The Pelicans traded Jordan Hawkins, Micah Peavy, a future second-round pick and a second-round pick swap to Memphis for AJ Johnson and Taj Gibson. Gibson was then waived, making New Orleans' roster maneuvering the freshest NBA development in today's data. The Pelicans' market-blended power rating moved 88.8 points since September 8 and is up 74.9 points versus September 5.
The Clippers posted the day's largest move at +116.5 points, alongside news that the NBA stripped the franchise of five first-round picks, fined it $30 million and suspended owner Steve Ballmer over Kawhi Leonard-related cap circumvention. Indiana went the other way at -93.9, Philadelphia fell 85.1 and Oklahoma City dropped 65.2; OKC's correlated news includes the three-team trade sending Luguentz Dort to Atlanta. These are preseason futures-market reactions, not results on the floor.
For Detroit-Boston on October 20, the model starts with a 62-point Elo difference including home court, producing a 58.76% base probability for Detroit. Jayson Tatum's day-to-day status supplies a weighted impact of about 0.06 points, while there is no rest adjustment; the final Detroit probability is 58.78%. The live home-court constant is 50 Elo points.
Across 1,246 graded games, COURT VISION has 68.3% accuracy and a 0.206 Brier score. Its mean predicted home-win rate is 56.44% against an actual 55.06%, leaving a 1.38-percentage-point systematic home bias. That bias is small but positive, so the model has historically leaned slightly too far toward the home side.
A new self-heal hypothesis reports a +1.38% systematic home-probability bias across 1,246 real-game replays. The patch has not been applied yet, so this is a measured drift in the current Elo home-court prior, not a confirmed fix. The replay signal is evidence-backed for elo_hca, making it the most actionable lead since the last notebook entry. That bias is visible in the five biggest misses since the last write-up. Four of them are home teams getting blown out: DET lost to CLE 94-125 (35.9 margin error) despite a 68.9% home probability; CLE lost to NY 93-130 (33.8 margin error) with the model leaning NY at 62.5%; PHI lost to NY 114-144 (25.2 margin error); and MIN lost to SA 109-139 (25.1 margin error). Only the SA-OKC game went the model's way at home. The common thread is the model anchoring too hard to home status in games that turned into road routs. Injury noise is still present but does not cleanly explain these margins. Jalen Brunson appears as Out with a 0.0 impact in two separate games, while day-to-day tags carry a 0.1 impact. That mismatch was already flagged in the prior entry; the new signal is that even after accounting for those table issues, home probability looks systematically too high. The upcoming season's first slate already shows the market disagreeing sharply with the model on home-team strength. The WSH-MIL line has the market at MIL -6.5 with a 66.5% home probability, while the model makes MIL a 2.7-point underdog with only a 39.1% win probability — a 27.4 percentage-point gap. That will be a live test of whether the home-court prior is overstating WSH or understating MIL. Bottom line: the replay evidence points to elo_hca as the parameter to re-examine, not the injury weights. The bias is small in expectation but appears amplified in lopsided games. [Flagged for the fitter, not applied: elo_hca -- The +1.38% measured home-probability bias across 1,246 replays and the cluster of home-team blowout losses in the biggest misses suggest the Elo home-court advantage prior is inflating home win probability, particularly in games that become road blowouts.]
"New Orleans: +88.8 points in the market-blended power rating since September 8."
Read live predictions and the full interactive site →