ISSUE 11

LIVEPythonClaude APIMCPFlask

CONTEXT

Short-dated temperature contracts get mispriced when forecast distributions disagree with market-implied probabilities.The hard part is the meteorology, not the execution.

BUILD

Pulls multi-source forecasts from NWS, HRRR, GFS, and ECMWF ensembles, calculates probability distributions across temperature buckets, and acts only when modelled edge clears a threshold.Claude-powered agents re-evaluate every 30 minutes with position recovery and risk management.APScheduler drives the engine, SQLAlchemy persists state, and a Flask dashboard surfaces live performance.

RESULT

Runs autonomously with daily outcome tracking that feeds back into agent decisions, closing the loop between forecast accuracy and position sizing.

LINKS

SCREENSHOTS

Kalshi Weather Trader screenshot 1
← All projects