JudyAI WaveRider — AI Trading Agent

JudyAI WaveRider is an autonomous crypto trading agent that proves performance on unseen data, not curve-fit backtests. The Problem Most AI trading agents backtest on the same data they optimize on. This is overfitting. They offer no on-chain proof and treat risk management as an afterthought. Our Solution ‧Walk-Forward Validation: 82.2% win rate across 366 out-of-sample trades using 8 rolling windows (90-day train, 30-day test). Every parameter proven on unseen data. ‧Three Strategy Engines: WaveRider (EMA crossover + RSI + volume), BB Squeeze (Bollinger Band breakout), and MACD Divergence (price-momentum reversal). A 36-cell strategy matrix routes the best strategy per coin per market regime. ‧Dual-AI Ensemble: MiniMax M2.7 + Qwen 2.5 cross-validate every signal. Disagreement = no trade. Rule-based fallback if APIs fail. ‧7-Layer Risk Management: Position sizing, daily loss limit, max drawdown, consecutive loss scaling, per-pair throttle, and regime filter. 87% of raw signals rejected. Result: 0.4% max drawdown over 11 days of adverse markets, preserving 99.6% of capital. ‧ERC-8004 On-Chain Identity: Agent #17 on Sepolia. 79 EIP-712 signed trade intents. 214 validation artifacts with SHA-256 Merkle integrity. Validation 98/100, Reputation 94/100, Rank #5 of 58. ‧Radical Transparency: Live win rate was 40% during ranging markets — we show it alongside the 82.2% backtest. The risk system held losses to $377 on $100K. Capital preservation > cherry-picked demos. ‧Kraken CLI Integration: OHLC data for 7 pairs, real-time tickers, paper trading execution, balance tracking. Fully verifiable: make install && make test && make validate && make verify