Look ahead bias is a backtest looking at the future. A trade that buys at the open because the trader knows the candle closes higher, or a signal that fires on a level that only became clear later. The trades look smart because they were made with the answer already on the screen... and that edge disappears the first day you trade it live.
What is look ahead bias?
Any decision in a test that uses data the strategy wouldn't have had yet. In live trading that information doesn't exist when you click, so the result the backtest showed can't happen.
Where does it hide?
Rarely where you'd look first:
| Where it hides | What it looks like |
|---|---|
| Same candle signals | entering at the open of a candle using its close, high or low |
| Indicators that repaint | a signal that moves or vanishes once later candles print |
| Settings fit on the whole history | choosing a parameter with data from the full test, then using it from day one |
| Revised data | earnings or economic numbers corrected later, tested as if they were known at the time |
| Manual backtests | seeing the next candles on the chart before you mark the trade |
The last one catches discretionary traders more than any code bug. Scroll a chart and your eyes already know where price went, so every entry you mark is a little better than the one you would have taken.
How do you catch it?
Make every trade enter one candle after its signal, then run the test again. A real edge barely notices the delay, because it never depended on that candle. For manual tests, use a bar by bar replay and hide everything to the right of the current candle, and write the trade down before you see its outcome. If the results look too good, a win rate far above what traders in that market usually manage or a curve with almost no drawdowns, check this first.
Why it costs more than it looks
Most mistakes shrink an edge when you go live. This one deletes it, because the advantage came from seeing the future. It's also the reason forward testing matters: on new data you can't see tomorrow, so the bias has nowhere to hide. It's one of three ways a backtest lies to you without showing an error, next to survivorship bias and data snooping, and it's a common reason behind a big live vs backtest gap.
Frequently asked questions
Related terms
- Expectancy — What one average trade is worth to you, after the wins and losses cancel out.
- Overfitting — When a strategy is tuned so tightly to the past that it only works on the past.
- Out of Sample — The data your strategy never saw while you built it, kept back to test if the edge is real.
- Walk Forward — Out of sample testing done over and over, rolling through your whole history instead of one split.