Tornado Prediction Markets vs. NOAA: Preliminary Counts, Final Counts and How Kalshi Settles

By Arthur Angelo, PredictionScout Research · Updated Oct 15, 2026, 7:59 AM ET

Key findings

• Start-of-month forecast: median miss 44 tornadoes. Over 20 months, the market’s implied median on the first day of the month was off by a median 39% of the count. Threshold contracts went the way the market leaned 138 of 171 times (81%).
• Preliminary vs. final: 3,571 vs. 3,504. Preliminary counts come from storm reports filed within hours; NWS offices then survey damage and merge or drop reports. Final numbers appear in NOAA’s Storm Data and the SPC tornado database, currently complete through 2025. Largest gaps: May 2025: 329 preliminary, 278 final; Dec 2024: 90 preliminary, 131 final; Mar 2025: 268 preliminary, 241 final.
• 10 months would settle differently on final counts. Dec 2023 (settled 24, final 21); Feb 2024 (settled 48, final 45); Oct 2024 (settled 50, final 54); Nov 2024 (settled 39, final 60); Dec 2024 (settled 90, final 131). Kalshi settles on the preliminary number and markets are not reopened.
• Settlement check: 6 of 35 months match the sum of SPC’s daily filtered reports exactly. The daily report files are a close proxy for the preliminary monthly count, not the identical source. September 2026: Kalshi settled on 55; the daily filtered reports sum to 57.

Preliminary vs. final tornado counts

The National Weather Service publishes a preliminary tornado count from local storm reports within hours. Reports can double-count a single tornado seen by several people, or miss tornadoes found later in damage surveys. Months later, NOAA publishes the final count in Storm Data, which is what climate records use. Kalshi’s rules settle on the preliminary count, so a market can pay out on a number the official record later changes.

Month by month

* Some daily report files were unavailable, so the sum is incomplete. Final counts: SPC tornado database (one row per tornado, full track).

Download the data

Tornado markets vs NOAA counts (CSV)

One row per settled market period. Per-contract price histories are not included. Free to reuse under CC BY 4.0 with a link back to this page.

Methodology

• Markets. Kalshi’s monthly tornado series (KXTORNADO, earlier TORNADO- events): “above X tornadoes” contracts, from Kalshi’s public API. Prices are hourly bid/ask midpoints (or last trade) at midnight ET on the first day of the month; the implied median is where the chance of “above X” crosses 50%.
• Preliminary counts. Kalshi’s recorded settlement value, cross-checked by summing SPC’s daily filtered tornado report files for the month. SPC’s monthly preliminary page (Kalshi’s stated source) is generated in the browser and is not read directly.
• Final counts. The SPC Severe Weather Database “actual tornadoes” file, counting one row per tornado (full-track record).
• Limits. Final counts lag by many months, so recent months can only be compared with preliminary data; storm-report days run 12Z to 12Z, not calendar days. Nothing here is investment advice. Last run: Oct 15, 2026, 7:59 AM ET.

Data sources: prices and volumes come from the public APIs of Kalshi and Polymarket, with tornado data from the NOAA Storm Prediction Center. All trademarks belong to their owners. PredictionScout is independent and not affiliated with any exchange. Nothing here is investment, betting, or legal advice.

Related: earthquake markets · markets vs official numbers · hottest-year markets · All research