Backtest Results

Small-Edges Cross-Family Portfolio

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

Sixteen strategies went through this pipeline; four kept a real out-of-sample edge but individually failed our quality gates. Combined at pre-registered volatility-parity weights, the four uncorrelated streams pass every gate we have - the first strategy on this site to do it.

Portfolio

Passed all 14 gates

$10k → $11.5k

+$1,496 net profit

Win Rate: 42.97% · PF: 1.232 · Max DD: 3.2%

Strategy Rules

Component 1 - Gotobi Tokyo fix (USD/JPY, 38.1% weight): buy 9h before the Tokyo fix on gotobi calendar days, exit at the fix

Component 2 - WMR fix reversal (GBP/USD, 54.1% weight): fade 15-pip pre-fix moves into the 4pm London fix, hold 30 minutes

Component 3 - Asian range breakout (USD/JPY, 3.6% weight): trade the daily breakout of the 00:00-04:00 UTC range, flat by 18:00

Component 4 - BTC Donchian trend (BTC/USD, 4.2% weight): H4 channel breakout with ATR stop, signal exit on the opposite channel

Weights are volatility-parity from IN-SAMPLE data only, frozen before the combined out-of-sample curve was computed

One-shot evaluation: no re-weighting, no component swaps, no second attempt after seeing the result

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 Trade1.0% of equity

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

OOS Runs Consumed0 new (offline combination of frozen component runs; components' own counts: gotobi 6, range breakout 2, WMR 1, BTC Donchian 1 - and components were selected on this window, see key takeaway)

Methodology

No new backtests were run: the portfolio is an analysis-layer combination of the four components' already-frozen out-of-sample runs. The design was pre-registered before the combined curve existed - universe by rule (every strategy whose OOS edge held), weights from in-sample volatility parity only, a fixed benchmark formula, and one shot with no re-weighting. Verdicts gated on net profit factor, drawdown, IS-to-OOS retention, cost stress, Monte Carlo resampling, and a weight-blended buy-and-hold/momentum baseline.

Key Takeaway

Diversification did the work: the four components' individual max drawdowns were 7.0%, 23.7%, 2.2% and 25.2%, yet the combined book never drew down more than 3.2% out-of-sample because their daily P&L streams are effectively uncorrelated (average pairwise correlation -0.01). Profit factor 1.232 with 87% retention from in-sample, profitable in all four out-of-sample years, still profitable with commission stressed to $8.50 a side, 0.4% Monte Carlo probability of loss, and 3.9x the risk-adjusted return of the best naive baseline blend. One honest caveat, pre-registered before the combined curve existed: the components were selected BECAUSE they survived out-of-sample, so this study characterizes how known edges combine - it is not untouched holdout validation. The live forward test on this site is the real exam, and the headline return is modest by construction (~3.6%/yr at these conservative weights): this is a risk-adjusted result, not a get-rich chart.

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.