PolymarketWeatherEdgeGaussianNo

This strategy is a refined version of weather market trading that specifically targets overpriced temperature markets by betting the NO side. The BallesJr/polymarket-weather-edge repository implements a paper-trading bot that ingests live METAR (aviation weather) observations and Open-Meteo forecast data, then applies a Gaussian probability model to estimate the true probability of a temperature threshold being exceeded. When the market's implied probability for YES exceeds the model's estimated probability by a sufficient margin, the bot bets NO.

The core edge is informational: METAR data provides real-time, high-quality temperature observations from airports, often before they are fully reflected in Polymarket prices. The Gaussian model converts forecast and observation data into a probability distribution over the temperature outcome, allowing the bot to identify mispriced contracts. The NO side is favored because retail traders tend to overpay for YES on sensational weather events (e.g., 'Will NYC hit 100F?'), creating a persistent overpricing bias.

Implementation: Python-based, 11.4MB, with topics including algorithmic-trading, metar, open-meteo, paper-trading, polymarket, prediction-markets, quantitative-finance, and weather. The bot is explicitly paper-trading only, meaning no live capital is at risk by default. Entry occurs when model probability diverges from market price; exit is at resolution or when the market price converges to the model. The Gaussian model likely assumes temperature anomalies are normally distributed, with parameters estimated from historical data and forecasts.

Risks: Weather models can be wrong, especially for extreme events where the Gaussian assumption breaks down (fat tails). METAR data may have gaps or quality issues. Polymarket weather markets can have thin liquidity, leading to slippage. Since it is paper-trading, real-world execution may differ. The repository has 6 stars, indicating some community interest.

Verified results: As a paper-trading bot, there are no verified live PnL figures. However, the strategy logic is sound and aligns with known inefficiencies in weather prediction markets. The repository is actively maintained (updated September 15, 2026).