AlgoXpert Alpha Research Framework: A Rigorous IS-WFA-OOS Protocol for Mitigating Overfitting in Quantitative Strategies

Transitioning a strategy from backtest to live operation is where many quantitative trading systems fail, primarily due to parameter overfitting, selection bias, and fragility under regime shifts. This paper proposes the AlgoXpert Alpha Research Framework-a standardized, decision-oriented protocol that evaluates strategies across three chronological stages: (i) In-Sample (IS), which prioritizes stable parameter regions (plateaus) over single optima; (ii) Walk-Forward Analysis (WFA) with rolling windows and purge gaps to mitigate information leakage, equipped with majority-pass and catastrophic-veto decision gates; and (iii) Out-of-Sample (OOS) holdout under strict parameter lock (no further tuning). The framework integrates a defense-in-depth architecture: structural (cliff veto), execution (spread/leverage guards), and equity protection (circuit breakers, kill switch). A case study on USDJPY M5 intraday illustrates the procedure for detecting overfitting through performance degradation and drawdown integrity under chronology. A post-validation report comparing four alpha variants (v1-v4) reveals rank reversal when switching the objective from maximizing Sharpe to minimizing MaxDD, underscoring the trade-off between risk-adjusted performance and tail-risk control.