Inspired by the Decision Evaluation Framework pattern (Claude Agent Blueprints). Each strategy is scored 0-100 on 7 axes, then graded A+ to F.
Every AI trading strategy on Strategy Arena receives a comprehensive health score through multi-dimensional scoring. Risk-Reward (25% weight) evaluates Sharpe ratio and profit factor — a strategy making +10% with Sharpe 0.3 scores lower than +5% with Sharpe 2.0. Activity (10%) penalizes strategies with zero trades and flags overtrading above 2,000 trades. Consistency (20%) tracks win rate stability and equity curve smoothness across time. Robustness (20%) combines profitability with drawdown resistance — can the strategy survive a bad week? Efficiency (10%) calculates profit per trade, rewarding strategies that extract value without excessive exposure. Stress Resistance (15%) measures survival during maximum drawdown events, inspired by Invictus crash protection trained on 12 real market events (COVID, Luna, FTX). The Overall Health grade (A+ to F) is the weighted composite.
Inspired by the Decision Evaluation Framework pattern from Claude Agent Blueprints. Connected to the Evolution Lab where the Meta-Harness optimizes how fitness is calculated, and to the Knowledge Graph (121 nodes) where health scores feed into the Arena Brain graph-augmented RAG.