Strategy Backtester · Advanced
Free to read
Avoiding Overfitting & Look-Ahead Bias
The easiest backtest to build is the one that has already happened. These are the traps that make a dead strategy look alive.
Overfitting
Add enough conditions and any rule will look brilliant on past data — you have simply described the history, not a repeatable edge. Favor few parameters, and be suspicious of a rule that only works with oddly specific thresholds.
Look-ahead bias
Using information that would not have been available at the moment of the trade silently inflates results. Signals must be computed only from data that existed before the entry bar. The engine fills close-to-close to keep this honest.
Out-of-sample is the only real test
A rule earns trust when it holds up on data it was not designed against. Reserve a slice of history you never look at while building, and judge the rule there. If the edge evaporates out of sample, it was never real.
Open the Backtester →Explained →How to read this tool →
Educational content from Nations Indexes. Backtests are descriptive of historical data and are not investment advice or a promise of future results.