Overfitting (curve fitting)
Tuning rules so tightly to past data that they describe history instead of the market.
Full explanation
Overfitting looks like success. The equity curve of the backtest is smooth, the win rate is high, and the parameters are oddly specific — a 37-period average, entries only between 08:14 and 09:02.
The giveaway is fragility. Shift the parameters slightly, or run the rules on a different period or pair, and the performance collapses. Live results then look nothing like the test.
The defences are simple: few parameters, a clear reason for each rule, testing on data you did not tune on, and forward testing before real money.
Why traders watch it
Overfitting is the most common reason a strategy that tested beautifully loses money from the first week it goes live.
Trading considerations
- Be suspicious of very specific parameter values.
- Hold back a period of data the strategy was never tuned on.
- Fewer rules with obvious logic beat many rules with perfect history.
Educational guidance only — never a trading signal or recommendation.