# Alpha Research Validation Lab > Institutional-grade strategy validation, walk-forward analysis, and regime-aware testing. Use when this capability is needed. - Skill: `tomevault-io/alpha-research-validation-lab` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/alpha-research-validation-lab` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/alpha-research-validation-lab/raw - Safety review: pending (external: skill-scanner PASS, skillspector PASS) - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Research & Search - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/alpha-research-validation-lab --- # Alpha Research Skill This skill provides the "Truth Layer" for all trading strategies. Follow these protocols to move any strategy from "Idea" to "Production": ### 1. The WFA Protocol (Walk-Forward Analysis) No strategy is considered institutional if it only works on historical "backtest" data. - **5-Fold split**: Every strategy must be tested across at least 5 walk-forward windows. - **IS/OOS Ratio**: Use 80% In-Sample (IS) for optimization and 20% Out-Of-Sample (OOS) for validation. - **Robustness Score**: A strategy passes if its OOS performance (Profit Factor/Sharpe) is within 20% of its IS performance. ### 2. Sensitivity Protocol - Call `scripts/sensitivity_analyzer.py` to stress-test parameters. - **The "Dead Zone" Check**: If shifting a parameter (e.g., Moving Average period) by +/- 1 period causes the strategy to fail, it is curve-fitted and must be REJECTED. ### 3. Regime Awareness - Every alpha must specify its "Natural Regime" (Trending, Mean-Reverting, or Volatility-Expansion). - Call `Regime Scout` (Sub-skill) to verify the strategy only runs during its alpha-positive window. ## Core Tools: - `scripts/wfa_engine.py`: Multi-fold Walk-Forward validator. - `scripts/sensitivity_analyzer.py`: Parameter stress-tester. - `scripts/regime_scout.py`: HMM-based market state classifier. --- > Converted and distributed by [TomeVault](https://tomevault.io/claim/thehaywire) — claim your Tome and manage your conversions.