Weather Market Mean-Reversion (NO-Side Bias)

How It Works Polymarket lists temperature-threshold markets (e.g., 'Will NYC reach 95°F on July 4th?'). Retail participants tend to overweight salient or dramatic outcomes, causing YES prices to be systematically inflated above the meteorologically correct probability. The BallesJr bot ingests live METAR (aviation weather) observations for the relevant station and Open-Meteo probabilistic forecasts, fits a Gaussian probability model to the temperature distribution for the target period, and compares the model probability against the current market YES price. When the market YES price exceeds the model probability by a configurable threshold, the bot places a NO trade.

Entry / Exit Conditions Entry: market_YES_price > gaussian_model_probability + edge_threshold. The NO contract is purchased at the prevailing ask. Exit is at market settlement (binary resolution based on official temperature measurement). No mid-contract exit is described; the strategy is hold-to-settlement.

Verified Results The repo is in paper-trading mode only, meaning no real capital has been deployed. No backtest statistics are published. However, the methodology is grounded in established meteorological data sources (METAR, Open-Meteo), which lends credibility to the edge thesis.

Implementation Notes The Gaussian model assumes temperature on a given day is normally distributed around a forecast mean with forecast-derived variance. This is a reasonable first-order approximation but ignores fat tails from unusual weather patterns. The use of METAR data (updated hourly) for intraday recalibration is a meaningful implementation detail. Paper-trading mode allows validation before capital deployment.

Risks - Gaussian assumption understates tail probabilities; rare heat events (exactly the scenarios most likely to be traded) may be systematically undermodeled. - Market resolution may use a different temperature station or measurement standard than the METAR source, creating basis risk. - Weather forecast uncertainty compounds significantly beyond 3–5 days, limiting the strategy to near-term contracts. - Retail overcrowding on the NO side (other arbitrageurs running similar models) will compress the edge over time.