★ My

Encyclopedia › Trading terms ›

Overfitting

In one line: When a trading rule fits old data so tightly that it only learned the noise.

Imagine a key cut so precisely for one lock that it opens nothing else. That is an overfit strategy. It looks amazing on old data and falls apart on new data.

Sullivan, Timmermann and White (1999) checked nearly 8,000 trading rules. Old results survived their data-snooping test, but the best rule did not beat the market in the next period. Park and Irwin (2007) warn that many technical-analysis studies suffer from data snoopingTrying many tests and showing only the lucky one..

Defences: keep fresh data aside (In-sample vs out-of-sample), use few rules, and compare with Random entry (random twin).

Related

Sources & studies

  1. Sullivan, Timmermann & White (1999), Journal of Finance
  2. Park & Irwin (2007), Journal of Economic Surveys
  3. Investopedia (secondary): Overfitting

Trading termsP-hacking and data snoopingIn-sample vs out-of-sample

Last checked: 2026-11-04 · Educational, not financial advice.

Education only, not financial advice. No signals. Backtests are hypothetical: real trading can be worse. Past results don't predict future results.