V237 EUR/USD M5 New York first break + ADX14 + Bollinger20/2
Educational historical Bid-reference simulation,2024,December2023 warmup.
401310 unique M1;80721 observed M5. HistData fixedEST+5h -> UTC, then historical
America/New_York DST. 08:00NY is chosen FXreference, not an exchange auction.
2023 rawarchive had60 identical duplicate timestamps;2024 hadnone. Rawarchive
SHA256 in source-manifest.json; original ZIPs retained in editorialbackup.

Python3 replay_strategy.py --model C --segment all --cost 1.2
Python3 replay_strategy.py --date 2024-01-11
Only standardlibrary needed. Script recalculates M5,ATR,ADX,Bollinger from
retained M1 OHLC. Stored indicators and returns are not calculation inputs.
S on charts is first signal; data key B stores S, distinct from modelB.
See fixed-method.json for full causal indicator seeds, coverage and exit rules.
A first strict closing break of complete08:00..08:30NY range, known<=10:00;
B addsADX14>=25 at sameS; C adds strict directionalClose outsideBB20/2 atS.
No laterreplacement, no waitforlaterADX, no priorcompression requirement.
Next exactM1Open,1.5ATR risk roundedup0.1pip,stop1R,target2R,90mincap.
Opengapstop atOpen;Openbeyondtarget atTarget;deadline atOpen;touches stopfirst
ifboth. Missingminute beforeexit ->unknown; intraminuteexit is[minute,next).
No Ask-awareexecution, measuredspreads, tickorder or brokerfills claim.
Allincost once/resolvedtrade combines spread,commission,slippage. It changes
netaccounting only, not path/barriers. Pips equalunits; additiveR normalizes
by eachtrade initialrisk. No position sizing/accountreturn/compounding.
C at1.2:30resolved,0unknown,-35pips,-6.2661067249R. H1/H2 descriptiveonly.
First CJan11 isalsofirstloss; firstwinJan31;ADXrejectionJan4;bandrejectionJan2.
Examples chronological across fullyear, not a win-selected success sample.
