V238 — USD/JPY M5 Asian false breakout, first reentry and RSI14

Educational 2024 Bid-reference test with December 2023 indicator warmup.
Retained inputs: 401,025 unique M1 records and 80,533 observed M5 candles.
HistData timestamps use fixed EST: add five hours for UTC. Tokyo 09:00–12:00
is UTC 00:00–03:00 all year. Each original annual archive contained 60
identical timestamp duplicates, removed only after equality checks.
Source SHA-256 hashes are included; original ZIPs are in the source backup.

REPLAY (Python 3, standard library only; run from this extracted folder)
python3 replay_strategy.py --model C --segment all --cost 1.2
python3 replay_strategy.py --date 2024-06-14

The script rebuilds M5, RSI14, ATR14, first events, trades and metrics from
M1 OHLC. Stored indicator/result columns are not used as calculation inputs.
Prices are integers in 0.000001 JPY per USD. One USD/JPY pip is 0.01;
0.1 pip is 0.001. Integer quote units must not be mistaken for pips.

FIXED METHOD
fixed-method.json contains the protocol frozen before the results.
S is the first complete close strictly outside the frozen Tokyo 09:00–12:00
range, known by 15:00 Tokyo inclusive. R is the first later complete close
strictly inside, known within 30 minutes of S (deadline included).
A wick alone is not S. Boundary equality is not R. No later replacement.
A fades S at the first R. B also needs RSI_R < 50 after upward S, or > 50
after downward S. C additionally needs RSI_S >= 50 after upward S, or <= 50
after downward S. C does not require exactly one crossing of 50.
The chart's first break S is stored under data key B, distinct from model B.

Entry uses the exact M1 Open at R's known time, in the opposite direction
to S. Missing exact Open means unavailable entry, with no later replacement.
Risk is 1.5*ATR14 at R, rounded upward to 0.1 pip. Stop 1R, target 2R,
90-minute cap. Target is not forced to the range midpoint or opposite edge.
Exit order: missing minute -> unknown; adverse opening gap -> Open;
favourable opening gap beyond target -> target; time cap -> Open;
otherwise High/Low touches, assuming stop first if both touch in one minute.
Intraminute exits are intervals [minute, next minute), not exact fill seconds.

COSTS AND INTERPRETATION
One hypothetical round-trip allowance includes spread, commission and
slippage. It is deducted once per resolved trade and changes no entries or
barriers. Exact integer 0.1-pip units prevent a true zero being counted as a
win. No historical Ask fills, variable spreads, tick order or broker fills
are reconstructed. Pips assume equal position units. Additive R divides each
trade result by its own initial risk before summing; it is not account return.

A: 85 entered paths, 84 resolved and one unknown. B/C: the same two resolved
shorts, both losses. C at 1.2 pips: -36.10 pips, -1.4440456133 R.
There is no C winner and no B-accepted/C-rejected case in this sample.
First C: June 14. Second C: July 31. First RSI rejection (A): January 4.
First A winner rejected by RSI: January 5. First unknown A: May 14.
Additional illustrations are recorded in illustration-selection.json;
these presentation choices do not change strategy parameters.
Each half-year has only one C trade, so it cannot validate stability.
Unknown outcomes are excluded from performance totals and counted separately.
