Backtesting Statistics Interview Guide
Backtesting statistics interview guide for train/test splits, leakage, bias, multiple testing, robustness, and toy-strategy caveats.
Quant researcher candidates discussing historical strategy evaluation.
Backtests are evidence, not proof
A backtest evaluates historical behavior under assumptions. It does not prove live performance, especially if research choices were made after seeing results.
Check leakage and bias
Lookahead bias, survivorship bias, selection bias, and data leakage can make a strategy look better than it would have in real time.
Concrete example
If a strategy uses revised data that was not available on the trade date, the backtest may include future information and overstate live feasibility.
Use validation
Out-of-sample testing, walk-forward logic, robustness checks, and cost assumptions can improve credibility, but none remove all uncertainty.
Common mistakes
Candidates often focus only on return. A strong backtest discussion includes risk, costs, turnover, bias, and robustness.
Practice the pattern
Use the LeetQuidity curriculum and calibration to turn this topic into a focused practice plan.