Weather Market NO-Side Bias (Overpriced Temperature Markets)

How It Works Polymarket temperature markets (e.g., 'Will NYC reach 95°F this week?') are often mispriced because retail participants overestimate tail weather probabilities. This strategy uses live METAR station observations and Open-Meteo forecast data to compute a Gaussian probability model for the temperature outcome, then compares this model price to the Polymarket price. When the market price significantly exceeds the model's estimated probability, the bot bets NO.

Entry/Exit Conditions Entry is triggered when the Polymarket YES price exceeds the Gaussian model's probability estimate by a configurable threshold (implying the market is overpriced relative to meteorological data). The bot exclusively trades the NO side. Exit occurs at market resolution or if the spread closes to an unfavorable level.

Verified Results The system is described as an automated paper-trading bot, meaning no live capital results are available. The use of METAR (official aviation weather observations) and Open-Meteo (open-source numerical weather prediction) provides a credible, real-time data foundation. No backtested statistics or Sharpe ratios are disclosed.

Implementation Notes The Gaussian probability model is a reasonable approximation for temperature distributions over short horizons. METAR data provides high-frequency ground-truth observations. Open-Meteo provides free, high-quality forecast data. The strategy is niche but structurally sound — weather markets on prediction platforms are likely to be less efficiently priced than financial derivatives markets.

Risks Gaussian temperature models can underestimate tail probabilities during extreme weather events. METAR observations may not perfectly match the geographic reference point used by Polymarket's market definition. Weather forecasts degrade rapidly beyond 3–7 days. If the strategy becomes known, arbitrageurs may close the pricing gap. Paper-trading results may not reflect live execution friction.