Trading Weather Markets on Polymarket: A Practical Guide

Weather markets are among the most transparent on Polymarket. The question is usually something like "Will the high temperature in New York City be above 70°F on November 13?" The resolution source is a public weather service. The data is free. The forecast models are available to anyone.

That transparency is why weather markets attract both careful traders and careless ones. A simple question does not mean a simple edge. The difference between profit and loss is often the resolution station, the time zone and the model's error distribution.

How a temperature market resolves

Most Polymarket weather markets resolve against a specific official source, such as the National Weather Service for US cities. The market rules specify:

• The station used for the reading.
• The variable measured (daily high, daily low, precipitation).
• The time window and time zone.
• The rounding convention.
• The source of record if the station reports late or changes.

Those details matter. A market that says "NYC" may use Central Park, LaGuardia or JFK. The three stations can differ by several degrees on the same day. A trader who assumes the wrong station can be right about the weather and wrong about the contract.

The resolution guide explains how to read the rules before trading.

The forecast edge

Weather forecasting has improved dramatically, but it is still probabilistic. The best models, such as the ECMWF and GFS ensembles, produce a distribution of possible outcomes, not a single number.

A useful workflow is:

1. Pull the ensemble forecast for the relevant station and date.
2. Convert the distribution into a probability for each temperature bucket.
3. Compare that probability with the market price.
4. Trade only when the gap is larger than the spread and your model error.

For example, suppose the ensemble gives a 35% chance that the high will be 70-74°F, but the market prices that bucket at 25 cents. If your model is well calibrated, the expected value is positive: 0.35 × $1 minus $0.25 = $0.10 per share, before costs.

The hard part is calibration. A model that is overconfident will produce false edges. Backtesting against historical forecasts is essential.

Why weather markets are not free money

Three risks dominate.

First, model error. The forecast can be wrong, and the market can be right. A 35% probability does not mean the outcome will happen. It means it should happen about 35% of the time over many similar trades.

Second, resolution risk. A station outage, a late report or a change in the source can create ambiguity. The UMA dispute article explains how contested resolutions work.

Third, manipulation and tampering. In 2026 the French weather service alerted police to possible tampering with sensors after suspicious Polymarket bets, according to the Financial Times. That case is a reminder that weather markets can create incentives to interfere with physical infrastructure. The manipulation cases article covers similar patterns.

How to read an ensemble forecast

An ensemble forecast runs the same model many times with slightly different initial conditions. The spread of the results is a measure of uncertainty. A tight cluster around 72°F means the forecast is confident. A wide spread from 68°F to 78°F means the outcome is uncertain.

To convert that into a probability, count the share of ensemble members that fall into each bucket. If 30 of 100 members land in the 70-74°F bucket, the model's implied probability is about 30%. Compare that with the market price, not with the single deterministic forecast.

The method is not perfect. Ensembles can be biased, and local effects such as sea breezes or urban heat islands can shift the outcome. That is why the resolution station matters as much as the model.

Liquidity and timing

Weather markets are often thin. The spread can be wide, especially in the hours before the observation. A trader who enters early may get a better price but take more model risk. A trader who waits for the observation may find no liquidity.

The best windows are usually after the latest model run and before the market becomes crowded. The liquidity and slippage guide explains how to measure the cost.

A simple strategy for a beginner

Start with one city and one variable. New York or London temperature is a good choice because the data is reliable and the markets are relatively liquid.

1. Read the contract rules and identify the exact station.
2. Follow the ensemble forecasts for a month without trading.
3. Record your predicted probability and the market price for each bucket.
4. After a month, compare your predictions with the outcomes.
5. Only trade when you have a documented calibration edge.

This is slow. It is also the only way to know whether your edge is real.

A useful discipline is to backtest the model against at least one full year of observations before risking money. If the ensemble's 30% buckets happened 20% of the time, your probabilities are overconfident. If they happened 40% of the time, they are underconfident. Either way, the market may already know.

The broader lesson

Weather markets are a good example of the general rule for prediction markets: the edge comes from a specific, verifiable information advantage, not from a vague feeling about the weather. The same is true for sports, elections and crypto.

If you enjoy weather, the market can be a useful place to apply that interest. If you are looking for a quick profit, the market will happily take your money. The how to make money guide explains where durable edges come from.