V236 GBP/USD M5 London first break + EMA50 + ATR14

Educational Bid-reference simulation, January-December2024, December2023 warmup.
400913 unique M1 records,80689 observed M5 bars. Source fixedEST+5h to UTC;
London uses historical Europe/London seasonal time. Original archive SHA-256
values retained in source-manifest.json.60 identical duplicate timestamps in each
original annual archive were removed only after equality checks. Raw provider
ZIPs are retained in the editorial backup, not redistributed here.

Use Python3 replay_strategy.py --model C --segment all --cost 1.2
Use Python3 replay_strategy.py --date 2024-01-09
The script uses standard library only and recomputes M5,EMA50 and ATR14 from
M1 OHLC. It does not read stored indicator or return columns. M5 CSV is supplied
for cross-checks, not as the script's source. No terminal or broker connection.

Data key B stores the first-break signal S on charts, distinct from modelB.
Full immutable rules are in fixed-method.json. A first break, B adds EMA,
C adds inclusive range/ATR-at-freeze2..8. One firstB per date, no replacement.
Next exact M1Open at signal-known minute, riskceil(1.5ATR/0.1pip)*0.1pip,
stop1R,target2R,90min cap. Minute priority: adverseOpen through stop=>Open;
favourableOpen beyond target=>target;timecap=>Open;barriertouches=>stopfirst
ifboth. Missing minute before resolution=>unknown. Intraminute exits represent
[minute,nextminute), not exact seconds. All prices are historical Bid references.

Roundtrip cost is one hypothetical combined allowance for spread/commission/
slippage, charged once in accounting. It never changes entries or barrier tests.
No historicalAsk,variable spread,short Ask-triggered stops or executable fills
are reconstructed. Pips use equal units;R divides each result by initial risk.
No account-% return, sizing or compounding. Unresolved trades are excluded from
metrics and counted. C at1.2:211resolved,1unresolved,-138.60pips,-39.64additiveR.
H1/H2 are descriptive chronological halves, not claimed untouched holdout data.
