Market field guides / Weather & Climate Markets

🌦 Weather & Climate Markets

Markets settled by a specific thermometer at a specific station. The forecast is public; the conventions are the edge.

The honest overview

Weather markets look like the fairest fight on Kalshi: the settlement data comes from government stations, the forecasting models are public, and no insider knows tomorrow's high before the atmosphere does. That surface fairness hides a game of conventions, which station, which time window, which rounding, that decides a surprising share of outcomes.

The honest structure of the category: within about three days, modern forecasts are excellent and prices track them tightly, so edges are thin and convention-driven. Beyond five to seven days, forecast skill decays toward climatology, and prices that imply confidence at that horizon are usually implying too much.

How these markets actually work

Every market names its exact observation station and time window. A city's 'high temperature' means the reading at one specific station (often the major airport), on the measurement day the rules define, not what your phone's weather app, averaging different sources, shows for the metro area.

Thresholds settle on the official recorded value at the stated precision. A market on '90 degrees or above' is decided by whether the station logs 90, and days that feel like near-misses are decided by instrument and convention, not vibes.

Precipitation markets carry their own conventions: trace amounts, the measurement window's clock (local standard time matters), and which gauge counts. Rules first, forecast second.

Where real edges come from

Model literacy is the honest edge. The major global and regional models publish on known schedules several times daily, and their agreement or divergence is readable by anyone willing to learn. When models converge and the market still prices yesterday's uncertainty, the reprice is predictable in direction.

Climatology is the base rate casual money skips: how often this station has historically hit this threshold on this calendar date. For horizons past a week, climatology plus a trend beats any point forecast, and prices set by point-forecast confidence at those horizons are the mispricing.

Station quirks persist: some stations read systematically warm or cool versus the metro perception, coastal stations behave differently from inland ones, and the person who has watched one station's behavior for a season prices its markets better than the person forecasting 'the city'.

Red flags

Conditions where the trade is usually worse than it looks. Any one of these firing is a reason to pass.

🚩 Trading your weather app against the settlement station
Your app blends sources for a metro area; the market settles on one station's instrument. The three-degree gap between them is the whole trade, and it is against you if you never checked which station rules.
🚩 Confidence past day seven
Forecast skill at long horizons decays toward climatology, and a ten-day point forecast is closer to a coin flip against a threshold than the confident graphic implies. Prices that reward long-horizon confidence are usually selling it to you.
🚩 Betting the near-miss threshold on feel
Whether today tops out at 89 or 90 is decided by one instrument's afternoon. Threshold markets within a degree of the forecast are bets on measurement noise, and the fee eats coin flips.
🚩 Ignoring the clock convention
Measurement days and windows run on defined local times. Storms that straddle midnight, or daylight-saving quirks, land precipitation in a different market day than your intuition files it.

Orange flags

Proceed only after you have checked the specific thing named.

Station outages and substitutions
Instruments fail. The rules define what happens when the named station has missing data; know it before an event where outages are plausible.
Model disagreement
When the major models diverge meaningfully, the honest probability is wider than any single run. Proceed only if the price reflects the spread, not one model's confidence.

Green lights

The conditions under which taking the trade is actually defensible.

You checked the station, the window, and the climatology
You know exactly which instrument settles it, what the calendar base rate is, and what the current model consensus says, and the price disagrees with that stack by more than the fee.
Models converged, the price hasn't
Successive model runs agreed and tightened while the market still shows the old uncertainty. You are trading the reprice, not the weather.

The tools that pair with this

Settlement Calendar Weather markets close daily; see the week's board at a glance.Strategy Lab How weather-category price zones have actually settled, measured.Trade Checker Near-threshold weather trades live or die on the fee math.

Questions traders actually ask

What decides a Kalshi weather market?
The official reading at the specific observation station and time window named in the market rules, at the stated precision. Not your weather app, which blends different sources across a metro area.
How far out are weather forecasts actually reliable?
Skill is strong inside three days, meaningful to about a week, and decays toward climatology beyond that. Long-horizon threshold markets are closer to base-rate bets than forecast bets, and pricing them requires the historical rate for that station and date, not a ten-day graphic.

Field guides are educational and describe historical patterns and mechanics; nothing here is a recommendation to trade any market. Rules quoted generically; the specific market’s rules page always governs. Not financial advice.