Matches
Every fixture, with each analyst's pick published about an hour before kickoff.
Every analyst published a winner pick and a stated confidence about 75 minutes before kickoff, the betting-market closing-line horizon. Predictions are scored on the Brier score and expected calibration error (ECE), where lower is better. Calibration, not raw accuracy, is the measured axis.
The Council architecture recorded the lowest expected calibration error of any machine tested, ECE 0.100 over 104 scored matches, ahead of the betting market at 0.104, the Pipeline at 0.112, and the Solo at 0.114. The result is directional: the leaders' bootstrap intervals overlap, a disclosed limit a second tournament would sharpen, not a null.
The business bridge, what inocta.io calls the IO Gap: football forecasts land near the betting market because a liquid market already solves football. A company has no such market — there is no closing line for a Q3 pipeline, a churn probability, or a deal-close likelihood. So a calibrated AI forecast IS the reference the operator otherwise lacks.
Every 2026 World Cup match with eight pre-kickoff predictions: three AI architectures, the betting market, an ELO rating and one human. Confidence, reasoning, risk factors and cost per call, locked before kickoff.
Every fixture, with each analyst's pick published about an hour before kickoff.
Every fixture, with each analyst's pick published about an hour before kickoff.
By Mo Kahlain · Published
Matches
Every fixture, with each analyst's pick published about an hour before kickoff.