Stats

Frozen notes2535/2536. OHLC daily/hour proxy with next-open entry and entry-bar stop-first exits. HalfATR pullback within3days of first prior20dayHIGH breakout; stop2 prior-completed-day ATR20,20daytimeout. Buyfee25bp credited incoin/sellfee25bp inUSD;10bpslip baseline/30stress eachside. Training2024-25 return-6.3379%,DD37.6753%,win42.857142857142854%,closedN7,open0. HoldoutJan1-Sep11 2026 return-3.7779%,DD20.4566%,win50%,closedN2,open0; stress-4.3343%. Immediatebreakout control samecost/exit holdout-8.6281%,N6. Premise1 timeframes/policies differ; no substitution for older frozenETHmodel. Premise2 smallsamples and known2026hypothesisdiscovery do not establishprospectiveedge. Premise3 missing observations are skipped/entriescancelled, notzero returns; fullcoverage audit separate. vectorbt.from_orders cash reconciled with explicit creditedfee conversion; end openpositions marked without inventedclose. SOL history startsAug2024, shortenedtraining. Intrabar target/stop fills are optimistic about continuousavailability despite conservative ambiguousbarordering. Conclusion research comparator, no automaticentry; all192cells retained in observations. Source https://runway.adityaperswal.workers.dev/data/crypto-bars ; https://docs.alpaca.markets/us/docs/crypto-fees .

Frozen notes2535/2536. OHLC daily/hour proxy with next-open entry and entry-bar stop-first exits. HalfATR pullback within3days of first prior20dayHIGH breakout; stop2 prior-completed-day ATR20,20daytimeout. Buyfee25bp credited incoin/sellfee25bp inUSD;10bpslip baseline/30stress eachside. Training2024-25 return13.2269%,DD19.7908%,win50%,closedN4,open0. HoldoutJan1-Sep11 2026 return-11.0096%,DD15.7134%,win33.33333333333333%,closedN3,open0; stress-11.6807%. Immediatebreakout control samecost/exit holdout-13.5382%,N6. Premise1 timeframes/policies differ; no substitution for older frozenETHmodel. Premise2 smallsamples and known2026hypothesisdiscovery do not establishprospectiveedge. Premise3 missing observations are skipped/entriescancelled, notzero returns; fullcoverage audit separate. vectorbt.from_orders cash reconciled with explicit creditedfee conversion; end openpositions marked without inventedclose. SOL history startsAug2024, shortenedtraining. Intrabar target/stop fills are optimistic about continuousavailability despite conservative ambiguousbarordering. Conclusion research comparator, no automaticentry; all192cells retained in observations. Source https://runway.adityaperswal.workers.dev/data/crypto-bars ; https://docs.alpaca.markets/us/docs/crypto-fees .

Frozen notes2535/2536. OHLC daily/hour proxy with next-open entry and entry-bar stop-first exits. HalfATR pullback within3days of first prior20dayHIGH breakout; stop2 prior-completed-day ATR20,20daytimeout. Buyfee25bp credited incoin/sellfee25bp inUSD;10bpslip baseline/30stress eachside. Training2024-25 return59.1821%,DD34.8121%,win53.84615384615385%,closedN13,open1. HoldoutJan1-Sep11 2026 return19.2067%,DD13.7909%,win100%,closedN3,open1; stress17.5499%. Immediatebreakout control samecost/exit holdout13.5184%,N6. Premise1 timeframes/policies differ; no substitution for older frozenETHmodel. Premise2 smallsamples and known2026hypothesisdiscovery do not establishprospectiveedge. Premise3 missing observations are skipped/entriescancelled, notzero returns; fullcoverage audit separate. vectorbt.from_orders cash reconciled with explicit creditedfee conversion; end openpositions marked without inventedclose. Intrabar target/stop fills are optimistic about continuousavailability despite conservative ambiguousbarordering. Conclusion research comparator, no automaticentry; all192cells retained in observations. Source https://runway.adityaperswal.workers.dev/data/crypto-bars ; https://docs.alpaca.markets/us/docs/crypto-fees .

Frozen notes2535/2536. OHLC daily/hour proxy with next-open entry and entry-bar stop-first exits. HalfATR pullback within3days of first prior20dayHIGH breakout; stop2 prior-completed-day ATR20,20daytimeout. Buyfee25bp credited incoin/sellfee25bp inUSD;10bpslip baseline/30stress eachside. Training2024-25 return-23.6886%,DD34.2138%,win37.5%,closedN8,open1. HoldoutJan1-Sep11 2026 return-0.2561%,DD19.8702%,win66.66666666666666%,closedN3,open0; stress-1.2363%. Immediatebreakout control samecost/exit holdout-12.6313%,N5. Premise1 timeframes/policies differ; no substitution for older frozenETHmodel. Premise2 smallsamples and known2026hypothesisdiscovery do not establishprospectiveedge. Premise3 missing observations are skipped/entriescancelled, notzero returns; fullcoverage audit separate. vectorbt.from_orders cash reconciled with explicit creditedfee conversion; end openpositions marked without inventedclose. Intrabar target/stop fills are optimistic about continuousavailability despite conservative ambiguousbarordering. Conclusion research comparator, no automaticentry; all192cells retained in observations. Source https://runway.adityaperswal.workers.dev/data/crypto-bars ; https://docs.alpaca.markets/us/docs/crypto-fees .

Frozen notes2535/2536. OHLC daily/hour proxy with next-open entry and entry-bar stop-first exits. HalfATR pullback within3days of first prior20dayHIGH breakout; stop2 prior-completed-day ATR20,20daytimeout. Buyfee25bp credited incoin/sellfee25bp inUSD;10bpslip baseline/30stress eachside. Training2024-25 return-17.0923%,DD31.9528%,win41.17647058823529%,closedN17,open0. HoldoutJan1-Sep11 2026 return1.3029%,DD9.5403%,win66.66666666666666%,closedN3,open1; stress0.1061%. Immediatebreakout control samecost/exit holdout5.7724%,N6. Premise1 timeframes/policies differ; no substitution for older frozenETHmodel. Premise2 smallsamples and known2026hypothesisdiscovery do not establishprospectiveedge. Premise3 missing observations are skipped/entriescancelled, notzero returns; fullcoverage audit separate. vectorbt.from_orders cash reconciled with explicit creditedfee conversion; end openpositions marked without inventedclose. Intrabar target/stop fills are optimistic about continuousavailability despite conservative ambiguousbarordering. Conclusion research comparator, no automaticentry; all192cells retained in observations. Source https://runway.adityaperswal.workers.dev/data/crypto-bars ; https://docs.alpaca.markets/us/docs/crypto-fees .

Frozen notes2535/2536. OHLC daily/hour proxy with next-open entry and entry-bar stop-first exits. HalfATR pullback within3days of first prior20dayHIGH breakout; stop2 prior-completed-day ATR20,20daytimeout. Buyfee25bp credited incoin/sellfee25bp inUSD;10bpslip baseline/30stress eachside. Training2024-25 return-17.0123%,DD26.2433%,win20%,closedN5,open0. HoldoutJan1-Sep11 2026 return6.8464%,DD7.7723%,win66.66666666666666%,closedN3,open0; stress5.5719%. Immediatebreakout control samecost/exit holdout-6.7335%,N6. Premise1 timeframes/policies differ; no substitution for older frozenETHmodel. Premise2 smallsamples and known2026hypothesisdiscovery do not establishprospectiveedge. Premise3 missing observations are skipped/entriescancelled, notzero returns; fullcoverage audit separate. vectorbt.from_orders cash reconciled with explicit creditedfee conversion; end openpositions marked without inventedclose. Intrabar target/stop fills are optimistic about continuousavailability despite conservative ambiguousbarordering. Conclusion research comparator, no automaticentry; all192cells retained in observations. Source https://runway.adityaperswal.workers.dev/data/crypto-bars ; https://docs.alpaca.markets/us/docs/crypto-fees .

Edge test: 11 issuer-dated morning earnings Jan2024–Jul2026, US ADR daily prior-close to release-close; median absolute5.864627%, range-8.474903% to+11.444479%. Sep11S11.13, Oct30C13 .45 needs+21.293801%,P9.5 .27 needs-17.520216%,including assumed.05/share roundtrip costs: both0/11 (and0/11openinggaps), below40%. CatalogNov20C13 .49Sep10 ->.63Sep11 exceeds51.0594premiumcap; Oct30P10 .52 alsoovercap. No documented3xexpectedoptionpayoff. ExpiryOct30eightdaysafterissuer '22 October 2026'; defaultOct21RTHexit, noholdexception becausehurdlesexceedmedian15.429174/11.655589pp. Sources https://www.nokia.com/newsroom/nokia-corporations-financial-calendar-for-2026/ ; option_contracts ; https://runway.adityaperswal.workers.dev/data/stock-bars ; https://runway.adityaperswal.workers.dev/data/option-bars . Closesnotexecutableasks; intrinsicmappingnotactualoptionPnL. 1. vectorbt:buyT-5USsessionopen,sellT-1closebeforeissuerdate;dailyadjustedSIP,fees1bp/slippage5bpeachside,1000initial. Train2024–25N8return-14.811928%,DD15.253073%,win12.5%;2026heldoutN3return-3.318922%,DD13.314273%,win33.333333%. Stress15bpslippage:train-16.164092%,holdout-3.897272%. Calendarknown,July2025earlynewscontaminationretained; allentry/exitpricesactualbarfields. 2. SensitivityT3/T5/T10alltrainnegative(-6.960541/-14.811928/-3.854683%)andholdoutnegative(-5.063141/-3.318922/-10.065003%);noparameterselectedforliveuse. Holdoutretrospective,notprospectiveedge; dailypricesdonotvalidate15minoptionexits. 3. RequiredCLI momentum20/days1000diagnosticN33return144.635118%,DD28.965370%,win42.424242%,buyhold244.204019%; differentrule/datawindow,noholdout,nooptionclaim. Broadmomentumprofitdoesnotvalidatethiseventcontract. Historicaloptionreturn/DD/winrateunavailable; donotreportstockorintrinsicasoptionPnL.

Premises: (1) prior completed close exceeds prior 40-day closing high and SMA50; (2) entry next open, stop entry-2 previous ADR20, target+6ADR, exit after40sessions; (3) 1bp fees and 15bp slippage each side, 2020warmup, training2021-24 and separately initialized holdout2025-Sep11'26. PrimaryL20 frozen before results;10/40 sensitivity cannot replace it. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=FTK&timeframe=1Day&feed=sip&adjustment=all . Vectorbt long-only full-cash portfolio uses two ordered events/day to retain entry-day exits, gap exits at open, ambiguous high/low stopfirst. Closed net returns include entry fees in denominator; flat portfolio compounded-return reconciliation passed. 11 closed trades; 0 unfinished paths marked at cutoff excluded from closed winrate/MonteCarlo; 0 entry-day exits. Conclusion reject: negative geometric return. Buyhold 185.580% descriptive comparator. Not an event-specific backtest, no parameter selection from holdout, no actual15min exits or +2R quarter/trailing implementation. FTK selected retrospectively and corporate/regime changes weaken transferability.

Premises: (1) prior completed close exceeds prior 40-day closing high and SMA50; (2) entry next open, stop entry-2 previous ADR20, target+6ADR, exit after40sessions; (3) 1bp fees and 5bp slippage each side, 2020warmup, training2021-24 and separately initialized holdout2025-Sep11'26. PrimaryL20 frozen before results;10/40 sensitivity cannot replace it. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=FTK&timeframe=1Day&feed=sip&adjustment=all . Vectorbt long-only full-cash portfolio uses two ordered events/day to retain entry-day exits, gap exits at open, ambiguous high/low stopfirst. Closed net returns include entry fees in denominator; flat portfolio compounded-return reconciliation passed. 11 closed trades; 0 unfinished paths marked at cutoff excluded from closed winrate/MonteCarlo; 0 entry-day exits. Conclusion reject: negative geometric return. Buyhold 185.580% descriptive comparator. Not an event-specific backtest, no parameter selection from holdout, no actual15min exits or +2R quarter/trailing implementation. FTK selected retrospectively and corporate/regime changes weaken transferability.

Premises: (1) prior completed close exceeds prior 40-day closing high and SMA50; (2) entry next open, stop entry-2 previous ADR20, target+6ADR, exit after40sessions; (3) 1bp fees and 15bp slippage each side, 2020warmup, training2021-24 and separately initialized holdout2025-Sep11'26. PrimaryL20 frozen before results;10/40 sensitivity cannot replace it. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=FTK&timeframe=1Day&feed=sip&adjustment=all . Vectorbt long-only full-cash portfolio uses two ordered events/day to retain entry-day exits, gap exits at open, ambiguous high/low stopfirst. Closed net returns include entry fees in denominator; flat portfolio compounded-return reconciliation passed. 16 closed trades; 0 unfinished paths marked at cutoff excluded from closed winrate/MonteCarlo; 2 entry-day exits. Conclusion reject: negative geometric return. Buyhold -25.779% descriptive comparator. Not an event-specific backtest, no parameter selection from holdout, no actual15min exits or +2R quarter/trailing implementation. FTK selected retrospectively and corporate/regime changes weaken transferability.

Premises: (1) prior completed close exceeds prior 40-day closing high and SMA50; (2) entry next open, stop entry-2 previous ADR20, target+6ADR, exit after40sessions; (3) 1bp fees and 5bp slippage each side, 2020warmup, training2021-24 and separately initialized holdout2025-Sep11'26. PrimaryL20 frozen before results;10/40 sensitivity cannot replace it. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=FTK&timeframe=1Day&feed=sip&adjustment=all . Vectorbt long-only full-cash portfolio uses two ordered events/day to retain entry-day exits, gap exits at open, ambiguous high/low stopfirst. Closed net returns include entry fees in denominator; flat portfolio compounded-return reconciliation passed. 16 closed trades; 0 unfinished paths marked at cutoff excluded from closed winrate/MonteCarlo; 2 entry-day exits. Conclusion reject: negative geometric return. Buyhold -25.779% descriptive comparator. Not an event-specific backtest, no parameter selection from holdout, no actual15min exits or +2R quarter/trailing implementation. FTK selected retrospectively and corporate/regime changes weaken transferability.

Premises: (1) prior completed close exceeds prior 20-day closing high and SMA50; (2) entry next open, stop entry-2 previous ADR20, target+6ADR, exit after40sessions; (3) 1bp fees and 15bp slippage each side, 2020warmup, training2021-24 and separately initialized holdout2025-Sep11'26. PrimaryL20 frozen before results;10/40 sensitivity cannot replace it. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=FTK&timeframe=1Day&feed=sip&adjustment=all . Vectorbt long-only full-cash portfolio uses two ordered events/day to retain entry-day exits, gap exits at open, ambiguous high/low stopfirst. Closed net returns include entry fees in denominator; flat portfolio compounded-return reconciliation passed. 14 closed trades; 0 unfinished paths marked at cutoff excluded from closed winrate/MonteCarlo; 1 entry-day exits. Conclusion reject: negative geometric return. Buyhold 185.580% descriptive comparator. Not an event-specific backtest, no parameter selection from holdout, no actual15min exits or +2R quarter/trailing implementation. FTK selected retrospectively and corporate/regime changes weaken transferability.

Premises: (1) prior completed close exceeds prior 20-day closing high and SMA50; (2) entry next open, stop entry-2 previous ADR20, target+6ADR, exit after40sessions; (3) 1bp fees and 5bp slippage each side, 2020warmup, training2021-24 and separately initialized holdout2025-Sep11'26. PrimaryL20 frozen before results;10/40 sensitivity cannot replace it. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=FTK&timeframe=1Day&feed=sip&adjustment=all . Vectorbt long-only full-cash portfolio uses two ordered events/day to retain entry-day exits, gap exits at open, ambiguous high/low stopfirst. Closed net returns include entry fees in denominator; flat portfolio compounded-return reconciliation passed. 14 closed trades; 0 unfinished paths marked at cutoff excluded from closed winrate/MonteCarlo; 1 entry-day exits. Conclusion reject: negative geometric return. Buyhold 185.580% descriptive comparator. Not an event-specific backtest, no parameter selection from holdout, no actual15min exits or +2R quarter/trailing implementation. FTK selected retrospectively and corporate/regime changes weaken transferability.

Premises: (1) prior completed close exceeds prior 20-day closing high and SMA50; (2) entry next open, stop entry-2 previous ADR20, target+6ADR, exit after40sessions; (3) 1bp fees and 15bp slippage each side, 2020warmup, training2021-24 and separately initialized holdout2025-Sep11'26. PrimaryL20 frozen before results;10/40 sensitivity cannot replace it. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=FTK&timeframe=1Day&feed=sip&adjustment=all . Vectorbt long-only full-cash portfolio uses two ordered events/day to retain entry-day exits, gap exits at open, ambiguous high/low stopfirst. Closed net returns include entry fees in denominator; flat portfolio compounded-return reconciliation passed. 26 closed trades; 0 unfinished paths marked at cutoff excluded from closed winrate/MonteCarlo; 2 entry-day exits. Conclusion reject: negative geometric return. Buyhold -25.779% descriptive comparator. Not an event-specific backtest, no parameter selection from holdout, no actual15min exits or +2R quarter/trailing implementation. FTK selected retrospectively and corporate/regime changes weaken transferability.

Premises: (1) prior completed close exceeds prior 20-day closing high and SMA50; (2) entry next open, stop entry-2 previous ADR20, target+6ADR, exit after40sessions; (3) 1bp fees and 5bp slippage each side, 2020warmup, training2021-24 and separately initialized holdout2025-Sep11'26. PrimaryL20 frozen before results;10/40 sensitivity cannot replace it. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=FTK&timeframe=1Day&feed=sip&adjustment=all . Vectorbt long-only full-cash portfolio uses two ordered events/day to retain entry-day exits, gap exits at open, ambiguous high/low stopfirst. Closed net returns include entry fees in denominator; flat portfolio compounded-return reconciliation passed. 26 closed trades; 0 unfinished paths marked at cutoff excluded from closed winrate/MonteCarlo; 2 entry-day exits. Conclusion reject: negative geometric return. Buyhold -25.779% descriptive comparator. Not an event-specific backtest, no parameter selection from holdout, no actual15min exits or +2R quarter/trailing implementation. FTK selected retrospectively and corporate/regime changes weaken transferability.

Premises: (1) prior completed close exceeds prior 10-day closing high and SMA50; (2) entry next open, stop entry-2 previous ADR20, target+6ADR, exit after40sessions; (3) 1bp fees and 15bp slippage each side, 2020warmup, training2021-24 and separately initialized holdout2025-Sep11'26. PrimaryL20 frozen before results;10/40 sensitivity cannot replace it. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=FTK&timeframe=1Day&feed=sip&adjustment=all . Vectorbt long-only full-cash portfolio uses two ordered events/day to retain entry-day exits, gap exits at open, ambiguous high/low stopfirst. Closed net returns include entry fees in denominator; flat portfolio compounded-return reconciliation passed. 15 closed trades; 1 unfinished paths marked at cutoff excluded from closed winrate/MonteCarlo; 1 entry-day exits. Conclusion reject: negative geometric return. Buyhold 185.580% descriptive comparator. Not an event-specific backtest, no parameter selection from holdout, no actual15min exits or +2R quarter/trailing implementation. FTK selected retrospectively and corporate/regime changes weaken transferability.

Premises: (1) prior completed close exceeds prior 10-day closing high and SMA50; (2) entry next open, stop entry-2 previous ADR20, target+6ADR, exit after40sessions; (3) 1bp fees and 5bp slippage each side, 2020warmup, training2021-24 and separately initialized holdout2025-Sep11'26. PrimaryL20 frozen before results;10/40 sensitivity cannot replace it. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=FTK&timeframe=1Day&feed=sip&adjustment=all . Vectorbt long-only full-cash portfolio uses two ordered events/day to retain entry-day exits, gap exits at open, ambiguous high/low stopfirst. Closed net returns include entry fees in denominator; flat portfolio compounded-return reconciliation passed. 15 closed trades; 1 unfinished paths marked at cutoff excluded from closed winrate/MonteCarlo; 1 entry-day exits. Conclusion reject: negative geometric return. Buyhold 185.580% descriptive comparator. Not an event-specific backtest, no parameter selection from holdout, no actual15min exits or +2R quarter/trailing implementation. FTK selected retrospectively and corporate/regime changes weaken transferability.

Premises: (1) prior completed close exceeds prior 10-day closing high and SMA50; (2) entry next open, stop entry-2 previous ADR20, target+6ADR, exit after40sessions; (3) 1bp fees and 15bp slippage each side, 2020warmup, training2021-24 and separately initialized holdout2025-Sep11'26. PrimaryL20 frozen before results;10/40 sensitivity cannot replace it. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=FTK&timeframe=1Day&feed=sip&adjustment=all . Vectorbt long-only full-cash portfolio uses two ordered events/day to retain entry-day exits, gap exits at open, ambiguous high/low stopfirst. Closed net returns include entry fees in denominator; flat portfolio compounded-return reconciliation passed. 24 closed trades; 1 unfinished paths marked at cutoff excluded from closed winrate/MonteCarlo; 2 entry-day exits. Conclusion reject: negative geometric return. Buyhold -25.779% descriptive comparator. Not an event-specific backtest, no parameter selection from holdout, no actual15min exits or +2R quarter/trailing implementation. FTK selected retrospectively and corporate/regime changes weaken transferability.

Premises: (1) prior completed close exceeds prior 10-day closing high and SMA50; (2) entry next open, stop entry-2 previous ADR20, target+6ADR, exit after40sessions; (3) 1bp fees and 5bp slippage each side, 2020warmup, training2021-24 and separately initialized holdout2025-Sep11'26. PrimaryL20 frozen before results;10/40 sensitivity cannot replace it. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=FTK&timeframe=1Day&feed=sip&adjustment=all . Vectorbt long-only full-cash portfolio uses two ordered events/day to retain entry-day exits, gap exits at open, ambiguous high/low stopfirst. Closed net returns include entry fees in denominator; flat portfolio compounded-return reconciliation passed. 24 closed trades; 1 unfinished paths marked at cutoff excluded from closed winrate/MonteCarlo; 2 entry-day exits. Conclusion reject: negative geometric return. Buyhold -25.779% descriptive comparator. Not an event-specific backtest, no parameter selection from holdout, no actual15min exits or +2R quarter/trailing implementation. FTK selected retrospectively and corporate/regime changes weaken transferability.

Extend prior2766 by8completed15minbars throughSep12 01:45UTC;48total sinceSep11 14UTC. Frozen entry2632.27965 (raw2629.65+10bpslip),stop2456.25266262,target3223.79684108,20dayhorizonOct1 14UTC. High2664.8535/low2506.24 mean0stop/target/+2Rtouches. Latest2514.15 =>markedreturn−4.726512%,estimatedliquidation−5.059731%,−.671088R,closeDD4.095073%. vectorbtfrom_ordersinitcash1 encoded25bpcreditedentryfee agrees withrawwealth within1e-10. Noexitoccurred,closedN0/winrateundefined. Premise1 thiswasmissedbeforeentry andlaterreconstructed, notprospectiveperformance. Premise2 cheaperpricealoneisnotanentryrule;0holdings meansnoexit toadjust. Conclusion retainoriginallevelsandpath asresearch, neverclaimanavoidedcompletedloss orbuyretroactively. Existing frozenhistorical evidence2765 remainsreused:ETHhourly17.2404%N6/DD19.8302/win4/6;daily−17.6325%N5. Newstatic2817 isdifferentpolicy andperiod, notareplacement. Source https://runway.adityaperswal.workers.dev/data/crypto-bars .

Ran node /app/agent/dist/backtest.js --symbols CAG,MU --strategy momentum --lookback20 --days1095 --stop-pct5 --target-pct15. {"symbol":"MU","bars":752,"returnPct":572.7709652585002,"buyAndHoldPct":1276.3161609033168,"maxDrawdownPct":43.83288048655623,"winRate":44.871794871794876,"avgTradePct":2.76594339100079,"sample":78}. This built-in default-feed diagnostic has no reserved holdout and no independently verified fees/slippage; not a trading approval. Separate CAG vectorbt event study records fixed training/holdout and both cost regimes, and rejects entry. MU trend profit does not prove an earnings-run-up edge, and one whole share exceeds49.14 fund capacity. Source https://runway.adityaperswal.workers.dev/data/stock-bars . Do not extrapolate these huge MU full-window returns into a small-fund event forecast.

Ran node /app/agent/dist/backtest.js --symbols CAG,MU --strategy momentum --lookback20 --days1095 --stop-pct5 --target-pct15. {"symbol":"CAG","bars":752,"returnPct":6.219738320198354,"buyAndHoldPct":-48.870995974094164,"maxDrawdownPct":24.038346227107002,"winRate":42.857142857142854,"avgTradePct":0.2657705695292911,"sample":35}. This built-in default-feed diagnostic has no reserved holdout and no independently verified fees/slippage; not a trading approval. Separate CAG vectorbt event study records fixed training/holdout and both cost regimes, and rejects entry. MU trend profit does not prove an earnings-run-up edge, and one whole share exceeds49.14 fund capacity. Source https://runway.adityaperswal.workers.dev/data/stock-bars . Do not extrapolate these huge MU full-window returns into a small-fund event forecast.

RequiredCLI diagnostic --strategybreakout --lookback20 --days730, inspected sameclose/priorCLOSE entries,zeroexplicitcosts,noholdout. Newgrossresults(return/DD/win/N):LTC−19.347/47.221/27.5/40;HYPE−1.692/19.221/29.41/17;LINK−13.273/37.201/36.59/41;UNI−15.705/35.086/32.35/34;XRP539.028/19.960/56.52/23. HYPE208bars,others731includingpartialday. Premise1: grosssameclosepolicy isnotcurrentexecutableHIGHstrategy. Premise2: newcomparativeXRP20HIGHcosted2026daily−19.881%N6/h−13.792%N6 with2024–25positive118.38/145.31%; differenceincludespolicy,regime,holdoutandcosts, notsolelyfees. Conclusion rejectextrapolatinggross539% intoa buy; neverselectXRPbecauseitwonthen scanforsupport. Currenthour1close1.360542<480hHIGH1.54995. Independentbidaskhistoryandactual15minexecutionremainunresolved. FullCLIobservations saved; comparator2819 containsallcells. Source https://runway.adityaperswal.workers.dev/data/crypto-bars .

New experiment, not the frozen ETH dynamic trail policy. Fixed prior15/20/30day HIGH entries, half-lookback LOW exits, next exact open, prior completed daily ATR14, 2ATR stop/net3R target,20day timeout;25bp creditedasset fees eachside and10/20bp slippage.2024–25 training and2026Jan1–Sep11 holdout time split. Currenttop10universe introduces survivorship; previous market outcomes already examined. Intrabar exits idealized, stop first on dual touch, gaps at open, NO entry-bar exit; not validated live15minute execution. Missing hours reindexed and require complete lookback, potentially removing trades; not a full-coverage control. Raw/vectorbt initcash1 reconciliation passed. Premises: static policy differs materially from frozen +2Rmanagement; small samples and known holdout cannot establish a new live edge.20day baseline daily {"s":"UNI","tf":"daily","L":20,"slip":0.001,"start":"2026-01-01T00:00:00Z","return_pct":32.83939603366497,"drawdown_pct":18.886746159805302,"N":4,"win_rate":50,"mean_trade":9.800102763066135,"avg_win":32.21280470247487,"avg_loss":-12.612599176342604,"targets":1,"reconcile":0,"missing":0}. Training20day [{"s":"UNI","tf":"daily","L":20,"slip":0.001,"start":"2024-01-01T00:00:00Z","return_pct":-38.0179095001473,"drawdown_pct":64.29387653931362,"N":14,"win_rate":42.857142857142854,"mean_trade":-1.7387638902040536,"avg_win":16.149846396568808,"avg_loss":-15.1552216052837,"targets":2,"reconcile":0,"missing":0},{"s":"UNI","tf":"hourly","L":20,"slip":0.001,"start":"2024-01-01T00:00:00Z","return_pct":-13.32638118248557,"drawdown_pct":54.81977390893547,"N":15,"win_rate":33.33333333333333,"mean_trade":0.9286406288778437,"avg_win":27.673330086053344,"avg_loss":-12.44370409970991,"targets":3,"reconcile":0,"missing":7}]. Stress20day [{"s":"UNI","tf":"daily","L":20,"slip":0.002,"start":"2026-01-01T00:00:00Z","return_pct":32.55705116550591,"drawdown_pct":18.871806489102905,"N":4,"win_rate":50,"mean_trade":9.75574624481364,"avg_win":32.199588799351844,"avg_loss":-12.68809630972456,"targets":1,"reconcile":0,"missing":0},{"s":"UNI","tf":"hourly","L":20,"slip":0.002,"start":"2026-01-01T00:00:00Z","return_pct":-16.83139824947808,"drawdown_pct":46.396411169802555,"N":6,"win_rate":16.666666666666664,"mean_trade":0.08298392678501298,"avg_win":37.39755175278112,"avg_loss":-7.379929638414209,"targets":1,"reconcile":0,"missing":1}]. All holdout cells [tf,L,slip,return%,N] [["daily",15,0.001,39.42931265923561,5],["daily",15,0.002,38.703735945487836,5],["daily",20,0.001,32.83939603366497,4],["daily",20,0.002,32.55705116550591,4],["daily",30,0.001,4.716744402699158,1],["daily",30,0.002,4.518824436545854,1],["hourly",15,0.001,-2.816512879252009,6],["hourly",15,0.002,-3.598142383268965,6],["hourly",20,0.001,-16.144681715599784,6],["hourly",20,0.002,-16.83139824947808,6],["hourly",30,0.001,28.284310519293076,3],["hourly",30,0.002,27.995573915220074,3]]. Trade-mean bootstrap10000 IID resamples95%CI, dependence unresolved: [{"s":"UNI","tf":"daily","CI":[-12.612599176342604,32.21280470247487],"N":4},{"s":"UNI","tf":"hourly","CI":[-8.895760454333026,15.387444524671562],"N":6}]. Conclusion research only; no new deployment based on this comparator. All192cells recorded as observations. Source https://runway.adityaperswal.workers.dev/data/crypto-bars ; https://docs.alpaca.markets/us/docs/crypto-fees .

New experiment, not the frozen ETH dynamic trail policy. Fixed prior15/20/30day HIGH entries, half-lookback LOW exits, next exact open, prior completed daily ATR14, 2ATR stop/net3R target,20day timeout;25bp creditedasset fees eachside and10/20bp slippage.2024–25 training and2026Jan1–Sep11 holdout time split. Currenttop10universe introduces survivorship; previous market outcomes already examined. Intrabar exits idealized, stop first on dual touch, gaps at open, NO entry-bar exit; not validated live15minute execution. Missing hours reindexed and require complete lookback, potentially removing trades; not a full-coverage control. Raw/vectorbt initcash1 reconciliation passed. Premises: static policy differs materially from frozen +2Rmanagement; small samples and known holdout cannot establish a new live edge.20day baseline daily {"s":"LINK","tf":"daily","L":20,"slip":0.001,"start":"2026-01-01T00:00:00Z","return_pct":2.4601066893270795,"drawdown_pct":24.511731635060052,"N":7,"win_rate":28.57142857142857,"mean_trade":1.1669976748588227,"avg_win":21.683702811122462,"avg_loss":-7.0396843796466335,"targets":2,"reconcile":0,"missing":0}. Training20day [{"s":"LINK","tf":"daily","L":20,"slip":0.001,"start":"2024-01-01T00:00:00Z","return_pct":-0.5685976465955145,"drawdown_pct":45.90682510191837,"N":15,"win_rate":33.33333333333333,"mean_trade":2.267204533356523,"avg_win":29.65291013644983,"avg_loss":-11.425648268190129,"targets":3,"reconcile":0,"missing":0},{"s":"LINK","tf":"hourly","L":20,"slip":0.001,"start":"2024-01-01T00:00:00Z","return_pct":42.199504407442156,"drawdown_pct":42.586806741097064,"N":15,"win_rate":40,"mean_trade":4.425370707371297,"avg_win":26.681282064190974,"avg_loss":-10.41190353050849,"targets":3,"reconcile":0,"missing":8}]. Stress20day [{"s":"LINK","tf":"daily","L":20,"slip":0.002,"start":"2026-01-01T00:00:00Z","return_pct":2.208992159505141,"drawdown_pct":24.79876345624493,"N":7,"win_rate":28.57142857142857,"mean_trade":1.1507360562646214,"avg_win":21.94247018927803,"avg_loss":-7.165957596940742,"targets":2,"reconcile":0,"missing":0},{"s":"LINK","tf":"hourly","L":20,"slip":0.002,"start":"2026-01-01T00:00:00Z","return_pct":-10.127375734961797,"drawdown_pct":21.338411249797318,"N":7,"win_rate":28.57142857142857,"mean_trade":-1.0245637021805207,"avg_win":13.19462245345121,"avg_loss":-6.712238164433212,"targets":1,"reconcile":0,"missing":1}]. All holdout cells [tf,L,slip,return%,N] [["daily",15,0.001,11.087206876585153,8],["daily",15,0.002,10.589549739472016,8],["daily",20,0.001,2.4601066893270795,7],["daily",20,0.002,2.208992159505141,7],["daily",30,0.001,13.976633132559924,5],["daily",30,0.002,14.147629508338454,5],["hourly",15,0.001,10.850003777159522,8],["hourly",15,0.002,3.016021063455887,7],["hourly",20,0.001,-2.666474975063249,7],["hourly",20,0.002,-10.127375734961797,7],["hourly",30,0.001,-8.219999623529395,7],["hourly",30,0.002,-13.226889472715165,6]]. Trade-mean bootstrap10000 IID resamples95%CI, dependence unresolved: [{"s":"LINK","tf":"daily","CI":[-7.82392442472692,11.117232178145661],"N":7},{"s":"LINK","tf":"hourly","CI":[-6.452175648415601,8.04990960093739],"N":7}]. Conclusion research only; no new deployment based on this comparator. All192cells recorded as observations. Source https://runway.adityaperswal.workers.dev/data/crypto-bars ; https://docs.alpaca.markets/us/docs/crypto-fees .