The market question
The Polymarket event asked a narrow weather question: what would be the highest temperature in Shenzhen on August 24? Resolution did not depend on local media or airport screenshots. It hinged on a single official reading—the highest whole-degree Celsius value shown in the NOAA/NWS WRH Time Series Viewer for station ZGSZ (Shenzhen Bao’an International Airport) on calendar date 24 August 2026, in metric units. That made the market a pure forecast of one station’s daily maximum, not a city-wide “feel” or an average of multiple sensors.
What the council concluded
The report was issued on 2026-08-23, the day before the observation window. All four models—Gemini, Grok, Claude, and GPT—participated, with Grok acting as chairman. The council’s formal action was recorded as BUY_NO, with an implied probability of 30% and confidence of 68%. Its most-likely single outcome, and the pick later used for grading, was that the high would be 31°C, assigned 37%.
The published reasoning stayed tightly scoped to the resolution source:
“Council recommends BUY_NO (NO). Probability: 30%, Confidence: 68%. Resolution is determined solely by the single highest whole-degree Celsius temperature appearing in the NOAA/NWS WRH Time Series Viewer for station ZGSZ (Shenzhen Bao’an International Airport) on calendar date 24 August 2026 (metric units).”
In short, the panel treated the problem as a station-specific daily max rather than a broader climate narrative, concentrated probability on the mid-30s band, and put the single largest mass on 31°C.
How the council framed the odds
Within the council’s own distribution, 31°C was the mode at 37%. No other outcome was listed as higher. The BUY_NO action and the 30% probability figure sat alongside that mode, reflecting a view that concentrated “yes” claims on any one exact degree were still more likely wrong than right—typical for discrete temperature ladders where adjacent bins compete. Confidence at 68% signaled moderate conviction in the overall shape of the distribution, not certainty on the exact bin.
What actually happened
The market resolved to exactly the council’s most-likely call: the highest temperature in Shenzhen on August 24 was 31°C under the ZGSZ NOAA/NWS rule. Graded on the most-likely basis, the council was correct. The Brier score on the graded pick was 0.3969—respectable for a multi-outcome temperature market where a 37% favorite leaves substantial residual probability on neighboring degrees, but still a clear hit on the outcome that paid.
Takeaway
This was a clean verification case. The council did not overclaim precision; it named 31°C as the single most probable high, documented the exact resolution series, and let the official station reading decide. When that reading landed on 31°C, the graded pick matched reality. Weather markets this granular remain noisy day-to-day, and a Brier score near 0.40 on a ~37% call is a reminder that even a correct mode is not a high-certainty lock. The useful discipline here was sticking to the stated data source and resisting the urge to invent a sharper story than the forecast distribution supported.
AI-generated analysis for informational purposes only. Not financial advice.