Naismith Nerds
RAPM ratings, win probabilities and lineup analytics from a thrice-weekly pickup run. Every game since January 11, 2025, tracked by hand courtside.
Both axes are adjusted for the other player's own rating. Without that they would be mirror images — the raw pair numbers carry both players' individual value, so the chart would collapse onto a diagonal that just re-reads the leaderboard. Even adjusted, a single dot is noisy: split each pair's games in half and the two halves agree at about r = 0.2. Read the spread, not any one player.
Outscoring expectation is how a rating improves. The more games played, the closer RAPM sits to raw APM (the y = x line). Small-sample players are pulled toward zero by the ridge penalty.