Question 1

What is a weather prediction market?

Accepted Answer

A weather prediction market is a financial exchange where traders buy and sell binary contracts based on future weather outcomes. Each contract resolves to YES ($1.00) or NO ($0.00) depending on whether a specific weather event occurs — for example, whether New York City's high temperature will exceed 50°F on a given day. The contract's trading price represents the market's implied probability of that event. Platforms like Kalshi, a CFTC-regulated US exchange, offer daily weather contracts covering temperature highs and lows, precipitation, snowfall, and wind speed across major US cities. Polymarket, a decentralized prediction market, also lists weather events using CLOB and USDC settlement. Weather markets are unique among prediction markets because numerical weather prediction models like GFS and ECMWF provide independent, objective probability estimates that can be systematically compared against market prices to identify mispriced contracts.

Question 2

What is edge in weather trading?

Accepted Answer

Edge in weather prediction market trading is the numerical difference between what forecast models predict and what the market prices. When GFS and ECMWF forecast models both indicate a 90% probability of New York City's high temperature exceeding 50 degrees Fahrenheit, but the Kalshi contract trades at 72 cents (implying 72% probability), that represents an 18-point edge. Celsi calculates edge scores by fetching real-time contract prices from Kalshi and Polymarket, querying GFS and ECMWF ensemble forecasts for the same weather event, computing model consensus probability, and scoring the divergence while adjusting for historical model bias specific to each city. Contracts with edge scores above 60 on Celsi's scanner typically represent the strongest mispricing signals. Paper trading these high-edge contracts first helps traders build intuition before risking real capital in weather prediction markets.

Question 3

What is the difference between GFS and ECMWF weather models?

Accepted Answer

GFS (Global Forecast System) is NOAA's primary global weather model with approximately 13 km grid spacing, 31 ensemble members, and free public access. ECMWF (European Centre for Medium-Range Weather Forecasts) operates at approximately 9 km resolution with 51 ensemble members and is widely considered the most accurate global model. Both update every 6 hours. ECMWF generally produces slightly more accurate temperature forecasts for days 1-3, with mean absolute error approximately 0.5-1.0°F lower than GFS. GFS tends to run warm for southern US cities like Miami and Houston. For precipitation, ECMWF's higher resolution and superior convective parameterization provide measurable advantages. When both models agree on an outcome, the consensus signal is strong and contracts trading far from that consensus represent high-confidence edge opportunities for weather prediction market traders.

Question 4

How does Celsi find mispriced weather contracts?

Accepted Answer

Celsi identifies mispriced weather contracts through a five-step process. First, it fetches current contract prices and implied probabilities from Kalshi and Polymarket. Second, it queries GFS and ECMWF forecast models including their full ensemble runs via Open-Meteo. Third, it calculates model consensus probability by weighting both models' deterministic and ensemble outputs. Fourth, it scores the divergence between market price and model consensus on a 0-100 scale. Fifth, it applies historical bias correction specific to each city and weather metric — for example, GFS consistently overpredicts Phoenix temperatures by approximately 2°F in summer. The resulting edge score accounts for raw divergence, model confidence (ensemble agreement), and known biases. Celsi covers 13 major US cities including NYC, Chicago, Miami, Denver, and more, tracking temperature, precipitation, wind, and snowfall contracts across both exchanges.

Question 5

How accurate are weather forecast models for prediction markets?

Accepted Answer

Weather forecast model accuracy varies by lead time, location, and variable. For day 1-2 temperature forecasts, both GFS and ECMWF achieve mean absolute errors of 2-4°F for most US cities, with ECMWF slightly more accurate. Accuracy decreases with lead time — day 3+ forecasts introduce substantially more uncertainty. Precipitation forecasting is less reliable than temperature, with both models struggling with exact timing and amounts. City-specific factors matter significantly: NYC benefits from dense observational data making models most accurate there, while Denver's complex terrain causes both models to struggle with upslope snow events. Celsi tracks rolling verification data comparing model forecasts against NWS observed outcomes for each city, calculating bias and accuracy metrics. For weather market trading, the most profitable signals come from day 1-2 contracts where model accuracy is highest and both GFS and ECMWF ensembles agree while the market price diverges significantly.

Question 6

What is paper trading on Celsi?

Accepted Answer

Paper trading on Celsi is a risk-free simulation system that lets users practice weather prediction market trading without committing real money. Users can take positions on any weather contract displayed in the scanner, tracking their picks against actual weather outcomes settled by NWS observed data. Celsi's paper trading portfolio tracks key performance metrics including hit rate (percentage of correct predictions), paper profit and loss, return on investment, and current win/loss streaks. The platform recommends completing at least 50 paper trades to build a statistically meaningful track record before risking real capital on Kalshi or Polymarket. Paper trading also feeds into Celsi's gamification features including leaderboard rankings across profit, accuracy, and streak categories, achievement badges, rank tier progression, and weekly tournaments. The free tier provides limited paper trading access, while Pro subscribers at $19 per month get full access to all cities and advanced analytics.