The question
Polymarket asked a straightforward weather question: what would be the lowest temperature in Tokyo on May 16? The market resolved to a specific degree reading, so the council’s job was to identify the most probable bucket before the day fully played out.
The report was filed on 2026-05-15. Models that participated were Gemini (chairman), Claude, and GPT.
What the council concluded
The formal action was SKIP, with 99% confidence and a stated council probability of 0.0% on taking a position. Internally, the panel still named a most-likely outcome: “Will the lowest temperature in Tokyo be 17°C on May 16?” at 80% probability. That most-likely call is what this grade is scored against.
The reasoning was time-zone and observational, not a long-range forecast:
“Council recommends SKIP. Confidence: 99%. At 22:11 UTC on May 15, it is already early morning on May 16 in Tokyo (JST). The overnight low has already been established at around 17°C (62-63°F). Because the daily low is already well below 20°C,”
In plain terms: by the time the council met, Tokyo’s calendar day had already begun, the overnight minimum had already printed near 17°C, and further cooling looked unlikely. The information edge was real—but it arrived after the low was effectively set, which is why the panel refused to treat the remaining market as a clean trade.
How the council framed the odds
The graded pick carried an 80% probability on the 17°C outcome. No separate live Polymarket mid-price is recorded in the report packet, so the honest comparison is internal: the models converged on 17°C as the modal answer while simultaneously arguing that the timing made a new position unattractive. Action and probability were decoupled on purpose—high conviction on what would resolve, low conviction that betting still made sense.
What actually happened
The winning outcome was exactly the council’s most-likely pick: the lowest temperature in Tokyo on May 16 was 17°C. Graded on that most-likely basis, the council was correct. The Brier score was 0.04—tight calibration for an 80% call that hit.
Takeaway
This is a clean illustration of when SKIP is the right institutional move even though the directional read is good. The panel used local clock time and already-observed overnight data to lock 17°C as the modal outcome, then refused to chase a market whose resolving fact was largely in the past. Correct call, strong Brier, no forced trade: the kind of boring discipline that keeps a public grade log honest when the next weather market is still genuinely open.
AI-generated analysis for informational purposes only. Not financial advice.