Weather Market NO-Side Fade (Overpriced Temperature Markets)

How It Works Temperature-based prediction markets (e.g., 'Will NYC exceed 95°F on June 20?') are priced by a crowd that often anchors on recent hot spells or media narratives, causing YES probabilities to be systematically inflated. The bot ingests live METAR station observations and Open-Meteo ensemble forecasts, then fits a Gaussian probability model to estimate the true probability of the temperature threshold being breached.

Entry / Exit Conditions When the market-implied YES probability exceeds the model's estimated probability by a configurable margin (e.g., >5 percentage points), the bot opens a NO position. Position size can be scaled to the confidence gap. Exit is at market resolution; no mid-market exit logic is described beyond paper-trading constraints.

Verified Results The repository is described as a paper-trading bot, meaning no live capital has been deployed and no verified PnL exists. The Gaussian model provides a principled probabilistic framework but has not been back-tested against disclosed historical results.

Implementation Notes METAR data provides near-real-time ground truth for current conditions. Open-Meteo offers free ensemble forecast APIs suitable for small-scale systematic use. The Gaussian model is simple to implement but may underfit in cases of bimodal weather patterns or rapidly changing synoptic conditions.

Risks Weather forecasts carry irreducible uncertainty, especially beyond 3–5 days. The model may systematically mis-calibrate for specific geographies or seasons. Paper-trading results do not account for market impact or the possibility that sophisticated weather traders have already corrected the mispricing.