AlphaCloud

Weather derivatives trading on Kalshi

What it is

AlphaCloud is a fully automated trading bot for Kalshi's temperature markets. It trades contracts like "Will the high in NYC tomorrow be between 45 and 46°F?" — binary bets priced from 1¢ to 99¢ that pay out $1.00 if they're right.

14 cities. 300+ markets per scan. 12 scans per day.

Why this exists

Honestly? This is the data scientist side of me wanting a playground. Weather prediction markets are a fascinating niche — the underlying data is completely public (weather forecasts), the markets are liquid enough to trade, and there's real edge available if you can blend models better than the average participant.

It's also one of the best systems engineering problems I've worked on. Real-time data ingestion, ensemble model blending, probabilistic edge detection, execution algorithms, risk management — every piece has to work together or the whole thing falls apart.

How the pipeline works

It starts with pulling ensemble forecasts from four weather model sources:

GEFS — NOAA's Global Ensemble Forecast System
ECMWF — the European model, generally considered the best in the world
ICON — Germany's DWD model, strong for short-range
NBM — NWS National Blend, a meta-model that combines many sources

Each source runs multiple simulations with slightly different starting conditions. My blending layer combines them with freshness weighting — newer model runs matter more — and produces a probability distribution for each city's temperature.

Then comes the fun part. The edge detector converts those probabilities into bracket prices using Gaussian CDF analysis and compares them against what Kalshi's market is actually pricing.

If my models say 40% and the market says 25¢ — that's a potential edge.

Position sizing uses adaptive Kelly criterion at three confidence levels. Execution runs an IOC/GTC hybrid — high-confidence trades execute immediately, lower-confidence trades sit as limit orders.

The exit system is something I spent a lot of time on. It watches for forecast updates and runs a 4-tier response: EMERGENCY (the model completely flipped), HARD (significant shift), SOFT (minor drift), HOLD (no change). Combined with daily loss limits and position caps, it keeps things from going sideways.

Where it's at

Paper trading — running the full pipeline against real markets with simulated money. I've got 12 parameter sweep configurations testing different edge thresholds, sizing aggressiveness, and exit sensitivity. The goal is to find a config that's consistently profitable before putting real capital behind it.

No rush. This one's for fun and learning.