Is the weather market smarter than the forecast? We check every day.
The market’s implied high vs NOAA’s National Blend of Models vs the NWS point forecast, each scored against the official high Kalshi settles on — city by city, hour by hour, on rolling 30, 90 and 365-day windows.
Scored through · 13 cities · updates daily after settlement · Settlement tracker · Weather markets hub
Quick Answer
Over the last 90 days (117 scored city-days, 2026-10-31 to 2026-11-08), Kalshi’s ladder-implied daily high sat 0.96°F from the official high on average against 1.66°F for NOAA’s National Blend of Models — a ratio of 0.58 — and the market was closer than the blend in 12 of 13 cities. The night before alone: 1.10°F vs 1.74°F (0.64) on 117 city-days.
K/N is the market’s mean miss divided by the National Blend’s, over the same hours. Below 1 the market was closer; 0.78 means its miss was 78% of the blend’s. The archived 12-month measurement (Oct 27, 2025 – Oct 26, 2026, 97,912 hourly comparisons, 24 cities, run 2026-10-31): 0.78 the night before, 0.74 before 11 AM, 0.63 from 11 AM to 2 PM, 0.49 after 3 PM — under 1 in every city at every hour.
By city
Night-before K/N on each window, then the same-day ratio and the market against the NWS point forecast over the last 90 days. A dash means too little scored data in that window.
| City | Days | Night before · 30d | Night before · 90d | Night before · 365d | Same day · 90d | vs NWS · 90d |
|---|---|---|---|---|---|---|
| Atlanta, GAKATL | 9 | 0.47 | 0.47 | 0.47 | 0.45 | 0.45 |
| Austin, TXKAUS | 9 | 0.48 | 0.48 | 0.48 | 0.42 | 0.63 |
| Boston, MAKBOS | 9 | 0.44 | 0.44 | 0.44 | 0.32 | 0.41 |
| Chicago, ILKMDW | 9 | 0.74 | 0.74 | 0.74 | 0.49 | 0.82 |
| Dallas, TXKDFW | 9 | 0.91 | 0.91 | 0.91 | 0.80 | 0.91 |
| Denver, COKDEN | 9 | 0.70 | 0.70 | 0.70 | 0.78 | 0.70 |
| Los Angeles, CAKLAX | 9 | 0.45 | 0.45 | 0.45 | 0.69 | 0.38 |
| Miami, FLKMIA | 9 | 1.00 | 1.00 | 1.00 | 0.89 | 1.13 |
| New York City, NYKNYC | 9 | 0.64 | 0.64 | 0.64 | 0.66 | 0.50 |
| Philadelphia, PAKPHL | 9 | 1.40 | 1.40 | 1.40 | 1.02 | 0.88 |
| Phoenix, AZKPHX | 9 | 0.43 | 0.43 | 0.43 | 0.29 | 0.33 |
| San Antonio, TXKSAT | 9 | 0.54 | 0.54 | 0.54 | 0.53 | 0.65 |
| Seattle, WAKSEA | 9 | 0.60 | 0.60 | 0.60 | 0.61 | 0.50 |
By hour of day
The market sharpens through the day while the blend does not — it is watching the station. Last 90 days, every city pooled; same-day hours count only before the day’s high was reached.
| Window | Hours scored | Market miss | NBM miss | NWS miss | K/N | Market closer |
|---|---|---|---|---|---|---|
| Night before (9 PM local) | 117 | 1.10°F | 1.74°F | 1.75°F | 0.64 | 50% (35 ties) |
| Before 11 AM | 1157 | 1.04°F | 1.62°F | 1.63°F | 0.64 | 46% (391 ties) |
| 11 AM – 2 PM | 321 | 0.70°F | 1.73°F | 1.55°F | 0.41 | 59% (97 ties) |
| 3 PM – 9 PM | 58 | 0.60°F | 1.98°F | 1.31°F | 0.30 | 71% (13 ties) |
How it is measured
The market’s number is the whole ladder read as a distribution: each bucket’s mid-price is spread over the degrees it covers, each tail puts its price on the three degrees past its strike, and the implied high is the median. A ladder whose prices do not sum to roughly one is not scored.
NOAA’s number is the National Blend of Models station guidance (NBS, via the Iowa Environmental Mesonet archive): at each scored hour, the newest run at least two hours old, its high for the climate day being the larger of its daytime maximum and any 3-hourly temperature inside that day. The NWS number is the point forecast for the station’s own grid cell that our logger stored beside the price.
The answer is the official high the market settled on. A row is scored at 9 PM local the night before and at each local hour on the day while the station’s running maximum had not yet reached its final value. Misses are summed per city, day and hour window; every ratio on this page is a ratio of those sums.
The method is the one we ran on a year of Kalshi candles in October 2026; this page runs it daily on our own 5-minute price snapshots. What it does not say: anything about which side to take. The market being closer than the forecast is exactly why no forecast-driven side is published on Weather Edge.
Frequently Asked Questions
Is the Kalshi weather market more accurate than the NWS forecast?
On a year of data, yes, and by a wide margin: Kalshi’s daily-high ladder implied the official high with a 1.44°F mean miss at 9 PM the night before against 1.84°F for NOAA’s National Blend of Models, a ratio of 0.78, and the market was closer in every one of 24 cities. The ratio fell to 0.49 after 3 PM on the day. This page re-measures that every day on the cities we log, so the claim is checked rather than remembered.
How is the market’s implied high computed?
From the whole ladder, not one strike. Each bucket’s mid-price is spread over the degrees it covers and each tail puts its price on the three degrees past its strike; the result is a probability for every whole degree, and the implied high is that distribution’s median. A ladder whose prices do not add to roughly one (a one-sided or stale book) is not scored.
What is the National Blend of Models?
NOAA’s NBM is a statistically bias-corrected blend of dozens of weather models, issued for each station on a schedule. We read the station guidance (NBS text) from the Iowa Environmental Mesonet archive and, at every scored hour, use the newest run that had been out for at least two hours — the run a careful trader could actually have had in hand.
Why are same-day hours only scored before the high is reached?
Because after the thermometer has printed the day’s high the question is closed and the market is simply reading the sensor. Each same-day hour counts only while the station’s running maximum had not yet reached its final value for the day; the night-before row has no such gate.
Does this mean weather markets are unbeatable?
It means no forecast built from public data has beaten the price after fees in our testing — every disagreement strategy we ran lost out of sample. The remaining structure is in how the markets settle and how their books are made, which is why the free product of this lane is the settlement tracker and this scorecard rather than a forecast.
Related: what each weather market will settle at if nothing changes · weather markets hub and MCP server · the live board.