Your backtest shows a 2.5 profit factor. You tell yourself this is the one. You go live. Six weeks later your account is down 22% and your PF is 1.08. What happened? Overfitting. We built 20 strategies on EURUSD in-sample data (2021-2024) on FXAbsolute, then tested them on completely unseen 2025-2026 data. Here is how much they collapsed.
| Strategy Type | In-Sample PF | Out-of-Sample PF | Collapse % |
|---|---|---|---|
| 50 EMA Crossover | 1.42 | 1.18 | -17% |
| 50/200 EMA + RSI Filter | 1.88 | 1.32 | -30% |
| EMA + RSI + Session Filter | 2.14 | 1.24 | -42% |
| Multi-Indicator (4+ params) | 2.52 | 1.08 | -57% |
| SMC Structure (discretionary) | 2.31 | 1.15 | -50% |
The pattern is clear: the more parameters you tune, the more your strategy memorizes the in-sample data, and the harder it collapses on unseen data. The simple 50 EMA — one parameter — only dropped 17%. The multi-indicator strategy with 4+ tuned parameters dropped 57% — from a "great" 2.52 PF to an essentially breakeven 1.08.
FXAbsolute gives you 5 years of data. Train on 2021-2024. Test on 2025-2026. See if your edge survives reality.
▶ Start Free BacktestingEvery additional parameter you tune — every indicator, every session filter, every "exclude that weird week" — increases in-sample PF and decreases out-of-sample PF. The ratio is brutal: roughly 10-15% collapse per additional tuned parameter on average. The strategy with 4 indicators might look incredible (PF 2.5), but it collapses to PF 1.1 when shown data it has never seen. That is not a strategy. That is a memory of the past.
Split your data. On FXAbsolute, use 2021-2024 for development. Lock your rules. Then test on 2025-2026 without changing a single thing. If the PF drops more than 20%, you overfitted. Go back. Remove parameters. Simplify. Test again. The best strategies have 1-2 parameters and survive out-of-sample with minimal degradation.
One afternoon on FXAbsolute with out-of-sample data will tell you. Free. No excuses.
▶ Test Your Strategy Now