GBP/USD H4 Williams %R(14) / EMA(50): fixed-rule educational replay

Actual unchanged four-decimal Forexite minute quotes, January-February 2024.
61,972 source records; 61,792 replay records through 2024-02-29T20:00:00Z.
Source filenames, winter CET and UTC timestamps, original daily-archive URLs
and hashes are retained. The original 60 daily archives were not downloaded
again for this release; 13 saved source-kit hashes were checked.

Run Python3 replay.py, then Python3 verify_replay.py. The frozen protocol hash
is required before replay. Seven result files reproduce byte-for-byte offline.
An independent Decimal70 calculation and a causal minute-by-minute ledger check
all decisions and 30 trades. Optional: Python3 verify_javascript.py (Node.js).
Original SVGs: Python3 make_charts.py; PNG export also requires Inkscape.

WPR14=-100*(HH14-Close)/(HH14-LL14), current 14 completed groups inclusive.
A zero range is unavailable: never create a crossing across missing values.
Long previous<=-80/current>-80; short previous>=-20/current<-20.
EMA50 is seeded at index49 with mean Close[0..49], then uses alpha2/51.
Both variants start at index50. Filtered requires Close>EMA>previous EMA for
long, reversed for short. These conditions are algebraically concordant,
not independent confirmations. Each variant has its own position schedule.
First later minute Open within five minutes, flat at signal checkpoint;
ATR14 risk rounded upward from 1.5*ATR to whole pips, stop1R, target2R,
six-observed-H4-group cap. Gap stop at Open, gap target at target, then cap
at Open, then minute touches with stop-first if both. Study-end at final Close.
The full priority and sample-end convention are in protocol.json.

Costs0/1.2/2.4 pips are hypothetical once-per-trade round-trip allowances.
UI0..5 step0.1 changes accounting only, never the entries or stops.
No historical Ask, spread, slippage, financing or broker fills reconstructed.
Minute times are source records, not exact execution seconds.
The EMA filter worsens the net-R result in this sample. Positive pips do not
prove positive fixed-risk results. No account percentage return is inferred.
