What problem does it solve? Quant strategy ideas are mostly noise, and a single attractive backtest often hides overfitting, cost fragility, or regime dependence. This Skill enforces a disciplined, time-boxed research loop in the market-pilot lab that separates genuine statistical edge from randomness, or honestly proves no edge exists in the available data. ## Core Features & Use Cases - Bounded research sessions: Run 1, 2, or 3-hour sessions with hard caps on hypotheses, batches, and experiments, plus periodic progress reports. - Mandatory validation pipeline: Every hypothesis passes through cost-aware backtesting, baseline comparison (buy-and-hold, cash), out-of-sample robustness gates, cross-asset validation, regime alignment, and a unified robustness score before earning candidate status. - Anti-overfitting discipline: Weak ideas are killed early, conclusions use constrained evidence language only, and no live trading, broker access, or trade signals are permitted. - Use Case: Ask for a 2-hour research session to test mean-reversion and breakout hypotheses across at least 3 instruments in 2 asset classes, and receive a final report listing surviving candidates, failed hypotheses with reasons, data gaps, and a bounded next-experiment plan. ## Quick Start Run a 2-hour research session with /research-agent 2 to test strategy hypotheses and report robust candidates or evidence of no edge.
