Weather Market NO-Side Bias Trading via METAR + Gaussian Model

How It Works Polymarket lists temperature markets such as 'Will the high temperature in [city] exceed X°F on [date]?' The strategy models the true probability of the outcome using a Gaussian distribution fitted to live METAR (aviation weather observation) data and Open-Meteo probabilistic forecasts. When the market's YES price exceeds the model's estimated probability by a meaningful margin, the bot bets NO (sells YES), capturing the overpricing.

Entry/Exit Conditions The bot pulls live METAR observations for the relevant weather station and current Open-Meteo forecast data. It computes a Gaussian probability P(temperature > threshold) using the forecast mean and standard deviation. If market_YES_price > model_probability + edge_threshold, the bot enters a NO position. The position is held to resolution, as weather markets have fixed expiration dates.

Verified Results The repo is described as an 'automated paper-trading bot,' indicating the strategy is being validated in simulation against real Polymarket prices and real weather data. The use of official METAR data (ground truth) and a named probabilistic model (Gaussian) gives the approach a concrete, testable basis. No live PnL is reported.

Implementation Notes METAR data is freely available from NOAA and aviation APIs. Open-Meteo provides free probabilistic forecast APIs. The Gaussian model is a reasonable approximation for short-horizon temperature distributions but may underestimate tail risk in severe weather events. The bot must correctly map Polymarket market definitions to the appropriate METAR station and threshold units.

Risks Temperature distributions are not perfectly Gaussian, especially in transitional seasons or unusual weather regimes. METAR stations may not match the exact location implied by the market. Resolution source used by Polymarket may differ from METAR (e.g., official NWS readings vs. airport sensors). The market may be efficiently priced if other weather-informed traders are already active.