ForexBestRobots.com — V224 confirmed-pivot MFI divergence / VWAP / activity replay

Historical EUR/USD quote simulation, January 2–31 2024; not real-account fills.
Source: HistData Generic ASCII EURUSD January 2024 Bid/Ask ticks.
SHA-256: 9a4ae15d87abff93311cc60adf7fe83de9078664c6baaf2984fcc70f160b3da9
Source clock: fixed EST UTC−05:00; add five hours for UTC.
Source page: https://www.histdata.com/download-free-forex-data/?/ascii/tick-data-quotes/eurusd/2024/1
The original raw archive is identified, not redistributed in this public kit.

protocol.json fixes every rule before the strategy results. No retuning.
Strict Low/High pivots require two complete neighboring candles on each side.
They become known only after both right-hand candles close. Entry is then
modeled at the first qualifying quote, never retrospectively at the pivot.
Use only the immediately previous same-type pivot, 5–30 minutes apart.
Regular divergence is lower Low / higher MFI or higher High / lower MFI.
MFI(14) uses Bid HLC3 times quote count N; equal TP is neutral and zero sums
are unavailable. Exact comparisons precede rounding. No 20/80 threshold.
A: divergence. B: plus confirmation Close on trade side of anchored VWAP.
C: plus confirmation N at least the preceding-20 median (current excluded).
VWAP uses a fixed 07:00 UTC anchor and is weighted by quote count, not lots.

Integer prices are millionths; one EUR/USD pip equals 100 integer units.
quote_units = side*(exit liquidation quote − entry execution-side quote).
Bid/Ask spread is already included. Subtract 70 commission units and
round(extra_pips*100) once per resolved round trip; default extra=0.2 pip.
Fees/allowance are illustrative assumptions. Cost controls change accounting
only, not signal eligibility, entry/exit timestamps or quoted prices.

There are 6613 confirmed pivot records across the source archive,
344 eligible shared divergence events, and 1032 variant decisions.
eurusd-pivot-events.csv retains each pivot and its pair decision, including
failed spacing/MFI/price pairs. A failed nearest pair cannot skip to an older one.
eurusd-m1-inputs.csv retains Bid/Ask OHLC, N/Q, exact MFI/VWAP sums and indices.
eurusd-trades.csv retains all 406 entries across independent A/B/C replays.
*_ms fields are Unix UTC milliseconds; bar t and *_start are UTC seconds.
Indices are zero-based original CSV record indices; equal times keep order.
Quote gaps >60 seconds while open make outcomes unresolved and exclude them
from performance while retaining lockout. Actual unresolved entries: zero.

C has 52 trades, −50.7 quote pips and −97.5 net pips at default costs.
All variants and both chronological segments are net negative. Both filters
worsen sample mean/PF. This is the same known source used in the guides,
V222 and V223; the chronological split is not untouched validation.

Standard-library cost accounting:
  python3 replay_costs.py
  python3 replay_costs.py --variant A --segment later --extra 1

Optional raw-quote reconstruction (NumPy and original ZIP required):
  python3 replay_quotes.py --archive /path/to/DAT_ASCII_EURUSD_T_202401.zip
It creates replayed-v224 and writes seven CSV files. No broker connection,
order submission, cash return, position size, funding or margin model.
