Kalshi Weather Trading Bot

Overview

Trading bot for Kalshi weather prediction markets using open source weather data, ensemble forecast models, and genetic evolution to optimize strategies.

Key Insight

Treat historical temperature data like financial charts — daily high/low as OHLC candles, 30-year climate normals as moving averages, record temps as support/resistance. Mean reversion is powerful: temps within 3°F of record lows revert toward 25th percentile ~78% within 48h.

Markets Available

• Daily temp highs: 15+ US cities (NYC $756K vol/day)
• Daily temp lows: 6+ cities
• Precipitation: NYC daily rain, monthly rainfall
• Snowfall: 16+ cities monthly
• Natural disasters: Tornadoes, hurricanes
• Climate: Hottest year/month, arctic ice

Data Sources

• Open-Meteo: 80+ years historical + 7-day forecasts, free, no API key
• GEFS/ECMWF ENS: 31/51-member ensemble models for probability estimation
• IEM ASOS: Airport station ground truth (what Kalshi settles against)
• NEXRAD on AWS S3: Real-time doppler radar for precipitation nowcasting
• HRRR: Hourly high-res model updates

Strategies

Strategy 1: "The Grinder" (Safe Bets)

• Ensemble model says >80%, market prices 60-75%
• Exploit longshot bias, recency bias, NWS wet bias
• Target: 75-85% win rate, 50-100 trades/month

Strategy 2: "The Storm Chaser" (Risk Bets)

• Detect extreme weather 6-18h before market reprices
• Radar integration for same-day precipitation
• Target: 45-55% win rate, 2-3x payoff ratio

Strategy 3: "Darwin" (Genetic Evolution)

• 70-80 parameter genome
• Sharpe-based fitness, 50 organisms/generation
• Island speciation by contract type
• Walk-forward validation

Exploitable Biases

1. Longshot bias: Extreme events overpriced 5-15%
2. Recency bias: Post-cold-snap overestimation of continued cold
3. NWS wet bias: Government systematically overforecasts rain

Honest Assessment

• Edge exists but liquidity-constrained
• Round-trip friction ~10% (fees + spread)
• Revenue ceiling: $1K-10K/year
• Primary value: Darwin R&D lab — proving genetic framework on fast-feedback data, then porting to forex/prediction markets

Reports

• /reports/weather-kalshi-markets.md — Market structure (Brianna)
• /reports/weather-data-sources.md — Data sources & tools (Ada)
• /reports/weather-trading-strategy.md — Strategy design (Axe)
• /reports/weather-edge-analysis.md — Edge analysis (Canary)