EUR/USD H4 CCI(20) return / SMA(50) research model

Actual unchanged four-decimal Forexite minute quotes, January–February 2024.
The source minute CSV preserves the original filename, CET time, converted UTC
and OHLC. Original daily-archive hashes and URLs are retained in sources.json;
those raw archives are not needed for the included offline replay.

Run, Python 3 standard library only:
  python3 replay.py
  python3 verify_replay.py
The fixed protocol hash is checked before any replay. Seven generated research
outputs reproduce byte-for-byte. Independent Decimal70 indicators and a causal
minute-by-minute ledger verify all decisions and trades, plus synthetic cases.

Optional JavaScript numerical check, with Node.js installed:
  python3 verify_javascript.py
Optional original SVG generation: python3 make_charts.py; PNG needs Inkscape.

Rules: completed CCI(20) return above -100 for long / below +100 for short.
Shared start at H4 index50. Filtered variant also requires Close and the SMA50
slope to agree. First later minute Open within five minutes; one position per
variant. ATR14 Wilder mean TR[1..14], then recursive weight 1/14. Risk is
ceil(1.5*ATR14_pips), one whole pip = 0.0001. Stop1R, target2R, six-observed-group
cap. Gap stop at Open, gap target at target, then cap at Open; same-minute both
touches assume stop first. Full conditions and priorities are in protocol.json.

The two variants are independent runs, not a combined portfolio. Transaction
allowances of 0.0/1.2/2.4 pips are hypothetical and deducted once per trade.
No historical Ask, spread, tick order, slippage, financing or broker fills are
reconstructed. R statistics do not establish account percentage returns.
Stop/target times identify minute records, not exact execution seconds.
The short sample and negative transaction-adjusted R result do not prove an edge.
