The whole record, including what failed.
Most trading sites show you a curve. This page shows the tests, the sample sizes, the drawdowns and the strategies we killed. Read the status column first — it tells you whether a number came from a backtest or from real money.
| Strategy | Instrument | Status | Profit factor | Expectancy | Trades | Win rate | Both halves + |
|---|---|---|---|---|---|---|---|
| Straddle (Donchian 20) | USD/JPY H1 | Backtest | 1.23 | +0.120R | 1,710 | 36.4% | Yes |
| Straddle (Donchian 20) | XAU/USD H4 | Backtest | 1.27 | +0.135R | 496 | 36.5% | Yes |
| Donchian + ADX gate | XAU/USD H1 | Backtest | 1.47 | +32.8R total | — | — | Yes |
| Single-direction breakout | USD/JPY H1 | Backtest | 1.11 | +0.059R | 1,072 | 35.6% | Yes |
| Mean-reversion scalper | 12 pairs M5 | Retired | 0.34 | −0.05 to −0.52R | — | 77% | No |
| Straddle | EUR/USD · GBP/USD · USD/CHF | Retired | 0.87–0.97 | negative | 1,400+ | ~32% | No |
What a strategy must clear to appear here.
The bar is deliberately awkward. Most candidates fail it, which is the point.
Positive expectancy
Average result per trade above zero after costs, measured in R rather than currency.
Adequate sample
At least 80–100 trades. Below that, results are noise wearing a suit.
Both halves positive
Split the history in two. Each half must be profitable independently — one good period is not an edge.
Survives stress
Costs doubled, parameters shifted ±20%, fills assumed pessimistic. A cliff edge means curve-fit.
The scalper cost us months. We published it anyway.
A mean-reversion scalper that won 77% of its trades still lost money — the losses were far larger than the wins and costs did the rest. We tested it across twelve pairs and none were profitable. That result is on this page permanently, because a track record that only contains winners is a brochure, not a record.
0 of 12 pairs profitable
77% win rate, but a break-even requirement of 90.9%. Net −$2,928 on the gold test, profit factor 0.34. Fades lose because price trends through the bands, and five-minute costs consume what is left.
Do not fund it. We did not.