Five signs a strategy is overfit (and one that says it isn't)

If you've spent any time looking at trading-strategy reports, you'll have noticed they all look amazing. Equity curve goes up and to the right, drawdown is tiny, profit factor is north of 2. Almost none of them work in production. The gap between 'looks amazing' and 'works' is overfitting — and you can usually spot it without running a single backtest yourself.

The five tells

1. Hyper-specific parameters

RSI(14) is fine. RSI(17.4) on a 26-bar Bollinger band with a 1.847 SD multiplier is a strategy fitted to the exact pattern of one historical period. The classical sniff test: nudge the parameters by ±10% and see if performance is stable. If +1.847 turns into +0.4 at 1.95, you're not looking at a strategy, you're looking at a memory of the past.

2. Massive in-sample / out-of-sample gap

If the strategy's profit factor is 3.2 in-sample and 0.9 out-of-sample, you've found the world's most expensive way to discover regression to the mean.

3. Parameter count > trade count / 10

A 5-parameter strategy needs hundreds of trades for the parameters to be meaningful. If it has 12 parameters and 80 closed trades, every parameter is fitted to ~7 trades — that's noise, not signal. Same logic as a regression: degrees of freedom matter.

4. The equity curve is too smooth

Real edge is bumpy. A perfectly smooth, almost-linear equity curve usually means stop placement was tuned to remove all the losses you'd actually take. Glance at the trade list — if losses are suspiciously rare or suspiciously small, the SL is fitted to the past, not to the strategy's natural risk.

5. It only works on one symbol

There's nothing wrong with a strategy designed for a specific symbol. But if it works on EURUSD M15 and breaks completely on every neighbour (GBPUSD, AUDUSD, EURJPY), you've probably tuned it to EURUSD's specific noise rather than a real microstructure feature.

The one sign that suggests it's not overfit

Boring. No flex. Profit factor 1.4, 55% win rate, drawdown that hurts but recovers. That's what a real edge looks like — small, durable, slightly disappointing on a good week. Strategies that beat that profile are almost always either overfit or running in a regime that's about to change.