It is 3 p.m. in Madrid on a July afternoon. The airport reports have looked flat for a while, so you read the highest one so far as the day's number. On 30 of the 31 July days in my records, the settled high finished a whole degree or more above it.
Make the same reading at 3 p.m. in Taipei and it held on 261 of 266 days. "The high comes in mid-afternoon" is true on average and hides how different airports are. So I counted it, airport by airport, hour by hour.
What I counted
49 airports named in the rules of Polymarket daily-high temperature markets. Every complete local day from November 2025 to mid-November 2026. That is 18,098 airport-days.
For each day I rebuilt the settled high the way the markets count it: routine airport reports only, the airport's local calendar day, the 0.1 °C remark value where one exists, converted to °F first for °F markets, rounded half up to a whole degree.
Then, for each local clock hour, one question: did the full day settle above the whole-degree value that the reports up to that hour had reached? If yes, the day "still rose" after that hour.
No prices are involved. I only read airport reports.
All 49 airports together
Share of airport-days where the settled high still rose after the hour:
- Noon: 74.5%
- 2 p.m.: 39.2%
- 3 p.m.: 21.5%
- 4 p.m.: 9.3%
- 5 p.m.: 3.5%
- 6 p.m.: 1.6% (294 days)
- 9 p.m.: 0.7% (130 days)
So at 3 p.m. the settled number was already final on 78.5% of days. That one figure covers a very wide spread.
The spread between airports and seasons
At 3 p.m., over the whole year:
- Under 10% of days still rose at 17 airports. Taipei 1.9%, Qingdao 2.2%, Panama City 2.5%, Shanghai 2.7%, Los Angeles 3.0%, Tel Aviv 3.0%.
- In the middle: Tokyo 10.2%, London 25.3%, New York (LaGuardia) 39.5%.
- 40% or more at 7 airports. Ankara 41.4%, Paris 44.4%, Seattle 51.8%, Atlanta 53.0%, Austin 57.7%, Dallas 58.9%, Madrid 68.0%.
Two things in the data go with this spread.
The hours are clock hours, not sun hours. Worked out from each airport's longitude, the sun is highest at about 2:20 p.m. clock time in Madrid in July, and at about noon in Taipei. So "3 p.m." in Madrid is early in the solar afternoon. In July the share still rising was 30 of 31 days at 3 p.m., 10 of 31 at 5 p.m., and 0 of 31 at 6 p.m.
The unit matters too. The 11 US airports settle in whole °F, a step a little over half a degree Celsius, so a smaller rise changes the settled number. At 3 p.m., 35.4% of days still rose at the °F airports and 17.5% at the °C airports. That gap mixes unit and geography. I have not separated the two.
Season moves the same airport a long way. Austin still rose after 3 p.m. on 29 of 31 August days and on 1 of 37 November days. Madrid went from 30 of 31 in July to 6 of 31 in December. Across all airports, the 3 p.m. share was 30.4% in July and 7.8% in November.
Late highs are a winter pattern
After 6 p.m. the settled high still rose on 1.1% of days from February to November. In December and January it was 4.1%.
It is concentrated in a few northern airports. Helsinki and Chicago each had 12 such days out of 62 in December and January (19.4%). Amsterdam had 10. London, New York and Moscow had 9 each. On those days the highest report came in the evening or at night, not in the afternoon.
At the other end, 15 airports had no day in the whole year that rose after 6 p.m. Singapore, Miami, Mexico City and Denver are among them.
What this does not tell you
It is one year. A single airport-month is about 30 days, so "29 of 31" is a count, not a law. Next August can differ.
Hours are local clock time, and only complete days are counted (at least 18 counted reports).
The year of history is my own rebuild from public reports. As a check, I recounted 8 of the 49 airports from the raw reports with separately written code, and all 2,208 airport-month-hour cells matched. The comparison against actual market results is shorter: at this week's check it covered 17 October to 16 November 2026, and 1,153 of 1,153 city-days fell in the range the market resolved to. Every city-day is published, including any day that differs:
https://isotherm-82f.pages.dev/check.html
If you want to run the hour-by-hour count yourself, there is a free single-file Python script that prints it for a 30-day Tokyo sample:
https://isotherm-82f.pages.dev/kit.html
This is information about how numbers are counted. It is not financial or betting advice, and ISOTHERM is not affiliated with any exchange or prediction market.


