Overfitting Quant Research Interview Guide
Overfitting quant research interview guide for train/test gaps, noise fitting, model complexity, validation, and backtest examples.
Candidates preparing for model validation and research process questions.
Overfitting learns noise
Overfitting happens when a model or strategy fits historical noise instead of a relationship that generalizes to new data.
Complexity increases risk
More features, parameters, and research attempts can improve in-sample results while weakening out-of-sample credibility.
Concrete example
A strategy tuned across hundreds of thresholds may find a great historical threshold by chance, especially without a proper holdout.
Mitigation
Use holdouts, cross-validation where appropriate, simpler models, pre-defined research rules, and economic reasoning to reduce overfitting risk.
Common mistakes
Candidates often say use train/test split and stop. Good answers explain why the split helps and what risks remain after validation.
Practice the pattern
Use the LeetQuidity curriculum and calibration to turn this topic into a focused practice plan.