GBP/USD Williams %R research data — January–February 2024

Actual unchanged Forexite quote records. No invented market prices or trades.
The 60 original daily MetaStock archives were checked using SHA-256 and
ZIP CRC. Source filenames, URLs and hashes are in research/sources.json.
Used minute records preserve four-decimal OHLC, original CET and UTC.
No Ask series, historical spread, execution or trading return is inferred.

Offline reproduction, Python 3 standard library only:
  python3 reproduce.py
  python3 verify_calculation.py
The first recreates research/bars.json, examples.json and the H4 CSV
byte-for-byte. The second independently groups the minute records and
uses direct Decimal arithmetic with 60-digit precision to verify %R.

For the current completed group and previous N-1 observed groups:
  HH = maximum High; LL = minimum Low
  %R = -100 * (HH - current Close) / (HH - LL)
N is 7, 14 or 28. No smoothing. First valid zero-based index is N-1.
If HH equals LL, %R and raw Fast %K are undefined (JSON null), not 0.
No crossing is accepted from an undefined reading.
For a nonzero range, raw Fast %K = %R + 100 on the same OHLC window.

Record timestamps belong to (end-4h,end], labelled by end in UTC.
Winter CET minus one hour. No missing prices filled. Short opening and
closing groups remain. The final source-boundary-cut group is omitted.
A broker's native H4 convention can differ.

Optional original-source extraction: put the 60 named original ZIP files
in raw/, then run python3 prepare_data.py. The included source manifest
is used. FBR_FOREXITE_ARCHIVES can point to another archive directory.
Optional chart generation: python3 make_charts.py. PNG export needs Inkscape.
The offline replay and numerical tests need no network or external package.
