How often our pre-match analysis picks the correct outcome (home win, draw, or away win), broken down by the confidence band we publish on each Edge Card. Last 90 days.
Same measure, split by competition — both pipelines shown. Leagues under 30 resolved matches stay calibrating.
Every upcoming match gets a probability split (home/draw/away) and a confidence band — HIGH, MED, or LOW — set by sample size and the gap between the top two probabilities. After full time, we record whether our top pick matched the actual result.
Bands stay marked "calibrating" until they hit 30 resolved matches — small samples produce noisy numbers and we'd rather show nothing than mislead. Postponed and abandoned matches are excluded.
This is a transparency tool, not a tipping service. We don't publish odds, recommend stakes, or link to bookmakers. Read about how the model works.