSPY.841656/corr.210142 in454barADSsample; marketassociationneitherFDApayoutnordiversificationguarantee. Default19holdoutmean95%bootstrapCI[-2.09237,3.59020]% crosseszero. Conclusionlivep/win magnitudeunknown,adoptedsize0;0/3isrecorddescriptive,notfuturep=0. Strongcountercaseforcedliquidationsandshortsampleunderstateeventstrategy; thatinvalidatesinferenceofpositiveandnegativeedgealike. Sources runway:query_trades?fund=catalyst ; https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=TLX,SPY&timeframe=1Day&feed=sip&adjustment=all .

1. Live unaudited monthly issuer table https://investor.tsmc.com/english/monthly-revenue/2026 gives July+August982386 NT$m, Jan-Aug3386870. Q2 monthly sum1270381 reconciles rounded quarterly1270.38bn. 2. July16 Q3 guide https://investor.tsmc.com/english/quarterly-results/2026/q2 is USD44.6–45.8bn at32NTD/USD. 3. Python residual = guide*32*1000 - July - August: September444814–483214 NT$m, -13.596% to -6.137% from August. Conclusion August does not mathematically force guidance raise. Conditional currency/recognition bridge, not a forecast; countercase actual FX or September demand changes. Eight monthly observations from one issuer, no independent corroboration or contemporaneous consensus surprise dataset. Do not infer pricing power solely from53.3% YoY. Earlier blank August search snippet versus live table reconciled as stale indexing, not contradictory company releases.

P1 query_tradesfundquant06:20returns3closed,0open/pending;alloperatorliquidations. ReturnsSNOW-5.39236448%,DELL+7.93994131%,GTLB-5.13049905%;win33.333%,avgW7.939941%,avgL5.261432%,mean-.860974%. QuarterKelly0. P2 Holdsendedbefore4–6weekstatedhorizons,notnormalstop/targetoutcomes;cannotestimateplannedexitoddsororderslippagewithoutarrivalquotes. Conclusiondonotborrowthree-tradewinprobabilityforcalendarstrategyanddonotfitliquidationlossesaway. Source query_trades(fund=quant),actualvenueledger;returnsbeforeunobservedfeeattribution.

P1 Fixed8ETF2020–25holdout568tradesare71sharedmonths;meantradecorrelationcalendar.7851/swing.6067. P2 Calendarpooledmean+.314209%baseCI[-.175992,.778329]%,stress+.113780%CI[-.369358,.576591]%;matchedsessions10–13advantage+.607433%CI[-.082291,1.273104]%. P3 Swingmean-.064592%baseCI[-.368358,.234912]%;stress-.264263%CI[-.568733,.034432]%. 20000circular3monthblockresamplesseed91006,preserveETFdependence. P4 HAC5lagdailycloseTOMregressions1508sessions:SPYincrement+.004056%t.05297;EEM+.119247%t1.4421;all8CIsinclude0. StrongestcounterexampleEEMunmanagedbase+.638174%meanCI[.058044,1.252755]% butstressCI[-.144644,1.043289]%;unmanagedexitsnotvenueplan. Conclusion chronologicalholdoutdoesnotestablishrobusttradablecalendaralpha;historicaleraalreadyreusedacrossresearch. No refit/promoteprofitablecells. Sources https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all ; https://business.purdue.edu/faculty/mcconnell/publications/Equity-Returns-at-the-Turn-of-the-Month.pdf .

Premises1 DARDEN06stressedgreenexecution6eventmean.76280%,CI[-1.49818,2.52709]%;4/6winsWilson[.29999,.90323]. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=DRI,SPY&feed=sip&adjustment=all .2 OLSnetDRIreturnongrosssamewindowSPY:train12alpha.251474%,beta.190686,corr.202725;test6alpha-.278871%,beta-1.678656,corr-.902852. Testmeanexcess+1.38334%doesnotestablishstablealpha; signreversal,smallNandpost-selectionknownholdoutremain. BenchmarkholdsfullT5->T1closewhileDRIexitsintraday,soexposuredurationsdiffer;notcausalfactorattribution.3 Ownquery_tradesfundcatalyst ledgerthreeclosed0wins,meanloss16.380601%,realized-31.937258USD;twooperatorliquidationscensorintendedhorizons. RAREloss45.860062%=6.440348listed-stopR. Noobservedaveragewinner,actualentryquoteshistoryabsent,soexecution'sspread/slippagecannotbeseparatedfromgapsandmonitorlatency. ConclusionoperationalKelly0. Validitypositiveexpectedreturnnotidentifiedbythreeheterogeneouslossesorsixselectedpaperreturns. Soundnesslimits sampling/censoring andregimes;0/3isnotfuturep=0. Countercasefirstrunlossesdonotdisprovenewstrategy; genuinelynewprospectivedatawouldchangeconfidence. Neverpooloperatorliquidationswithplannedstop/targetoutcomesasiftheywereidenticaltrials.

Premise1: fixed OOS exits, 87 trades/18 entry-month clusters, same-sector co-movement. Source https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all . Premise2: seed91006 10000 cluster bootstraps resample entire months with all constituent trades, preserving within-month overlap/correlation; 95% meanCI[-.87379,1.91693]%, 26.25% bootstrap means<=0. Premise3: pooled p=.528736,avgwin3.54074%,avgloss3.02751%,mean .445356% on investedcash denominator; limited clusters and cross-month serial dependence remain. Conclusion no defensible positive lower-bound edge, sizing0. OLS descriptive422daily returns equalrefinerbasket onXLE/SPY beta1.11797/.13699,intercept12.9571bp/day,R2.65442; paircorrelations.8565-.8852. Intercept reflects selected historical regime, not forecast alpha; no t-stat inference or causal shortage identification claimed. Counterexample prolonged true structuralmarginshift could make historical training irrelevant, but current mechanism does not calibrate new stock probability. Missing delisted peers and dated historical netbacks; no repeated sampling of the holdout to repair significance.

Premises:1.Common2502026dailyreturnsfrom24houraggregatesproducecorrelationsBTCETH.91395,BTCSOL.88286,ETHSOL.88250.2.OLSagainstBTCgivesETHbeta1.24655,R2.83530;SOLbeta1.20728,R2.77945.3.AnnualizedsamplevolBTC46.72%,ETH63.73%,SOL63.89%. Conclusion:dividingcashacrossthreesymbolsisnotthreeindependentedges;bootstrapentrymonthsjointlyandbudgetcombinedrisk. Validity:common-datesremoveavailabilitydifferencesinthisdescriptivecomparison. Soundness:short2026sample,regime-dependentcorrelation,quote-basedcloseandOLSpointestimates;notcausationorforwardbetaguarantee. InterceptETH-.02128%/day,SOL-.03004%/dayarein-sample,notshortrecommendations. Counterexample:idiosyncraticprotocolshockcanbreakcorrelation;thatstrengthensneedforliquidity/risklimitsratherthanassuminghedge. Source https://runway.adityaperswal.workers.dev/data/crypto-bars .

Premises:1.Dec2repeatfetchgivesdailyopen/hourlyopenBTC91625.132/86270.82,ETH3020.1375/2798.84,SOL139.716/126.67.2.Coinbase00UTCopens86282.35/2798.25/126.65arewithin2.11bpofhourlyquotes,butdaily/hourdifferences620.64/790.68/1029.92bp.3.RecentSep9exactaggregationmatchtestsdifferentperiod. Conclusion:recentagreementcannotclearolderconflicts;vendorrootcauseunresolved. Validity:matchingindependenttimestampisausefulcontrol,notproofofallhistory. Soundness:Coinbasevenue/tradesdifferfromAlpacaquote-midbars,soexactpriceequalitynotexpected;600–1000bp discrepanciesfar exceedobservedopencontrolspreadhere. Originalrule usesdailyprices;reconstructedhistoryandidenticalavailabilitycontrolmustbecomparedseparately. Noorders. Sources https://runway.adityaperswal.workers.dev/data/crypto-bars ; https://api.exchange.coinbase.com/products/BTC-USD/candles ; https://api.exchange.coinbase.com/products/ETH-USD/candles ; https://api.exchange.coinbase.com/products/SOL-USD/candles .

P1 Frozenbase693holdout2020–25events in72entrymonths; longnetreturn-4.5294% vs partner-4.59854%, onidenticaldates andcosts. P2 Meanpairedtrade difference+.00749122%, trade-returncorrelation.9941244. Monthclusterbootstrap10000 CI[-.0263933,.0421721]%; additional3consecutive-monthcircularblockCI[-.0246642,.0397757]%; neither supports positiveconditionaledge. P3 DailyOLS longportfolio onpartnercontrol: beta1.009636, annualizedintercept+.025113%, Newey-West10lagalpha t.12717. SPY-onlyregressionbeta.29198,alpha-5.2025%annual,t-2.2804; conditionalpartner isclosercontrolthanmarket. P4 Swingmean-.133650%,3monthblockCI[-.249193,-.015275]%;stress-.333183%,CI[-.448336,-.216552]%. Conclusion rejectdeployment; largecommonexposureandcosts dominate. SoundnesslimitedbycurrentETFsurvivors, historicalperiodreuse acrossresearch, assignedcosts/adjustmentproxy; this ischronologicalparameterholdout, notgenuinelyunseenprospectiveevidence. Counterexample: someindividualpairs/lowcostneighborcellspositive, but selectingthemnow contaminatesholdout. Need independentprospectivedifferenceCI>0 beforechange. Fullnumdiagnostics archived. Source https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all .

Premises: 1. Reconstructed daily bars require24observedhours; an incomplete day masks OHLC and invalidates subsequent full-window indicators. 2. Unmatched2026 entry-condition disagreements are10BTC/18ETH/7SOL, but conflate price changes with eligibility. 3. An added supplied-daily control uses the identical completeness mask and frozenSMA50, with no parameter selection. Training2025 returns exactly match reconstruction for all three assets; holdout ETH and SOL also match at-3.4189%N14/-18.9768%N8. BTC matched-10.7009%N14 versus reconstructed-9.8920%N14, a .80894percentage-point difference. Conclusion: most apparent aggregation-performance changes come from history availability, not the large flagged price discrepancies; BTC retains some sensitivity. Validity: identical masks isolate construction within this rule. Soundness: this selected rule may not trade flagged dates, so equality does not clear bad input history for other rules or stop/range models. All120cells and payoff CIs recorded19931–70; repeated historical holdout is not fresh validation. Source https://runway.adityaperswal.workers.dev/data/crypto-bars .