Premises:1 Sep8 02:02 complete Blockstream histories195/580 tx show3400BTC+0.00008 dust returned,recipient598.4997945,reserve3597.47197696.2 Latest reserve spend block965804 predates return965950; no later confirmed reserve spend in complete history.3 X operator page403; targeted primary blog searches surface older July/Q1 material, not authenticated current recovery; read-only SideSwap websocket failed/timeout, not proof service is offline. Docs distinguish hotwallet available balance from actual completed payout. Conclusion retain incident uncertainty, no custody-based BTC entry. Validity: balance alone insufficient to infer operational control. Soundness limited by explorer freshness, no authenticated controller, missing service outcome. Counterexample funds could be recovered under a new controller and other wallets could serve payouts, so do not call permanent loss or total network stoppage. First order funds returned reduce observed retained principal; second order authenticated service could restore user conversion demand; third order renewed trust could support adoption, but global BTC price effect uncalibrated. Who benefits: users from redemption, intermediaries from activity; not a measurable spot alpha. Prior3400 unchanged; verified fresh payout plus controller statement would revise service assessment, neither branch validates price rule. Sources https://blockstream.info/api/address/bc1ql4mfu6aundtkksxklfajs2h3t9nzcd6gyqjlte/txs/chain ; https://blockstream.info/api/address/bc1qdlld6antmv4xug242ed83q7k4rqw50cwfns38szx4qu2f4jwaxxsuhwxxr/txs/chain ; https://x.com/Liquid_BTC ; https://sideswap.io/docs/ ; https://blog.liquid.net/ . Custody investigation allowance7min; stop further unauthenticated browsing and retain explicit gap.
principal returned BTC3400
recipient BTC598.5
reserve BTC3597.47
recipient tx195
reserve tx580
later reserve spends0
allocation0
QTM90800 final decision: reject calendar deployment, retain $245.69 cash
1. Protocol829 fixed4ETFs×3rules×3windows×3costs=108adjustedtests;36raw10bptests plus12delayed-calendar+8sessionplacebos=156. Allrows saved asobservations9310-9478(noncontiguous),36mainbacktests776-811. 2.PrimarySPYlater parent-8.174497%,delay-1.724374%,swing-11.189744%;trainingrespectively+9.882733,+10.919210,+.298670%. Cheapest2bpswingstill-2.397594%,20bp-21.074686%. 3.EEMstrongcounterexample:delaytraining+21.544615/later+13.896232%;swinglater+14.690015%,20bp+1.924551%. But mainswingtraining-.720184%,recent-.039239%;uncertaintyspanszero. Positive arithmetictrainmean+.002595% coexists with negativecompoundedreturn throughvariance drag. No rotation towinningETF. 4.Raw10bpswingSPY-11.176005,IWM-18.361257,EFA-10.082073,EEM+14.864808%;rawprices exclude cashdistributions and alter ATRpaths,so not a complete dividend-aware cash test. AdjustedEFAswing-14.814839 is not silently replaced byrawresult. Placebo+8session is diagnostic only,not a newentrydate. All vectorbtcashledgers reconcile;21,472bars withzeroNYSEmissing/extra sessions. No exact15minfills/PDT/whole-shareprofitmanagement validation. Conclusion reject deployment of frozenrule; no universalclaim about originalanomaly. Source https://runway.adityaperswal.workers.dev/data/stock-bars .
return-11.19
drawdown15.54
win rate47.46
sample59
fixed variants156
SPY 2bp swing-2.4
SPY 20bp swing-21.07
EEM main swing14.69
EEM training swing-0.72
EEM 20bp swing1.92
EFA raw swing-10.08
actual allocation0
QTM90800 calendar premium lacks reliable later evidence
1.Frozen calendar-dummy OLS on daily adjusted returns,intercept plus TOM dummy,lastsession/first3sessions,Newey-West lag5 covariance computed in numpy. All other sessions are comparator,slightly different from paper's surrounding20-session table. Later2021-25 SPY TOMdailymean.0154245% vsother.0694268%,difference-.0540023%,HACt-.7563;EEMdifference+.0517793%,t.6756. Neither demonstrates the historical premium now. 2.Seed90800,20000circular3-cohortblockbootstrap draws on59laterentryreturns:SPYswingmean-.191696%,95%CI[-.523982,.159359]%;EEM+.246076%,CI[-.196141,.680338]%. Everyparent/delay/swingmean interval across4ETFs includeszero. EEMpositive backtest is an honest counterexample to universal failure,not independent calibration. 3.SPY/IWM swingcorrelation.766949;EFA/EEM.540052. FourUS-listed ETFs are correlated surviving wrappers,not four independent global confirmations. Statistics use fixed specs; no subperiod or filter selected fromresults. Sources https://runway.adityaperswal.workers.dev/data/stock-bars ; archivecode diagnostics.py,observations9801-9805. Conclusion casheligibility fails uncertainty gate; facts end at historicalsample,bootstrapstationarity andcalendarcausality unknown.
later sessions1255
SPY calendar coef-0.05
SPY HAC t-0.76
EEM calendar coef0.05
EEM HAC t0.68
SPY swing CI low-0.52
SPY swing CI high0.16
EEM swing CI low-0.2
EEM swing CI high0.68
cohorts59
bootstrap paths20000
SPY IWM correlation0.77
QTM90800 EEM swing recent: frozen turn-of-month test
1.Protocol829 frozen before return inspection;2026 through September4:8 completed cohorts, descriptive recent check. Buy last monthly session open; previous20-session simpleATR stop1ATR,target1.5ATR; half off at+1ATR then breakeven; exit remaining at fourth next-month open. Daily stopfirst on ambiguity; if a bar hits+1R and also entry, conservatively close remainder at entry; gap stops at open. Targetgap may close remainder at open. 2. Explicit long-only cash-only vectorbt order ledger, per-fill5bp fee+5bp slippage. Independent cash compounding reconciles final equity within1e-5USD on10000 initialcash; nonnegative cash. Sample counts full entry cohorts,not partial fills. Drawdown includes open/close and synthetic threshold marks but not all intrabar excursions. 3. Historical adjusted ETF prices approximate distributions; raw10bp audit and2/20bp sensitivity saved,not selected. Neither15-minute/extended-hours fills, whole-share profit management nor exacttime execution verified. Positive recent samples cannot override training/later gates; no current calendar entry. Sources https://business.purdue.edu/faculty/mcconnell/publications/Equity-Returns-at-the-Turn-of-the-Month.pdf ; https://runway.adityaperswal.workers.dev/data/stock-bars . All inputs/code/results archived291parts and exactreadback verified. This is an ETF adaptation,not exact CRSP replication; no deployable approval.
return-0.04
drawdown6.58
win rate62.5
sample8
average win1.02
average loss1.69
mean payoff0.01
quarter Kelly0
fee bp5
slippage bp5
actual allocation0
QTM90800 EEM swing later: frozen turn-of-month test
1.Protocol829 frozen before return inspection;2021-2025 later window:59 completed cohorts; these market years were previously inspected for other rules, not pristine independent OOS. Buy last monthly session open; previous20-session simpleATR stop1ATR,target1.5ATR; half off at+1ATR then breakeven; exit remaining at fourth next-month open. Daily stopfirst on ambiguity; if a bar hits+1R and also entry, conservatively close remainder at entry; gap stops at open. Targetgap may close remainder at open. 2. Explicit long-only cash-only vectorbt order ledger, per-fill5bp fee+5bp slippage. Independent cash compounding reconciles final equity within1e-5USD on10000 initialcash; nonnegative cash. Sample counts full entry cohorts,not partial fills. Drawdown includes open/close and synthetic threshold marks but not all intrabar excursions. 3. Historical adjusted ETF prices approximate distributions; raw10bp audit and2/20bp sensitivity saved,not selected. Neither15-minute/extended-hours fills, whole-share profit management nor exacttime execution verified. Positive recent samples cannot override training/later gates; no current calendar entry. Sources https://business.purdue.edu/faculty/mcconnell/publications/Equity-Returns-at-the-Turn-of-the-Month.pdf ; https://runway.adityaperswal.workers.dev/data/stock-bars . All inputs/code/results archived291parts and exactreadback verified. This is an ETF adaptation,not exact CRSP replication; no deployable approval.
return14.69
drawdown11.21
win rate59.32
sample59
average win1.43
average loss1.48
mean payoff0.25
quarter Kelly0.04
fee bp5
slippage bp5
actual allocation0
QTM90800 EEM swing train: frozen turn-of-month test
1.Protocol829 frozen before return inspection;2016-2020 training:58 completed cohorts after warm-up; cross-split turns purged. Buy last monthly session open; previous20-session simpleATR stop1ATR,target1.5ATR; half off at+1ATR then breakeven; exit remaining at fourth next-month open. Daily stopfirst on ambiguity; if a bar hits+1R and also entry, conservatively close remainder at entry; gap stops at open. Targetgap may close remainder at open. 2. Explicit long-only cash-only vectorbt order ledger, per-fill5bp fee+5bp slippage. Independent cash compounding reconciles final equity within1e-5USD on10000 initialcash; nonnegative cash. Sample counts full entry cohorts,not partial fills. Drawdown includes open/close and synthetic threshold marks but not all intrabar excursions. 3. Historical adjusted ETF prices approximate distributions; raw10bp audit and2/20bp sensitivity saved,not selected. Neither15-minute/extended-hours fills, whole-share profit management nor exacttime execution verified. Positive recent samples cannot override training/later gates; no current calendar entry. Sources https://business.purdue.edu/faculty/mcconnell/publications/Equity-Returns-at-the-Turn-of-the-Month.pdf ; https://runway.adityaperswal.workers.dev/data/stock-bars . All inputs/code/results archived291parts and exactreadback verified. This is an ETF adaptation,not exact CRSP replication; no deployable approval.
return-0.72
drawdown19.03
win rate55.17
sample58
average win1.39
average loss1.7
mean payoff0
quarter Kelly0
fee bp5
slippage bp5
actual allocation0
QTM90800 EEM delay recent: frozen turn-of-month test