SportiveWorldSportiveWorld

FORECAST · NBA

Season

projected standings, with range and movement since the previous snapshot

Season2026-27Update statusNot currentdata as of 20 Sept· last recalculation 20 Sept

What we expect from the seasonsnapshot of 20 Sept

Oklahoma City Thunder has the most projected wins in the WEST (61.5)Cleveland Cavaliers has the most projected wins in the EAST (52.2)
See sources

The range in this snapshot is the simulation's (10th and 90th percentiles of 800 simulated seasons with fixed strength). Measured coverage: 51.4% (n=210): it is NOT an 80%. The evaluated interval, from previous seasons' errors, is published with the next re-issue.

The seed is an ordering by expected wins, without play-in or tiebreakers. Injuries are received but the model does not consume them: they do not move the number.

  • known + TBD = 19.680 min equivalentes; el TBD (fichajes/rotación futura) NO se reparte
  • Known minutes + unallocated = 19,680 equivalents per team. “Unallocated” is future signings and rotation that don't exist yet: it is not force-distributed.

Who changes, and whysince the 19 Sept snapshot

Teams rising most · projected wins

No team moved more than a tenth.

Teams falling most · projected wins

No team moved more than a tenth.

What entered the model

3 official roster moves (GSW, MIN, NOP).

  • NOPNew Orleans Pelicans waived guard Markquis Nowell.
  • MINMinnesota Timberwolves waived guard Stephen Thompson.
  • GSWGolden State Warriors waived Elias Ralph.

Official NBA transactions, as recorded.

68 injury reports received in this window: the season model does not read them, so they do not move the number.

The change is measured between two snapshots, not a causal attribution; renormalising to 1,230 wins, one team's move nudges the other 29.

Track recordbacktest and prospective, kept separate

7.30
MAE · Wins
walk-forward 2019-2025 · n=210 · previous season's wins 8.40
51.4 %
measured coverage
n=210 · 2019–2025 · simulation range · no valid evaluation for this interval: it is not the 80% range

interval served: simulation range. no valid evaluation for this interval: it is not the 80% range.

  • Evaluated with the rosters actually used (information known after the fact): it does not accredit the real coverage or error of a preseason prediction.
  • Quantiles of the simulation with fixed strength: only game-to-game variance, no strength uncertainty. It is not an 80% range.

What Forecast covers in this competitionevery block declares its state: no empty tabs

Game

Available
  • Does not read the day's injury report.
  • Publishes a point per game, no band: calibration is measured in the track record.
  • No resolved games with a sealed pre-game call yet: the prospective record starts with the season.

Season

Available
  • The seed is an ordering by expected wins, without play-in or tiebreakers.
  • Injuries are received but the model does not consume them: they do not move the number.
  • The wins interval states in each snapshot which method it is and its measured coverage; the evaluated one (residuals, 80% with ex-post rosters) is served after the re-issue.

Playoffs & title

Not yet modeled
  • No evaluated model: no probability is claimed.

Players

Available
  • No projected stat line (points, rebounds, assists).

Awards

Not yet modeled
  • No model: nothing to claim.

Candidates per projection (not an award probability)

Track record

Available
  • No resolved games with a sealed pre-game call yet: the prospective record starts with the season.