Trust scores are evidence-weighted,not just hit rates.
Accura separates measurable predictions from commentary, applies confidence labels, and keeps unresolved claims out of the score until the deadline passes.
Clear claim
A prediction must have a source, date, subject, expected outcome, and preferably a timeframe.
Outcome review
Resolved claims are labeled verified, partly true, missed, pending, or too vague.
Confidence layer
Evidence quality, sample size, and category difficulty change the score impact.
Safety review
Early public scores stay Preliminary until a human review lands.
Score formula inputs
Resolved prediction accuracy
Prediction specificity
Difficulty and baseline probability
Clear timeframe
Citation and source quality
Sample size reliability
Recency weighting
Penalty for vague confident claims