Every published pick gets graded after the match. For each confidence band we show what we said (the label) and what actually hit (the realised rate). When the second bar matches or exceeds the first, we under-promised. When it falls short, we over-promised.
Match winner, player props, period markets — every market we graded, in the order it settled. No filters, no cherry-picks.
ECE is computed on bands with at least 10 settled picks so a tiny tail doesn’t skew the metric. Bands shown muted (n<20) carry too much sampling noise to drive the headline.