🦀 CRABZILLA — Weather Trading Intelligence

🤖

ML Resolution Model

ECMWF + GFS forecast features, trained on 5,914 resolved markets. 76% win rate, 61.7% ROI.

📡

Real-Time Monitoring

WebSocket blockchain feed, live weather data, whale alerts, accumulation detection.

🌡️

METAR Lag Strategy

Exploiting delay between airport observations and market updates. 5-12% edge window.

🐋

Whale Intelligence

Behavioral fingerprinting, farmer detection, copy-trading signals from 7,296 addresses.

📊

Forecast Divergence

ECMWF vs GFS model disagreement as alpha signal. Automatic edge detection per city.

⚡

Automated Execution

Half-Kelly sizing, stop-loss management, direct CLOB integration via Polygon.

crab@crabzilla ~

$ python3 train_model_v3.py

📂 Loading 5,914 resolved markets...

🌡️ Fetching ECMWF/GFS forecasts for 17 cities...

🔧 Building 20 features (7 forecast + 13 market)...

🎯 Training RandomForest (200 trees)...

✅ AUC-ROC: 0.911 | Win rate: 76% | ROI: 61.7%

💰 Edge >20%: 944 trades, $0.477 avg profit