← All marketsGet Poly — free

Accuracy of the AI Council’s calls — from the first session

The Council against the crowd on the same questions, counted from the open files since the first scored session.

Oct 19, 2026 – Oct 27, 2026: the Council’s side was right on 81.8% of 33 crowd questions; the crowd’s on 84.8%.

Oct 19, 2026 – Oct 27, 2026: the Council’s Brier skill against the crowd is −0.41 (95% interval −0.63 to −0.19), 7 scored sessions.

7 scored sessions · data as of Oct 27, 2026

Sealed before the open: level 2+ (two independent RFC 3161 timestamps)

Proof of Oct 27, 2026

The first line follows stages fixed in advance

  1. Under 5 scored sessions: the count — right sides of the Council and of the crowd on the same crowd questions, the number of sessions and the date the interval comes.
  2. From 5 scored sessions: the Council’s Brier skill against the crowd with its 95% interval by whole sessions.
  3. From 20 comparable questions: the shares of right sides in percent, beside the Brier skill and its interval. ← now

The stages are in force from Oct 22, 2026; they are not changed after the fact. The window always starts at the first scored session.

Open data

council-stats.json · CSV · JSONL · Method and conflicts →

An AI experiment — the same issue for every reader, not personal investment advice.

The scorecard on its own page →