About

How we keep score on ourselves.

IdeaForge runs a public calibration ledger. Every analysis emits falsifiable predictions; founders who opt in publish them to /predictionsfor anyone to vote on. When a prediction resolves, we score both the engine's confidence and the crowd's yes/no consensus against the actual outcome. The metric is Brier — lower is better, capped between 0 and 1.

Engine Brier

—

0 resolved · IdeaForge analyses' own confidence vs reality

Crowd Brier

—

0 resolved · Anonymous voters on the same predictions

Not enough resolved predictions to compare yet. The ledger fills as analyses age past their 30/60/90-day horizons.

Where the scores live

/predictions — live board with open and resolved predictions.
/graveyard — outcome ledger of analyses at 30/60/90 days.
/anti-portfolio — where the engine was wrong (and where it was right).
/leaderboard — opt-in founder rankings on calibration.

Last computed: 6/13/2026, 4:53:31 AM. Cache: 30 min.