Stats

Preregistered note1253 before data retrieval. Nine sector ETFs; rank prior 5 session adjusted close returns at weekly boundary, buy worst nextopen if its own return negative. SPY subtraction does not change cross-sectional ranks. Prior20 mean high-low range R: initialstop-1R,target+2R,half off+1R and remainingstop entry, maximumnextweeklyopen. Train2016–20/test2021–25/2026 retrospectiveextension; late boundary-crossingtrades omitted. All10series2685alignedsessionsJan2016–Sep8 2026,no missing-valuefill, adjustedOHLC, survivedETFset. Full normalized capital10000, cashonly,fees1bp/slippage15bp eachside includingpartials. Customevent cash and every equitypoint verified by vectorbt from_orders; exactfinal6939.161364714867. Stoppriority on dailycollisions, adversegapstopatopen,targetgaprecognized, partial/BE ambiguousdays assigned adverse sequence. Continuous touch is optimistic vs15minpolling even though collision rule conservative; current-sized wholelots/half-rounding/extendedhourfills notvalidated. Not exact Nagel replication: original48industry/individualstocks,long-short,5daily sleeves,conditionalbetahedge,closingliquiditysupplier andVIX dependence; ETFcash adaptation tests differentpayoff. Premises1.recordedreturns from fixedrule;2.basepositiveedge must survive costs/uncertainty;3.wholefamilyfailsstressandCI. Conclusion research-only/noallocation. Sample 219,p0.547945205479452,W1.3581621844578406%,L1.9731766003059112%,mean-0.14778548536686914%; years{"2021":-5.1554814620838885,"2022":-4.253855505472382,"2023":-5.28169377896509,"2024":-12.922522067412046,"2025":-7.352503039406888};exitcounts{"stop":111,"time":43,"breakeven":38,"target":22,"target_gap":5}. Trade samplecounts completed roundtrips,partiallegs aggregated. Return/drawdown are fullresearchcapital anddailyclose marks,not account performance orworstintradayDD. No fixedmonthlyplatformcharges deducted,which worsen net economics. Allhistoricalwindows previouslyused byotherfundstudies,notpristineOOS; no winnerchosen from sweep. Sources https://www.nber.org/system/files/working_papers/w17653/w17653.pdf ; https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all ; observations16049–16076; localresearch/reversal.py.

Preregistered note1253 before data retrieval. Nine sector ETFs; rank prior 5 session adjusted close returns at weekly boundary, buy worst nextopen if its own return negative. SPY subtraction does not change cross-sectional ranks. Prior20 mean high-low range R: initialstop-1R,target+2R,half off+1R and remainingstop entry, maximumnextweeklyopen. Train2016–20/test2021–25/2026 retrospectiveextension; late boundary-crossingtrades omitted. All10series2685alignedsessionsJan2016–Sep8 2026,no missing-valuefill, adjustedOHLC, survivedETFset. Full normalized capital10000, cashonly,fees1bp/slippage5bp eachside includingpartials. Customevent cash and every equitypoint verified by vectorbt from_orders; exactfinal10752.96360642542. Stoppriority on dailycollisions, adversegapstopatopen,targetgaprecognized, partial/BE ambiguousdays assigned adverse sequence. Continuous touch is optimistic vs15minpolling even though collision rule conservative; current-sized wholelots/half-rounding/extendedhourfills notvalidated. Not exact Nagel replication: original48industry/individualstocks,long-short,5daily sleeves,conditionalbetahedge,closingliquiditysupplier andVIX dependence; ETFcash adaptation tests differentpayoff. Premises1.recordedreturns from fixedrule;2.basepositiveedge must survive costs/uncertainty;3.wholefamilyfailsstressandCI. Conclusion research-only/noallocation. Sample 219,p0.5525114155251142,W1.5489653847268852%,L1.7960280711602061%,mean0.05211899807421413%; years{"2021":3.569097556488754,"2022":4.553662485482968,"2023":3.2246191431342774,"2024":-4.912419336412777,"2025":1.1699758614958533};exitcounts{"stop":111,"time":43,"breakeven":38,"target":22,"target_gap":5}. Trade samplecounts completed roundtrips,partiallegs aggregated. Return/drawdown are fullresearchcapital anddailyclose marks,not account performance orworstintradayDD. No fixedmonthlyplatformcharges deducted,which worsen net economics. Allhistoricalwindows previouslyused byotherfundstudies,notpristineOOS; no winnerchosen from sweep. Sources https://www.nber.org/system/files/working_papers/w17653/w17653.pdf ; https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all ; observations16049–16076; localresearch/reversal.py.

Preregistered note1253 before data retrieval. Nine sector ETFs; rank prior 5 session adjusted close returns at weekly boundary, buy worst nextopen if its own return negative. SPY subtraction does not change cross-sectional ranks. Prior20 mean high-low range R: initialstop-1R,target+1.5R,half off+1R and remainingstop entry, maximumnextweeklyopen. Train2016–20/test2021–25/2026 retrospectiveextension; late boundary-crossingtrades omitted. All10series2685alignedsessionsJan2016–Sep8 2026,no missing-valuefill, adjustedOHLC, survivedETFset. Full normalized capital10000, cashonly,fees1bp/slippage15bp eachside includingpartials. Customevent cash and every equitypoint verified by vectorbt from_orders; exactfinal7258.721592771103. Stoppriority on dailycollisions, adversegapstopatopen,targetgaprecognized, partial/BE ambiguousdays assigned adverse sequence. Continuous touch is optimistic vs15minpolling even though collision rule conservative; current-sized wholelots/half-rounding/extendedhourfills notvalidated. Not exact Nagel replication: original48industry/individualstocks,long-short,5daily sleeves,conditionalbetahedge,closingliquiditysupplier andVIX dependence; ETFcash adaptation tests differentpayoff. Premises1.recordedreturns from fixedrule;2.basepositiveedge must survive costs/uncertainty;3.wholefamilyfailsstressandCI. Conclusion research-only/noallocation. Sample 219,p0.547945205479452,W1.3956566398879215%,L1.97317660030591%,mean-0.12724057828189275%; years{"2021":-9.575724080843905,"2022":1.4530726723529508,"2023":-3.3536905856611954,"2024":-13.306595788986852,"2025":-5.5637869206533015};exitcounts{"stop":104,"target":41,"breakeven":38,"time":29,"target_gap":7}. Trade samplecounts completed roundtrips,partiallegs aggregated. Return/drawdown are fullresearchcapital anddailyclose marks,not account performance orworstintradayDD. No fixedmonthlyplatformcharges deducted,which worsen net economics. Allhistoricalwindows previouslyused byotherfundstudies,notpristineOOS; no winnerchosen from sweep. Sources https://www.nber.org/system/files/working_papers/w17653/w17653.pdf ; https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all ; observations16049–16076; localresearch/reversal.py.

Preregistered note1253 before data retrieval. Nine sector ETFs; rank prior 3 session adjusted close returns at weekly boundary, buy worst nextopen if its own return negative. SPY subtraction does not change cross-sectional ranks. Prior20 mean high-low range R: initialstop-1R,target+2R,half off+1R and remainingstop entry, maximumnextweeklyopen. Train2016–20/test2021–25/2026 retrospectiveextension; late boundary-crossingtrades omitted. All10series2685alignedsessionsJan2016–Sep8 2026,no missing-valuefill, adjustedOHLC, survivedETFset. Full normalized capital10000, cashonly,fees1bp/slippage15bp eachside includingpartials. Customevent cash and every equitypoint verified by vectorbt from_orders; exactfinal5139.891293176538. Stoppriority on dailycollisions, adversegapstopatopen,targetgaprecognized, partial/BE ambiguousdays assigned adverse sequence. Continuous touch is optimistic vs15minpolling even though collision rule conservative; current-sized wholelots/half-rounding/extendedhourfills notvalidated. Not exact Nagel replication: original48industry/individualstocks,long-short,5daily sleeves,conditionalbetahedge,closingliquiditysupplier andVIX dependence; ETFcash adaptation tests differentpayoff. Premises1.recordedreturns from fixedrule;2.basepositiveedge must survive costs/uncertainty;3.wholefamilyfailsstressandCI. Conclusion research-only/noallocation. Sample 225,p0.52,W1.350008917560452%,L2.035598201882041%,mean-0.2750824997719447%; years{"2021":-19.77584279508666,"2022":-16.816823077388243,"2023":4.845641283599722,"2024":-18.778632296479834,"2025":-9.553328538558715};exitcounts{"stop":123,"time":39,"breakeven":35,"target":24,"target_gap":4}. Trade samplecounts completed roundtrips,partiallegs aggregated. Return/drawdown are fullresearchcapital anddailyclose marks,not account performance orworstintradayDD. No fixedmonthlyplatformcharges deducted,which worsen net economics. Allhistoricalwindows previouslyused byotherfundstudies,notpristineOOS; no winnerchosen from sweep. Sources https://www.nber.org/system/files/working_papers/w17653/w17653.pdf ; https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all ; observations16049–16076; localresearch/reversal.py.

Preregistered note1253 before data retrieval. Nine sector ETFs; rank prior 3 session adjusted close returns at weekly boundary, buy worst nextopen if its own return negative. SPY subtraction does not change cross-sectional ranks. Prior20 mean high-low range R: initialstop-1R,target+2R,half off+1R and remainingstop entry, maximumnextweeklyopen. Train2016–20/test2021–25/2026 retrospectiveextension; late boundary-crossingtrades omitted. All10series2685alignedsessionsJan2016–Sep8 2026,no missing-valuefill, adjustedOHLC, survivedETFset. Full normalized capital10000, cashonly,fees1bp/slippage5bp eachside includingpartials. Customevent cash and every equitypoint verified by vectorbt from_orders; exactfinal8060.95807690537. Stoppriority on dailycollisions, adversegapstopatopen,targetgaprecognized, partial/BE ambiguousdays assigned adverse sequence. Continuous touch is optimistic vs15minpolling even though collision rule conservative; current-sized wholelots/half-rounding/extendedhourfills notvalidated. Not exact Nagel replication: original48industry/individualstocks,long-short,5daily sleeves,conditionalbetahedge,closingliquiditysupplier andVIX dependence; ETFcash adaptation tests differentpayoff. Premises1.recordedreturns from fixedrule;2.basepositiveedge must survive costs/uncertainty;3.wholefamilyfailsstressandCI. Conclusion research-only/noallocation. Sample 225,p0.52,W1.552911990645975%,L1.8394731261162218%,mean-0.07543286539987953%; years{"2021":-12.571197649678457,"2022":-9.164950188317588,"2023":14.719415636254652,"2024":-10.951740773413865,"2025":-0.6389174619807036};exitcounts{"stop":123,"time":39,"breakeven":35,"target":24,"target_gap":4}. Trade samplecounts completed roundtrips,partiallegs aggregated. Return/drawdown are fullresearchcapital anddailyclose marks,not account performance orworstintradayDD. No fixedmonthlyplatformcharges deducted,which worsen net economics. Allhistoricalwindows previouslyused byotherfundstudies,notpristineOOS; no winnerchosen from sweep. Sources https://www.nber.org/system/files/working_papers/w17653/w17653.pdf ; https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all ; observations16049–16076; localresearch/reversal.py.

Preregistered note1253 before data retrieval. Nine sector ETFs; rank prior 3 session adjusted close returns at weekly boundary, buy worst nextopen if its own return negative. SPY subtraction does not change cross-sectional ranks. Prior20 mean high-low range R: initialstop-1R,target+1.5R,half off+1R and remainingstop entry, maximumnextweeklyopen. Train2016–20/test2021–25/2026 retrospectiveextension; late boundary-crossingtrades omitted. All10series2685alignedsessionsJan2016–Sep8 2026,no missing-valuefill, adjustedOHLC, survivedETFset. Full normalized capital10000, cashonly,fees1bp/slippage15bp eachside includingpartials. Customevent cash and every equitypoint verified by vectorbt from_orders; exactfinal5418.042992912734. Stoppriority on dailycollisions, adversegapstopatopen,targetgaprecognized, partial/BE ambiguousdays assigned adverse sequence. Continuous touch is optimistic vs15minpolling even though collision rule conservative; current-sized wholelots/half-rounding/extendedhourfills notvalidated. Not exact Nagel replication: original48industry/individualstocks,long-short,5daily sleeves,conditionalbetahedge,closingliquiditysupplier andVIX dependence; ETFcash adaptation tests differentpayoff. Premises1.recordedreturns from fixedrule;2.basepositiveedge must survive costs/uncertainty;3.wholefamilyfailsstressandCI. Conclusion research-only/noallocation. Sample 225,p0.52,W1.3941282054574626%,L2.035598201882043%,mean-0.25214047006550006%; years{"2021":-21.661419014315452,"2022":-14.01779110571465,"2023":6.137142136057849,"2024":-18.073610246089235,"2025":-7.494630644957379};exitcounts{"stop":114,"target":39,"breakeven":35,"time":26,"target_gap":11}. Trade samplecounts completed roundtrips,partiallegs aggregated. Return/drawdown are fullresearchcapital anddailyclose marks,not account performance orworstintradayDD. No fixedmonthlyplatformcharges deducted,which worsen net economics. Allhistoricalwindows previouslyused byotherfundstudies,notpristineOOS; no winnerchosen from sweep. Sources https://www.nber.org/system/files/working_papers/w17653/w17653.pdf ; https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all ; observations16049–16076; localresearch/reversal.py.

Preregistered note1253 before data retrieval. Nine sector ETFs; rank prior 3 session adjusted close returns at weekly boundary, buy worst nextopen if its own return negative. SPY subtraction does not change cross-sectional ranks. Prior20 mean high-low range R: initialstop-1R,target+1.5R,half off+1R and remainingstop entry, maximumnextweeklyopen. Train2016–20/test2021–25/2026 retrospectiveextension; late boundary-crossingtrades omitted. All10series2685alignedsessionsJan2016–Sep8 2026,no missing-valuefill, adjustedOHLC, survivedETFset. Full normalized capital10000, cashonly,fees1bp/slippage5bp eachside includingpartials. Customevent cash and every equitypoint verified by vectorbt from_orders; exactfinal8497.18698967831. Stoppriority on dailycollisions, adversegapstopatopen,targetgaprecognized, partial/BE ambiguousdays assigned adverse sequence. Continuous touch is optimistic vs15minpolling even though collision rule conservative; current-sized wholelots/half-rounding/extendedhourfills notvalidated. Not exact Nagel replication: original48industry/individualstocks,long-short,5daily sleeves,conditionalbetahedge,closingliquiditysupplier andVIX dependence; ETFcash adaptation tests differentpayoff. Premises1.recordedreturns from fixedrule;2.basepositiveedge must survive costs/uncertainty;3.wholefamilyfailsstressandCI. Conclusion research-only/noallocation. Sample 225,p0.52,W1.597119605511995%,L1.8394731261162252%,mean-0.05244490566955085%; years{"2021":-14.626110737345076,"2022":-6.108440230685142,"2023":16.132542794177795,"2024":-10.178779321536025,"2025":1.622685403241042};exitcounts{"stop":114,"target":39,"breakeven":35,"time":26,"target_gap":11}. Trade samplecounts completed roundtrips,partiallegs aggregated. Return/drawdown are fullresearchcapital anddailyclose marks,not account performance orworstintradayDD. No fixedmonthlyplatformcharges deducted,which worsen net economics. Allhistoricalwindows previouslyused byotherfundstudies,notpristineOOS; no winnerchosen from sweep. Sources https://www.nber.org/system/files/working_papers/w17653/w17653.pdf ; https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all ; observations16049–16076; localresearch/reversal.py.

Preregistered note1253 before data retrieval. Nine sector ETFs; rank prior 5 session adjusted close returns at weekly boundary, buy worst nextopen if its own return negative. SPY subtraction does not change cross-sectional ranks. Prior20 mean high-low range R: initialstop-1R,target+1.5R,half off+1R and remainingstop entry, maximumnextweeklyopen. Train2016–20/test2021–25/2026 retrospectiveextension; late boundary-crossingtrades omitted. All10series2685alignedsessionsJan2016–Sep8 2026,no missing-valuefill, adjustedOHLC, survivedETFset. Full normalized capital10000, cashonly,fees1bp/slippage5bp eachside includingpartials. Customevent cash and every equitypoint verified by vectorbt from_orders; exactfinal11078.92489715759. Stoppriority on dailycollisions, adversegapstopatopen,targetgaprecognized, partial/BE ambiguousdays assigned adverse sequence. Continuous touch is optimistic vs15minpolling even though collision rule conservative; current-sized wholelots/half-rounding/extendedhourfills notvalidated. Not exact Nagel replication: original48industry/individualstocks,long-short,5daily sleeves,conditionalbetahedge,closingliquiditysupplier andVIX dependence; ETFcash adaptation tests differentpayoff. Premises1.recordedreturns from fixedrule;2.basepositiveedge must survive costs/uncertainty;3.wholefamilyfailsstressandCI. Conclusion research-only/noallocation. Sample 35,p0.6857142857142857,W1.3326076064494985%,L1.9265552735026028%,mean0.30829927275026664%; years{"2026":10.789248971576026};exitcounts{"stop":14,"breakeven":9,"target":8,"time":3,"target_gap":1}. Trade samplecounts completed roundtrips,partiallegs aggregated. Return/drawdown are fullresearchcapital anddailyclose marks,not account performance orworstintradayDD. No fixedmonthlyplatformcharges deducted,which worsen net economics. Allhistoricalwindows previouslyused byotherfundstudies,notpristineOOS; no winnerchosen from sweep. Sources https://www.nber.org/system/files/working_papers/w17653/w17653.pdf ; https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all ; observations16049–16076; localresearch/reversal.py.

Preregistered note1253 before data retrieval. Nine sector ETFs; rank prior 5 session adjusted close returns at weekly boundary, buy worst nextopen if its own return negative. SPY subtraction does not change cross-sectional ranks. Parent exits nextweeklyopen without protectiveexits, not venueeligible. Train2016–20/test2021–25/2026 retrospectiveextension; late boundary-crossingtrades omitted. All10series2685alignedsessionsJan2016–Sep8 2026,no missing-valuefill, adjustedOHLC, survivedETFset. Full normalized capital10000, cashonly,fees1bp/slippage5bp eachside includingpartials. Customevent cash and every equitypoint verified by vectorbt from_orders; exactfinal13996.983495286033. Stoppriority on dailycollisions, adversegapstopatopen,targetgaprecognized, partial/BE ambiguousdays assigned adverse sequence. Continuous touch is optimistic vs15minpolling even though collision rule conservative; current-sized wholelots/half-rounding/extendedhourfills notvalidated. Not exact Nagel replication: original48industry/individualstocks,long-short,5daily sleeves,conditionalbetahedge,closingliquiditysupplier andVIX dependence; ETFcash adaptation tests differentpayoff. Premises1.recordedreturns from fixedrule;2.basepositiveedge must survive costs/uncertainty;3.wholefamilyfailsstressandCI. Conclusion research-only/noallocation. Sample 35,p0.6857142857142857,W2.19947912344991%,L1.6359066382048224%,mean0.9940721697869942%; years{"2026":39.96983495286044};exitcounts{"time":35}. Trade samplecounts completed roundtrips,partiallegs aggregated. Return/drawdown are fullresearchcapital anddailyclose marks,not account performance orworstintradayDD. No fixedmonthlyplatformcharges deducted,which worsen net economics. Allhistoricalwindows previouslyused byotherfundstudies,notpristineOOS; no winnerchosen from sweep. Sources https://www.nber.org/system/files/working_papers/w17653/w17653.pdf ; https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all ; observations16049–16076; localresearch/reversal.py.

Preregistered note1253 before data retrieval. Nine sector ETFs; rank prior 5 session adjusted close returns at weekly boundary, buy worst nextopen if its own return negative. SPY subtraction does not change cross-sectional ranks. Prior20 mean high-low range R: initialstop-1R,target+1.5R,half off+1R and remainingstop entry, maximumnextweeklyopen. Train2016–20/test2021–25/2026 retrospectiveextension; late boundary-crossingtrades omitted. All10series2685alignedsessionsJan2016–Sep8 2026,no missing-valuefill, adjustedOHLC, survivedETFset. Full normalized capital10000, cashonly,fees1bp/slippage5bp eachside includingpartials. Customevent cash and every equitypoint verified by vectorbt from_orders; exactfinal11248.155939006509. Stoppriority on dailycollisions, adversegapstopatopen,targetgaprecognized, partial/BE ambiguousdays assigned adverse sequence. Continuous touch is optimistic vs15minpolling even though collision rule conservative; current-sized wholelots/half-rounding/extendedhourfills notvalidated. Not exact Nagel replication: original48industry/individualstocks,long-short,5daily sleeves,conditionalbetahedge,closingliquiditysupplier andVIX dependence; ETFcash adaptation tests differentpayoff. Premises1.recordedreturns from fixedrule;2.basepositiveedge must survive costs/uncertainty;3.wholefamilyfailsstressandCI. Conclusion research-only/noallocation. Sample 219,p0.5525114155251142,W1.5862244122917777%,L1.7960280711602088%,mean0.07270503613518095%; years{"2021":-1.2577553398408714,"2022":10.785560863044473,"2023":5.325769419946624,"2024":-5.33182331833888,"2025":3.123232803065701};exitcounts{"stop":104,"target":41,"breakeven":38,"time":29,"target_gap":7}. Trade samplecounts completed roundtrips,partiallegs aggregated. Return/drawdown are fullresearchcapital anddailyclose marks,not account performance orworstintradayDD. No fixedmonthlyplatformcharges deducted,which worsen net economics. Allhistoricalwindows previouslyused byotherfundstudies,notpristineOOS; no winnerchosen from sweep. Sources https://www.nber.org/system/files/working_papers/w17653/w17653.pdf ; https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all ; observations16049–16076; localresearch/reversal.py.

Preregistered note1253 before data retrieval. Nine sector ETFs; rank prior 5 session adjusted close returns at weekly boundary, buy worst nextopen if its own return negative. SPY subtraction does not change cross-sectional ranks. Parent exits nextweeklyopen without protectiveexits, not venueeligible. Train2016–20/test2021–25/2026 retrospectiveextension; late boundary-crossingtrades omitted. All10series2685alignedsessionsJan2016–Sep8 2026,no missing-valuefill, adjustedOHLC, survivedETFset. Full normalized capital10000, cashonly,fees1bp/slippage5bp eachside includingpartials. Customevent cash and every equitypoint verified by vectorbt from_orders; exactfinal10550.627692510181. Stoppriority on dailycollisions, adversegapstopatopen,targetgaprecognized, partial/BE ambiguousdays assigned adverse sequence. Continuous touch is optimistic vs15minpolling even though collision rule conservative; current-sized wholelots/half-rounding/extendedhourfills notvalidated. Not exact Nagel replication: original48industry/individualstocks,long-short,5daily sleeves,conditionalbetahedge,closingliquiditysupplier andVIX dependence; ETFcash adaptation tests differentpayoff. Premises1.recordedreturns from fixedrule;2.basepositiveedge must survive costs/uncertainty;3.wholefamilyfailsstressandCI. Conclusion research-only/noallocation. Sample 219,p0.547945205479452,W2.3086207067727176%,L2.6229487299582943%,mean0.07928109838746565%; years{"2021":17.555219237760223,"2022":-17.779409986678264,"2023":12.300994549576117,"2024":2.5599976586681006,"2025":-5.224913683076449};exitcounts{"time":219}. Trade samplecounts completed roundtrips,partiallegs aggregated. Return/drawdown are fullresearchcapital anddailyclose marks,not account performance orworstintradayDD. No fixedmonthlyplatformcharges deducted,which worsen net economics. Allhistoricalwindows previouslyused byotherfundstudies,notpristineOOS; no winnerchosen from sweep. Sources https://www.nber.org/system/files/working_papers/w17653/w17653.pdf ; https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all ; observations16049–16076; localresearch/reversal.py.

Preregistered note1253 before data retrieval. Nine sector ETFs; rank prior 5 session adjusted close returns at weekly boundary, buy worst nextopen if its own return negative. SPY subtraction does not change cross-sectional ranks. Prior20 mean high-low range R: initialstop-1R,target+1.5R,half off+1R and remainingstop entry, maximumnextweeklyopen. Train2016–20/test2021–25/2026 retrospectiveextension; late boundary-crossingtrades omitted. All10series2685alignedsessionsJan2016–Sep8 2026,no missing-valuefill, adjustedOHLC, survivedETFset. Full normalized capital10000, cashonly,fees1bp/slippage5bp eachside includingpartials. Customevent cash and every equitypoint verified by vectorbt from_orders; exactfinal9321.38308558911. Stoppriority on dailycollisions, adversegapstopatopen,targetgaprecognized, partial/BE ambiguousdays assigned adverse sequence. Continuous touch is optimistic vs15minpolling even though collision rule conservative; current-sized wholelots/half-rounding/extendedhourfills notvalidated. Not exact Nagel replication: original48industry/individualstocks,long-short,5daily sleeves,conditionalbetahedge,closingliquiditysupplier andVIX dependence; ETFcash adaptation tests differentpayoff. Premises1.recordedreturns from fixedrule;2.basepositiveedge must survive costs/uncertainty;3.wholefamilyfailsstressandCI. Conclusion research-only/noallocation. Sample 219,p0.5296803652968036,W1.2918659205613918%,L1.4892365755847583%,mean-0.01614118949821297%; years{"2016":-6.488625913214707,"2017":-11.238119385883005,"2018":-4.014194298051255,"2019":-0.5454962826797471,"2020":17.640795784197994};exitcounts{"stop":103,"target":38,"time":37,"breakeven":37,"target_gap":4}. Trade samplecounts completed roundtrips,partiallegs aggregated. Return/drawdown are fullresearchcapital anddailyclose marks,not account performance orworstintradayDD. No fixedmonthlyplatformcharges deducted,which worsen net economics. Allhistoricalwindows previouslyused byotherfundstudies,notpristineOOS; no winnerchosen from sweep. Sources https://www.nber.org/system/files/working_papers/w17653/w17653.pdf ; https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all ; observations16049–16076; localresearch/reversal.py.

Preregistered note1253 before data retrieval. Nine sector ETFs; rank prior 5 session adjusted close returns at weekly boundary, buy worst nextopen if its own return negative. SPY subtraction does not change cross-sectional ranks. Parent exits nextweeklyopen without protectiveexits, not venueeligible. Train2016–20/test2021–25/2026 retrospectiveextension; late boundary-crossingtrades omitted. All10series2685alignedsessionsJan2016–Sep8 2026,no missing-valuefill, adjustedOHLC, survivedETFset. Full normalized capital10000, cashonly,fees1bp/slippage5bp eachside includingpartials. Customevent cash and every equitypoint verified by vectorbt from_orders; exactfinal9339.544206340903. Stoppriority on dailycollisions, adversegapstopatopen,targetgaprecognized, partial/BE ambiguousdays assigned adverse sequence. Continuous touch is optimistic vs15minpolling even though collision rule conservative; current-sized wholelots/half-rounding/extendedhourfills notvalidated. Not exact Nagel replication: original48industry/individualstocks,long-short,5daily sleeves,conditionalbetahedge,closingliquiditysupplier andVIX dependence; ETFcash adaptation tests differentpayoff. Premises1.recordedreturns from fixedrule;2.basepositiveedge must survive costs/uncertainty;3.wholefamilyfailsstressandCI. Conclusion research-only/noallocation. Sample 219,p0.54337899543379,W1.9476171220379839%,L2.2486592099630833%,mean0.031509207882245496%; years{"2016":7.380320007257146,"2017":-5.6854951288370925,"2018":6.974500634784531,"2019":21.73115524058902,"2020":-29.182516886926045};exitcounts{"time":219}. Trade samplecounts completed roundtrips,partiallegs aggregated. Return/drawdown are fullresearchcapital anddailyclose marks,not account performance orworstintradayDD. No fixedmonthlyplatformcharges deducted,which worsen net economics. Allhistoricalwindows previouslyused byotherfundstudies,notpristineOOS; no winnerchosen from sweep. Sources https://www.nber.org/system/files/working_papers/w17653/w17653.pdf ; https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all ; observations16049–16076; localresearch/reversal.py.

Frozen note1254 before history download. Rule: prior close>SMA100 and RSI2 EWMA alpha .5 adjustfalse<10; next adjacent open; exit prior RSI>70 or10 bars. Development2025, holdout2026Jan1–Sep8, each starts flat. Development thresholds5/10/15 x baseline/stress all uncomputable because history startsFeb17 2026. Vectorbt from_signals fees25bp cash each side, slip30bp baseline60bp stress; fee approximation is not exact credited-asset accounting, no15minstop/target/profitmanagement simulated. {"quality":{"tf":"1Hour","pages":24,"n":4882,"first":"2026-02-17T13:00:00+00:00","last":"2026-09-08T23:00:00+00:00","missing":1,"zero":3805},"window":"holdout","threshold":10,"slip":0.003,"invalid":false,"return_pct":-60.17996814165475,"drawdown_pct":60.18793088766776,"n":89,"win_rate_pct":29.213483146067414,"avg_win_pct":0.7914118253239084,"avg_loss_pct":1.7598876033687094,"expectancy_pct":-1.0145641747618772,"open":0}. Stress return-76.655559%, N89,p13.483146%,meanwin.792083%,loss1.982278%,expectancy-1.608207%. Premises1 no training means no qualified model; 2 current top10 membership creates survivorship and reused2026 is not pristine unseen data; 3 zero-volume bars do not establish executable liquidity. Hourly one missing interval while flat retains warmup continuity limitation, no interpolation; held-gap would invalidate. Conclusion size0, do not move split date to manufacture training. Strong counterexample daily+18.4% is four outcomes and cannot rescue hourly failure. Source https://runway.adityaperswal.workers.dev/data/crypto-bars ; https://docs.alpaca.markets/us/docs/crypto-fees . Reproduce hype.py; detailed results observation16041/16042.

Frozen note1254 before history download. Rule: prior close>SMA100 and RSI2 EWMA alpha .5 adjustfalse<10; next adjacent open; exit prior RSI>70 or10 bars. Development2025, holdout2026Jan1–Sep8, each starts flat. Development thresholds5/10/15 x baseline/stress all uncomputable because history startsFeb17 2026. Vectorbt from_signals fees25bp cash each side, slip30bp baseline60bp stress; fee approximation is not exact credited-asset accounting, no15minstop/target/profitmanagement simulated. {"quality":{"tf":"1Day","pages":1,"n":204,"first":"2026-02-17T00:00:00+00:00","last":"2026-09-08T00:00:00+00:00","missing":0,"zero":16},"window":"holdout","threshold":10,"slip":0.003,"invalid":false,"return_pct":18.396617256845293,"drawdown_pct":10.624176942415907,"n":4,"win_rate_pct":75,"avg_win_pct":6.824850847960659,"avg_loss_pct":2.661979920842749,"expectancy_pct":4.453143155759807,"open":0}. Stress return15.588867%, N4, p75%, averagewin6.184312%, loss3.245767%. Premises1 no training means no qualified model; 2 current top10 membership creates survivorship and reused2026 is not pristine unseen data; 3 zero-volume bars do not establish executable liquidity. Hourly one missing interval while flat retains warmup continuity limitation, no interpolation; held-gap would invalidate. Conclusion size0, do not move split date to manufacture training. Strong counterexample daily+18.4% is four outcomes and cannot rescue hourly failure. Source https://runway.adityaperswal.workers.dev/data/crypto-bars ; https://docs.alpaca.markets/us/docs/crypto-fees . Reproduce hype.py; detailed results observation16041/16042.

1. Ran mandated CLI --strategy breakout --lookback20 --days365 across7majors; per-symbol return/DD/winrate/N observations saved. BTCgross5.929854%,DD9.036712%,p40%,N20; SOL31.42404%,XRP55.13033% counterexamples to blanket bearishness.2. Inspected engine.js/bars.js: close-derived signal enters sameclose, nofees/slippage, noholdout, fetch hasnoend andignoresnextpagetoken.366bars mayincludeincompleteday. Do not confuse these attractive results with executable futurepayoffs.3. Independentrecordedcosted daily1099zeroqualifiedtraining; hourly1112all30computableholdoutslose, selectedUNI-42.91764%,DD72.21537%,N61,p50.81967%,meanwin5.26321%,meanloss6.76089%; selectedSOLinvalid becausemissingheldexitbar. Exactfees/rangeexits/vectorbtreconciliation and splitwindows are inoriginalrecords, notnewlyrerunhere.4. ConclusionrejectCLIasdeploymentbasis. Soundnesslimitsglobalreuse2026, survivorship, missingbars and15minexitresolution. Source https://runway.adityaperswal.workers.dev/data/crypto-bars ; https://docs.alpaca.markets/us/docs/crypto-fees .

Ran node /app/agent/dist/backtest.js --symbols AVO --strategy momentum --lookback20 --days730.500bars,24trades. Premise1 genericmomentumdoesnotestimateadatedearningsedge.2 CLIreturnnegative,noindependentCLIholdout. Conclusionnegativecrosscheckonly; chronologicalvectorbteventtestseparatelyrecordedgoverns. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=AVO . Do not usebuyholdoutperformanceasrunupprobability.

Premises1 primary23earningsJan2021–Jun2026 from67SEC8Ks,actualearningssentences recorded;Dec19 2024excludedbecauseNov5preannouncement https://www.sec.gov/Archives/edgar/data/1802974/000180297424000046/avo-20241105.htm .2 Frozenbeforeoutcomes note1015,train2021–24N16/holdout2025–26N6,T1previoussessionopen-close;vectorbtlongonly,10bpfee+10bpslippage eachside;independentcompounding matches.3 Baseline train-15.6021%,holdout-7.3318%,DD9.372%,p50%,W1.31844%,L3.74363%,mean-1.21259%,95%bootstrapmeanCI[-3.75774,1.05684]%.1ATRstop/1.5ATRtargetconservativestopfirst:train-20.4866%,holdout-4.01016%,DD6.12344%,p50%,W1.31844%,L2.62743%,mean-.654495%.4 All16holdoutvariants(h0/1/3/5,time/barrier,slip10/25bp)lose. h0same-dateconditionaldiagnostic-5.51296%time/-9.55199%barrierN6,p0; firstreleaseAMCnotverifiedhistorically so notvalidtradablestrategy. CONCLUSION reject,size0; Sep8cannotborrowSep4prior-sessionwindow. Validitynecessaryedgefails; soundnesslimitsarchivaladvance-noticesmissing,shortselectedcohort,adjusteddailyOHLCnot15minfills,May28Calavostructuralbreakonlyonepostcloseevent. No assumptionofknownhistoricaldatesbeforeactualnotice. Statisticalnegativitysufficestostopcostlyarchive/fillresearch; notproofallrunupslose. Market https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=AVO&timeframe=1Day&feed=sip&adjustment=all . 100kbootstrapseed2026090808,reproresearch/test.py.

Executed node /app/agent/dist/backtest.js --symbols MATX --strategy breakout --lookback 20 --days 1700 --stop-pct 3 --target-pct 4.5. CLI1165bars,97trades,return-30.82459%,DD41.17376%,wins28.86598%,avgtrade-.35229%,buyhold158.23066%. Source https://runway.adityaperswal.workers.dev/data/stock-bars . ExistingCLIauditanalysis53 establishes IEX/default nofees/touchfills/samesignalclose semantics; notsame testasnewSIP nextopenADRrule. Do notattribute discrepancy solelytofeed and do notuseCLI as executablevalidation. Required diagnosticruncompleted; preserveoutput/command. Noholdoutpretended inCLI; customtestseparatelyholdsout2026.

Preregisterednote1013. ConsolidatedSIPadjustmentall daily1425bars2021–Sep4 2026 https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=MATX&timeframe=1Day&feed=sip&adjustment=all . Training2022–24,validation2025,holdoutJan1–Sep4 2026, resetflat. Close>previous20dailyHIGH AND>SMA50; nextopen long, previousADR20fixed1Rstop/1.5Rtarget,max10sessions. Entryday exitsincluded in two dailyexecution slots; gapstop paidopen,gaptarget cappedtarget,bothintradaytouchesstopfirst,no sameday reentry. Fees5bp+slippage10bp eachside,stress25bp slippage. vectorbt explicitlongonly tradecount/compoundedreturn reconciled independentledger; daily marked drawdown not15minvenueexecution. Train+12.7051%,DD13.9755%,N38,p52.6316%,avgwin3.4027/loss2.9987%; validation+6.3149%,DD2.977%,N5; holdout-1.1617%,DD10.7850%,N13,p46.1538%,avgwin3.4881/loss3.0562%;stress-4.9422%,DD13.6811%. Fixedbaseline notchanged aftertraining-only9neighbor sweep. Correction to code explicitly prioritizes gappedtarget before laterintradaystop, resultsunchanged. No claim weather causality or execution at15minobservations. Selectedcurrentticker/smallholdout/corporateactioncash and executableovernightquote gaps mean positive result alone would notqualify. Here pointedgealsofails; allocationzero.

Fixed L126/H20 monthly top1positive sector momentum, nineETFs; train2017-20/holdout2021-25/recent2026throughSep4. Managed1ATRstop1.5ATRtarget half+1ATR thenBE, daily stopfirst ambiguouspath; nonoverlap, nextopenlaggedsignal. 5bpfees+20bpslippageperfill. Signed eventledger independentlyreconciles vectorbt maxUSDerror3.637978807091713e-12. Fees on everypartial included. {"stop":4,"samebar_BE":1,"gap_stop":1,"target":1,"gap_target":1,"boundary":1}. Data2016floor2684bars/symbol, alladjustedtotalreturnproxy; cashdividends/exactfees/15minutequotes/whole-share/PDT unvalidated. OOS is chronological notpristine duepreviousmarketwindowreuse. Parentcannotauthorizeorderwithoutrequiredexits. Thisrule notactiveSep8monthlyentry. Source https://runway.adityaperswal.workers.dev/data/stock-bars ; frozen notes1011/1014, strategy1121.

Fixed L126/H20 monthly top1positive sector momentum, nineETFs; train2017-20/holdout2021-25/recent2026throughSep4. Managed1ATRstop1.5ATRtarget half+1ATR thenBE, daily stopfirst ambiguouspath; nonoverlap, nextopenlaggedsignal. 5bpfees+20bpslippageperfill. Signed eventledger independentlyreconciles vectorbt maxUSDerror3.637978807091713e-12. Fees on everypartial included. {"stop":27,"target":13,"samebar_BE":7,"gap_target":5,"gap_stop":4,"BE":3}. Data2016floor2684bars/symbol, alladjustedtotalreturnproxy; cashdividends/exactfees/15minutequotes/whole-share/PDT unvalidated. OOS is chronological notpristine duepreviousmarketwindowreuse. Parentcannotauthorizeorderwithoutrequiredexits. Thisrule notactiveSep8monthlyentry. Source https://runway.adityaperswal.workers.dev/data/stock-bars ; frozen notes1011/1014, strategy1121.

Fixed L126/H20 monthly top1positive sector momentum, nineETFs; train2017-20/holdout2021-25/recent2026throughSep4. Managed1ATRstop1.5ATRtarget half+1ATR thenBE, daily stopfirst ambiguouspath; nonoverlap, nextopenlaggedsignal. 5bpfees+20bpslippageperfill. Signed eventledger independentlyreconciles vectorbt maxUSDerror3.637978807091713e-12. Fees on everypartial included. {"target":9,"stop":23,"samebar_BE":7,"gap_target":7,"BE":2}. Data2016floor2684bars/symbol, alladjustedtotalreturnproxy; cashdividends/exactfees/15minutequotes/whole-share/PDT unvalidated. OOS is chronological notpristine duepreviousmarketwindowreuse. Parentcannotauthorizeorderwithoutrequiredexits. Thisrule notactiveSep8monthlyentry. Source https://runway.adityaperswal.workers.dev/data/stock-bars ; frozen notes1011/1014, strategy1121.

Fixed L126/H20 monthly top1positive sector momentum, nineETFs; train2017-20/holdout2021-25/recent2026throughSep4. Managed1ATRstop1.5ATRtarget half+1ATR thenBE, daily stopfirst ambiguouspath; nonoverlap, nextopenlaggedsignal. 5bpfees+5bpslippageperfill. Signed eventledger independentlyreconciles vectorbt maxUSDerror1.8189894035458565e-12. Fees on everypartial included. {"stop":4,"samebar_BE":1,"gap_stop":1,"target":1,"gap_target":1,"boundary":1}. Data2016floor2684bars/symbol, alladjustedtotalreturnproxy; cashdividends/exactfees/15minutequotes/whole-share/PDT unvalidated. OOS is chronological notpristine duepreviousmarketwindowreuse. Parentcannotauthorizeorderwithoutrequiredexits. Thisrule notactiveSep8monthlyentry. Source https://runway.adityaperswal.workers.dev/data/stock-bars ; frozen notes1011/1014, strategy1121.

Fixed L126/H20 monthly top1positive sector momentum, nineETFs; train2017-20/holdout2021-25/recent2026throughSep4. Managed1ATRstop1.5ATRtarget half+1ATR thenBE, daily stopfirst ambiguouspath; nonoverlap, nextopenlaggedsignal. 5bpfees+5bpslippageperfill. Signed eventledger independentlyreconciles vectorbt maxUSDerror3.637978807091713e-12. Fees on everypartial included. {"stop":27,"target":13,"samebar_BE":7,"gap_target":5,"gap_stop":4,"BE":3}. Data2016floor2684bars/symbol, alladjustedtotalreturnproxy; cashdividends/exactfees/15minutequotes/whole-share/PDT unvalidated. OOS is chronological notpristine duepreviousmarketwindowreuse. Parentcannotauthorizeorderwithoutrequiredexits. Thisrule notactiveSep8monthlyentry. Source https://runway.adityaperswal.workers.dev/data/stock-bars ; frozen notes1011/1014, strategy1121.