When a great result is just luck: overfittingIntermediate

How to tell a real edge from a curve-fitted one.

How it happens

Try enough variations of anything and one of them will look brilliant on past data. Change the RSI length, the stop, the session, the target, and keep the best: you have probably fitted the noise.

Three quick checks

Nearby settings: does RSI 12 and 16 also work, or only 14? Other periods: does it work in the years you did not tune on? Other markets: does it work on a similar market? A real edge usually survives all three, a little weaker.

The habit to build

Decide the rules before you look at the result. Change one thing at a time and write down why. Keep a final period of data you never tune on, and test there once at the end.

Try it on your own strategy

Write your rules in plain words and test them on real data in seconds.

Open WickCheck

More lessons

Why most strategies need a human behind the screenWhat a backtest can and cannot tell you about discretionary trading, and how to use both.Trading sessions and killzones: times, what happens in each, and how to test themAsian, London and New York sessions and the ICT killzones, with exact times in New York and London time, daylight saving notes and test rules.What a backtest is, and what it is notThe basics in five minutes: rules, data, costs and the limits of testing the past.How costs quietly eat small edgesSpread, commission and slippage explained, and why tight stops make them worse.Why 30 trades prove nothingHow luck hides in small samples and how many trades you actually need.R, risk per trade and drawdown in plain wordsHow position size, R-multiples and drawdown fit together.How to read your WickCheck resultsWhat each number means and the order to look at them in.

Education only, not financial advice. Definitions vary between teachers; these are ours, written for clarity.