Whether you're new to weather trading or prediction markets in general, this glossary covers every term you'll encounter.
Market Terms
Binary contract: A contract that resolves to YES ($1.00) or NO ($0.00) based on whether a specific event occurs.
Implied probability: The market price of a YES contract expressed as a percentage. A contract trading at $0.72 implies a 72% probability.
Settlement: When a contract expires and pays out based on the actual outcome. Weather contracts on Kalshi settle using NWS observed data.
Liquidity: How easily you can buy or sell a contract without moving the price. Higher liquidity = tighter spreads = better execution.
Spread: The difference between the best bid and best ask price. A contract with bids at $0.68 and asks at $0.74 has a 6-cent spread.
Forecast Model Terms
GFS (Global Forecast System): NOAA's primary global weather model. Updated 4 times daily. Free and publicly accessible.
ECMWF (European Model): Operated by the European Centre for Medium-Range Weather Forecasts. Generally considered the most accurate global model.
Ensemble: Multiple runs of a model with slightly different initial conditions. GFS runs 31 ensemble members, ECMWF runs 51. The spread of ensemble members indicates forecast confidence.
Deterministic run: The single "best guess" forecast from a model, as opposed to the full ensemble.
Model consensus: The average or agreed-upon forecast when multiple models are considered together.
Lead time: How far in advance a forecast is made. A "day 1" forecast predicts tomorrow; "day 3" predicts three days out. Accuracy decreases with lead time.
Edge & Analysis Terms
Edge: The difference between model consensus probability and market-implied probability. Positive edge suggests the market is underpricing the event.
Edge score: Celsi's proprietary 0-100 score combining raw divergence, model confidence, and bias adjustment.
Model bias: A systematic tendency for a model to over- or under-predict a variable. Measured as the average signed error over time.
MAE (Mean Absolute Error): The average absolute difference between predicted and observed values. Lower = more accurate.
RMSE (Root Mean Square Error): Similar to MAE but penalizes larger errors more heavily. Useful for identifying models with occasional big misses.
Bias correction: Adjusting model output based on known systematic biases for a specific location and variable.
Divergence: The raw point difference between model consensus and market price, before bias correction.
Weather Observation Terms
NWS (National Weather Service): The official US weather observation authority. Their observed data is used to settle Kalshi weather contracts.
Observed high/low: The actual maximum or minimum temperature recorded at the official station for a given day.
Trace precipitation: A measurable but extremely small amount of precipitation, typically < 0.01 inches. Important for contracts with "any precipitation" thresholds.
ASOS/AWOS: Automated Surface Observing Systems at airports that provide the official temperature and precipitation readings used for contract settlement.