Moneyline (who wins)

8 predictions logged without a market line (odds feed outage) — graded win/loss and counted in games tracked, but left out of the market Brier, every closing-line number and the totals record, which all need a price (those use the other 1273).

Totals (over/under)

How to read this: Brier score is the mean squared error of probabilities — lower is sharper; beating the market's Brier means our probabilities were more accurate than the closing price's. The four closing-price numbers are deliberately separate, because they are not the same measurement. Mean closing-line value is how far the market moved toward our pick after we logged it, in implied-probability points, and it is the number we hold ourselves to — for scale, the house margin is roughly 2 points, so anything below that is a real signal that still does not cover the toll. Mean CLV, anomalies screened is that same average with a handful of rows set aside: 4 of the 1207 priced picks moved so far between our price and the close (more than 1.0 in logit — one was a home side we logged at −605 that closed +106) that they are far likelier to be a scratched starter or a stale quote than a market opinion. Those few carry about a third of all the closing-line value in the record, so the honest thing is to publish the average both ways and let you see how much of the headline rests on them. We do not delete them. Beat the closing line is the plain share of picks the market moved toward at all, regardless of size. Still +EV at the close is a harder bar than either: the share that would remain profitable priced at the close after the house margin is removed. That last one is the figure this page previously showed under the "closing-line beat" label, which was wrong — they differ by about thirty points, so we now publish each under its own name. "Model lean" on totals grades whichever side of the line the model gave >50%. These samples grow daily; we publish them at every size.