AI-powered weather trading bot for Kalshi. 82-member dual ensemble forecasting with NWS bias correction. Full Python source code.
Real Kalshi trades with real money. Every settlement verifiable on-chain via the Kalshi API.
Strategy uses a go/no-go algorithm with 15-trade minimum, 60% win rate gate, and 1.50 profit factor gate before approving live deployment. Track record is from the current model epoch (April 2026).
GFS (31 members) + ECMWF (51 members) = 82 ensemble members. More data points than any competing bot. Open-Meteo API, zero cost.
Kalshi settles on NWS station observations. Our model applies 70% NWS weighting to correct for gridded-vs-station divergence. Unique to WeatherEdge.
Quarter-Kelly sizing. 8% minimum edge. Circuit breakers: daily loss limit, max drawdown halt, per-position caps. Live and paper modes with independent tracking.
Crosses the spread on high-edge signals (>10%) for guaranteed fills. Rests as maker on lower-edge signals for zero fees. Adaptive execution.
NYC, CHI, MIA, AUS, DEN, BOS, DAL, PHX, SEA, SFO, ATL, HOU, MIN, LAX. Per-city sigma calibration for market-specific accuracy.
Cron-scheduled scans. Automated settlement reconciliation. Position deduplication. Go/no-go live readiness evaluation. Set it and forget it.
| Feature | WeatherEdge ($97) | Predict & Profit ($67) | Open-Source Bots (Free) |
|---|---|---|---|
| Ensemble Members | 82 (GFS+ECMWF) | 62 (HGEFS) | 31 (GFS only) |
| NWS Bias Correction | Yes (70% weight) | No | No |
| Verified Win Rate | 81% (13W/3L) | 17% (1W/5L) | Unknown |
| Cities | 14 | 14 | 5 |
| Bracket Markets | Yes (192/day) | No | No |
| Circuit Breakers | Yes (3 layers) | Unknown | No |
| Smart Taker | Yes | No | No |
| Live/Paper Modes | Both | Both | Simulation only |
| Settlement Recon | Automated | Unknown | No |
| Source Code | Full Python | Full Python | Full Python |
Bot fetches all active weather markets across 14 cities. Threshold markets (will temp exceed X?) and bracket markets (will temp fall in range?).
82-member ensemble (GFS 31 + ECMWF 51) generates probability distributions. NWS bias correction aligns gridded forecasts to station observations.
Compare model probability to market price. Only trade when edge exceeds 8%. Quarter-Kelly position sizing limits risk.
Places orders via Kalshi API. Smart taker crosses spread on high-edge signals. Automated reconciliation tracks settlements and P&L.
Yes. Kalshi is a CFTC-regulated prediction market exchange. You'll need to create an account at kalshi.com and generate API credentials. The setup guide walks you through this.
Zero. The bot uses free APIs (Open-Meteo for weather data, Kalshi API for trading). No paid subscriptions required. You only need Python 3.9+ and an internet connection.
Yes. The bot defaults to paper (simulated) mode. Paper trades are tracked in a separate SQLite database with full P&L history. Switch to live with the --live flag when ready.
The bot was designed and tested with $100 starting capital. Circuit breakers protect against large drawdowns. You set your own position sizes and risk limits in the config.
No. This is a software tool for educational and informational purposes. Trading involves risk of loss. Past performance does not guarantee future results. You are solely responsible for your trading decisions.
Python 3.9 or higher. All dependencies install via pip (requirements.txt included). Tested on macOS and Linux.
One-time purchase. Full source code. No recurring fees. Start trading weather markets with an edge.
Coming Soon — $97Available soon on Lemon Squeezy. Full source code delivered instantly after purchase.