The Council, Graded
Every prediction our AI council makes is graded against the real outcome — and published here, wins and losses alike. See the full track record.
AI council assigned 1% to Iran ending uranium enrichment by July 31 — NO won
On a tightly defined Polymarket question about Iran publicly pledging to end all uranium enrichment by July 31, 2026, the council unanimously favored NO at 1% probability and was correct. The near-perfect Brier score underscored how resolution criteria, not headlines, decided the outcome.
Council skipped LEC Versus 2026 at 100% confidence but nailed G2 Esports as winner
The AI council issued a SKIP on the LoL LEC Versus 2026 Winner market yet assigned 100% to G2 Esports as most likely; G2 did win, producing a perfect Brier score of 0.0.
Council backed ‘no change’ at 64% on Bank of Russia’s July rate — they cut instead
The Prediction Council assigned 64% to the Bank of Russia holding its key rate in July 2026 and recommended BUY_YES. The central bank decreased the rate instead, producing a clear miss and a Brier score of 0.4096.
Council put 60% on a Group H nation winning the 2026 World Cup — Group H delivered
With the final hours away, the AI council skipped a bet but correctly identified Group H as the most likely origin of the champion at 60%. The pick resolved true for a Brier score of 0.16.
Council named Spain most likely at 17% for 2026 World Cup — Spain won
The AI council’s top pick in the 2026 World Cup winner market was Spain at 17%. Spain took the title, validating the ranking even as the Brier score reflected how open the field was.
Council assigned 100% to a completed Cattolica match — and was exactly right
All four models independently verified that the Roberto Carballes Baena vs Francesco Forti match had already finished, leading to a SKIP with a perfect Brier score of 0.0 on the graded outcome.
The AI council gave Japan 64% against Tunisia — Japan won
The council skipped a Tunisia vs. Japan market at 64% probability for a Japan win due to key absences, and Japan did win. Their most-likely call proved correct with a Brier score of 0.1296.
The AI council gave Türkiye 48% against Paraguay — Paraguay won
The council skipped Türkiye vs. Paraguay at 48% with 75% confidence, citing no edge over market odds. Paraguay won, and the graded most-likely pick was wrong.
England vs. Croatia: Council correctly flagged a settled 100% market and skipped
After England’s win was already confirmed, the council assigned 100% to the result, recommended SKIP, and posted a perfect Brier score of 0.0 on the graded outcome.
AI council skipped Mexico vs Korea Republic at 51% — Mexico won anyway
The council declined a position on Mexico’s home match against Korea Republic, calling the edge too thin at roughly coin-flip odds. Mexico still won, validating the lean but not the decision to sit out.
The AI council gave Brazil 62% to beat Morocco — the match ended in a draw
The Prediction Council skipped the Brazil vs. Morocco market after aligning with the ~58% favorite price, but its most-likely call of a Brazil win proved wrong when the sides drew. A clean miss on the graded pick.
Council favored 26°C for Seoul’s June 13 high at 25% — actual high was 27°C
The AI council’s most-likely pick of 26°C for Seoul on June 13 missed by one degree; the verified high was 27°C. A clear miss on a tightly clustered weather market.
AI council picked 33°C as Manila’s June 8 high — the actual high was 32°C
The council’s most-likely call of 33°C for Manila on June 8 missed by one degree. Real-time METAR evidence and converging forecasts still produced a wrong modal outcome.
AI Council Picked 56-57°F as Certain for LA’s June 5 High — Actual High Was 70-71°F
The council assigned 100% probability to a 56-57°F high in Los Angeles on June 5 and was graded wrong when the temperature landed in the 70-71°F bucket. A Brier score of 1.0 records the complete miss.
Council correctly flagged 17°C as Tokyo’s May 16 low — and skipped the trade
With the overnight low already locked in, the council’s most-likely pick of 17°C proved right (Brier 0.04), but the panel still recommended SKIP rather than a late bet.
AI council favored 36°C+ in Karachi at 30% — the high was 34°C
The Prediction Council backed a 36°C-or-higher outcome for Karachi on May 7 based on a PMD heatwave forecast. The actual maximum was 34°C, making the graded pick incorrect.
Council skipped Wellington's May 7 high after backing 18°C — 19°C won
The AI council unanimously recommended SKIP on Wellington's May 7 high, citing forecasts of 17–18°C and naming 18°C as most likely at 45%. The actual high was 19°C, and the graded pick was wrong.
Council skipped Chengdu’s May 6 high-temperature market — 28°C or higher still won
The AI council issued a high-confidence SKIP on Chengdu’s May 6 temperature market after forecasts clustered between 26–31°C. Their most-likely pick of 28°C or higher still matched the outcome, with a Brier score of 0.3025.
Council correctly called 32°C as Singapore’s May 5 high
The Prediction Council’s most-likely pick of 32°C for Singapore’s high on 5 May 2026 was correct. Graded on that selection, the 41% probability produced a Brier score of 0.3481.
Council assigned 0% chance to a 28°C high in Kuala Lumpur on May 2 — NO won
All four models unanimously bought NO on a low high-temperature outcome for Kuala Lumpur, citing clear forecast data. The call was correct with a perfect Brier score of 0.0.