Backtest Results

WMR Fix Reversal

Dukascopy Tick Data · OOS 2022–2026 · IS 2019–2022 · commission included

Fading the pressure move into the 4pm London WM/R fix on GBP/USD is still profitable after costs, but only just: out-of-sample it earned about 0.7% a year at 0.5% risk per trade. The reversal documented in the academic literature survives post-2015 as a residue, not as a tradeable edge.

GBP/USD

Real residue - too small to trade

$10k → $10.3k

+$294 net profit

Win Rate: 42.96% · PF: 1.225 · Max DD: 2.21%

Strategy Rules

Compute the pre-fix move: price(fix) - price(fix - PreFixWindowMin minutes), where fix = 16:00 London time converted to UTC with last-Sunday-of-March/October DST rules

If the pre-fix move is DOWN by at least MinMovePips, buy at the first M5 bar at/after the fix (fade the pressure)

Mirror: pre-fix move UP by at least MinMovePips -> sell at the fix

one trade per fix (per day)

hard time exit HoldMinutes after entry - the reversal horizon in the literature is 1-15 minutes, month-end effects up to ~1h; TP is a formality

MonthEndOnly toggle (default false) kept fixed, not optimized - month-end-only trading cannot reach the trade-count gates but is the strongest documented case; record its subset stats in the report

Backtest Parameters

In-Sample Window2019.01.01 – 2022.01.01

Out-of-Sample Window2022.01.01 – 2026.01.01

Data SourceDukascopy tick data (real ticks)

Starting Capital$10,000

Risk per Trade0.5% of equity

Commission$3.50/side/lot (IC Markets Raw, modeled)

OOS Runs Consumed1

Methodology

Genetic optimization on the in-sample window with a net-of-costs robustness criterion; parameter-plateau selection (top candidates must have profitable neighborhoods, not just a best pass); out-of-sample window run once per frozen candidate on real ticks; verdicts gated on net profit factor, drawdown, and IS-to-OOS retention.

Key Takeaway

This is the rare rejection where the effect is real: 11 of 12 gates passed, including profit-factor retention (0.88), a 2.2% max drawdown, and profitability under doubled costs. It failed exactly one gate - annualized-return retention (0.39 vs the 0.5 bar): the in-sample ~1.9%/yr shrank to ~0.74%/yr out-of-sample, roughly 70bps a year. Consistent with Evans' research, the month-end subset carried half the profit from just 13% of trades (19 trades, 58% win rate) - but a month-end-only variant cannot generate enough trades to be validated on our gates. A real anomaly, economically too small to trade at retail.

Disclaimer: Past performance is not indicative of future results. These backtest results are based on historical data with realistic commission assumptions. Real trading involves additional risks including execution delays, variable spreads, and emotional decision-making.