Weather Market NO-Side Fade (METAR + Open-Meteo Gaussian Model)

How It Works This strategy targets Polymarket temperature/weather markets where the YES side is systematically overpriced relative to statistically derived probabilities. The bot ingests live METAR (Meteorological Aerodrome Report) observations — real-time airport weather station data — alongside Open-Meteo forecast API data. It then fits a Gaussian probability model to estimate the true probability of the temperature threshold being met.

Entry Conditions A NO position is entered when the implied YES probability (market price) exceeds the Gaussian model's estimated true probability by a configurable margin. The edge source is the market's tendency to overprice dramatic weather outcomes, combined with the availability of high-quality free meteorological data that most retail traders do not systematically process.

Exit Conditions Positions are held to resolution (binary outcome at market close/event time). The paper-trading implementation suggests no active exit before resolution, relying on the statistical edge aggregating over many trades.

Verified Results & Implementation Notes Two related repositories exist: `BallesJr/polymarket-weather-edge` (the NO-fade automated bot) and `RiekertQuant/polymarket-weather-bot-poc` (a proof-of-concept with 21 stars and 5 forks, suggesting community validation). Both operate in paper-trading mode. The Gaussian model calibration is the core IP — the quality of METAR station selection and model fit to local microclimates is a key implementation risk. Size: 1579KB and 3092KB respectively, suggesting real code bases.

Risks Weather markets may be thin and subject to oracle/resolution disputes. Gaussian model may underfit tail weather events (storms, inversions). METAR stations may not correspond to the exact location referenced in the market question. Polymarket may delist or limit weather markets.