How to read a Kalshi temperature market as a forecast

A weather trader looks at tomorrow's high for a city and asks one question: is the market's forecast too warm, too cold, or too confident? Kalshi's daily temperature markets make that question answerable, because each day's prices form a complete forecast you can read and grade.

#How the market is structured

Each city and day is one event. The event lists a ladder of contracts on the day's high temperature: an open bottom tail ("65° or below"), a run of 2°F brackets ("66° to 67°", "68° to 69°", …) and an open top tail ("74° or above"). Exactly one contract settles YES.

Settlement uses the official reported temperature for the city's designated primary station. On Kalshi the source agencies are, in order, The Weather Company and then the National Weather Service. The first official non-preliminary report counts; later revisions do not. The measured day runs midnight to midnight local standard time, so trading stays open into the night and a late-evening reading can still set the high.

#Step 1: Turn the ladder into probabilities

Take a mid price for each bracket: the average of the YES bid and ask, or the ask alone when there is no bid. The mids will not sum exactly to $1.00, because of the bid-ask spread. Divide each by their sum so the probabilities add to 100%.

Here is a hypothetical ladder:

In a spreadsheet, with mids in column B:

Probability   =B2/SUM($B$2:$B$7)
Cumulative    =SUM($C$2:C2)

#Step 2: Read the forecast

• Most likely bracket: the largest probability, here 68° to 69° at 37%.
• Median: the bracket where the cumulative probability first passes 50%, again 68° to 69°.
• Range: the brackets where cumulative probability passes 10% and 90% give a rough 80% range, here 66° to 73°.

A wide range means the market is uncertain, for example ahead of a front. A narrow one means it is confident. Both can be wrong, which is what the next step measures.

#Step 3: Check the market against reality

Collect thirty or more past days for the same city. For each day, record the ladder's prices at the same morning time, compute the median and the 10–90 range as above, and note the bracket that actually won.

Then report three numbers:

• Bias: the average of (market median − actual high). A persistent positive number means the market ran warm.
• Average miss: the average absolute difference.
• Coverage: the share of days the actual high fell inside the market's 10–90 range. A well-calibrated market lands near 80%. Much lower means it was overconfident.

The snapshot time matters more than anything else. A price taken at 4pm, after the day's peak, will look superb because the answer is already mostly known. Fix one time before the information you want to test arrives, and use it every day.

#Pitfalls

• Correlated contracts. The six brackets on one day are one observation, not six. Judge sample size by days.
• Seasonal grids. The bracket boundaries move with the season. Compare results by distance from the market's median, not by absolute temperature.
• Station mismatch. A forecast for an airport or a city-wide average is not the settlement station. Compare like with like.
• Thin brackets. Far-out brackets trade at a cent or two with wide spreads. Their mids are noise, not tail probabilities.

#Run it in Jordan

"Which NYC temperature is Kalshi betting on tomorrow?" is one of the starting questions in Jordan. Jordan loads the ladder onto a sheet with each bracket's strike values, normalizes it with formulas, charts the distribution, and can collect a month of past days priced at a fixed morning time to compute bias and coverage. Every number is a formula you can inspect. For the general method, see Market calibration and Prediction market prices as probabilities.

#Frequently asked questions

#What decides how a Kalshi temperature market settles?

The official reported high (or low) for the city's designated primary weather station, from The Weather Company and then the National Weather Service, using the first non-preliminary report. Revisions after expiration do not change the result.

#Why do the bracket prices add up to more than $1?

Each contract is quoted with a bid-ask spread. Buying every bracket at the ask costs more than $1, which is the market's margin. Mid prices sum closer to $1, and normalizing removes the remainder.

#How many days do I need to judge the market?

Coverage near 80% over thirty days could still be luck, so treat a month as a first look. Several months, split by season, give a steadier read.