How we measure who you can trust

What goes into the Score

The Score combines four things: accuracy (how many resolved predictions turned out right), difficulty (calling something obvious is not the same as calling something risky), calibration (whether someone who says they are very confident tends to be right) and consistency (holding up results over time, not a single lucky streak).

Minimum sample

An analyst needs at least 5 resolved predictions to have a Score. With fewer, there is not enough of a track record to judge and no number is assigned. The more resolved predictions, the more reliable the figure.

That reliability rises in steps, not continuously: 5, 10, 20 and 50 resolved. Between two steps the Score barely moves even as predictions close, because the only thing changing is the average of the grades. That is by design: a small sample should not swing wildly.

Misses count

Failing with high conviction penalises more than failing with low conviction; being right and well-calibrated adds. Failed predictions do not disappear: they are part of the track record and of the Score.

Anti-manipulation

The system penalises duplicate predictions and contradictions (predicting both a rise and a fall on the same asset to "be right no matter what"). The goal is for the Score to reflect real judgement, not tricks.