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

How It Works This strategy identifies temperature-based prediction markets on Polymarket where the YES side is overpriced relative to what meteorological data suggests. It ingests live METAR (Meteorological Aerodrome Report) observations — real-time surface weather readings from airports — alongside Open-Meteo numerical weather forecasts, then fits a Gaussian probability distribution over expected temperatures to compute a fair-value probability.

Entry Conditions A trade is entered on the NO side when the market's implied YES probability exceeds the Gaussian model's estimated probability by a meaningful margin (the specific threshold would be defined in the bot's configuration). The edge is the mispricing between crowd sentiment/lazy pricing and physics-grounded meteorological models.

Verified Results The repository describes this as a paper-trading bot, so no live P&L is confirmed. However, the use of METAR data (authoritative, low-latency official observations) and a calibrated probabilistic model suggests the edge thesis is methodologically sound.

Implementation Notes Implemented in Python, automated via GitHub Actions for scheduling. Uses Open-Meteo's free API for forecasts and parses METAR strings for current conditions. The Gaussian model converts temperature forecast distributions into win probabilities directly comparable to market prices.

Risks Temperature markets may have thin liquidity, limiting position size. Extreme weather events (non-Gaussian tails) can cause model failure. METAR stations may not match the exact location specified in a market's resolution criteria, introducing basis risk.