Glossary

What every number on NBL Index means and how it's calculated. Box-score abbreviations: FGM/FGA field goals made/attempted, 3PM three-pointers made, FTA free throws attempted, ORB/DRB offensive/defensive rebounds, TOV turnovers.

Team ratings Four Factors Shooting and style Player stats Predictions Ladder and finals Model accuracy

Team ratings

Possession: One trip down the court that ends in a shot that isn't rebounded by the offense, a turnover or free throws. Every rating here is measured per possession, so fast and slow teams can be compared fairly. Estimated from the box score as field-goal attempts − offensive rebounds + turnovers + 0.44 × free-throw attempts, averaged over both teams.
AdjEM: Adjusted efficiency margin: how many points per 100 possessions better a team is than an average NBL team on a neutral court. AdjO minus AdjD. The headline rating; teams are ranked by it.
AdjO: Adjusted offense: points scored per 100 possessions against an average defense on a neutral court. Higher is better.
AdjD: Adjusted defense: points allowed per 100 possessions against an average offense on a neutral court. Lower is better.
AdjT: Adjusted tempo: possessions per 40 minutes against an average opponent. Higher means a faster pace. Pace isn't good or bad on its own.
Adjusted: Solved for every team at once so that the strength of each opponent, and home court, are taken into account. Beating good teams counts for more than beating weak ones by the same margin.
Rank (small numbers): The team's league rank for that stat, 1 being best. Ranks already account for direction, so the best defense is ranked 1 even though its number is the lowest.
SOS: Strength of schedule: the average AdjEM of the opponents a team has played this season. Higher means a tougher schedule.
Left SOS: The average AdjEM of the opponents a team still has to play.
Luck: Actual wins minus the wins the ratings expected for the games played. Positive means a team has won more close games than its play deserved; it tends to even out.
Résumé: Wins above what an average team would be expected to win against the same schedule, home and away. Credits wins by how hard they were to get.
Trend: Recent form: how a team's last few weeks of play compare with its season rating, in points per 100 possessions.
7d: How many places a team has moved in the ratings over the last 7 days.
Preseason rating: The rating before the season's first game, built from last season. Early in the year ratings lean on it; it fades as games are played.

Four Factors

eFG%: Effective field-goal percentage: (FGM + 0.5 × 3PM) ÷ FGA. Counts a three as one and a half makes, because it's worth 50% more.
TO%: Turnover percentage: turnovers ÷ possessions. On offense lower is better; on defense (turnovers forced) higher is better.
OR%: Offensive rebound percentage: offensive rebounds ÷ (offensive rebounds + opponent defensive rebounds). The share of a team's misses it gets back.
FTRate: Free-throw rate: free-throw attempts ÷ field-goal attempts. How often a team gets to the line.
Defensive four factors: The same four numbers for opponents: the eFG%, turnover rate, offensive rebounding and free-throw rate a team allows. Dean Oliver's four factors explain most of the difference between good and bad teams.

Shooting and style

3P%: Three-point field goals made ÷ attempted.
2P%: Two-point field goals made ÷ attempted.
FT%: Free throws made ÷ attempted.
3PA rate: Three-point attempts ÷ all field-goal attempts: how much a team lives behind the arc.
Ast/FGM: Assists ÷ made field goals: how much of a team's scoring comes from passing.
Steal%: Steals ÷ opponent possessions.
Block%: Blocks ÷ opponent two-point attempts.
Bench share: The share of a team's points scored by players who didn't start.
Raw efficiency: Points scored and allowed per 100 possessions, before any adjustment for opponents.

Player stats

Per 40 (P/40): A stat scaled to 40 minutes, a full NBL game, so bench players and starters can be compared.
TS%: True shooting percentage: points ÷ (2 × (FGA + 0.44 × FTA)). Scoring efficiency that counts threes and free throws.
USG%: Usage rate: the share of the team's possessions a player uses (shots, free-throw trips and turnovers) while on court.
AST%: Assist percentage: the share of teammates' made shots a player assisted while on court.
ORB% / DRB%: Rebound percentages: the share of available offensive or defensive rebounds a player grabbed while on court.
STL% / BLK%: Steals per opponent possession, and blocks per opponent two-point attempt, while the player was on court.
TOV%: Turnovers ÷ (FGA + 0.44 × FTA + turnovers): how often a player's possessions end in a turnover.
ORtg: Points produced per 100 possessions used: 100 × points ÷ (FGA + 0.44 × FTA + turnovers). A simple scoring-efficiency measure, not Dean Oliver's full offensive rating.
+/- per 40: The team's scoring margin while the player was on court, scaled to 40 minutes. Noisy over small samples and affected by teammates.

Predictions

Win probability: How often the model expects a team to win this game. Built from the two teams' ratings and home court, then calibrated against how far past predictions have missed.
Projected score: The expected score from both teams' adjusted offense, defense and tempo.
Projected margin: How many points the model expects the winner to win by.
Total: The projected combined points of both teams.
Thrill: A 0–100 pre-game watchability score: both teams' quality, how close the game should be, and how many points are expected.
Watchability stars: 1 to 5 stars for a played game, from how close it finished, overtime and scoring, without showing the score.

Ladder and finals

Simulated season: The rest of the regular season and the finals played out 10,000 times using the ratings, with a little random variation in team strength each time because ratings are estimates.
Projected record: Average wins and losses across all the simulated seasons.
Top 2: Chance of finishing first or second, which goes straight to the semi-finals.
Top 6: Chance of making the finals.
Play-in: 3rd plays 4th, with the winner going through to the semi-finals. 5th plays 6th, and the winner plays the loser of 3rd v 4th for the last semi-final spot.
Finish position: The share of simulated seasons a team finished in each ladder position, 1st on the left. The outlined square is the most likely finish.

Model accuracy

Backtest: A test on past seasons where the model only ever sees games played before each prediction, refitting every week. It shows how the model would really have done.
Log loss: A score for probability predictions that punishes confident misses hard. Lower is better; a coin flip scores 0.693.
Calibration: Whether the probabilities mean what they say: teams given a 65% chance should win about 65% of the time.
Prediction ledger: Every prediction published before tip-off, kept and graded afterwards, so the track record can't be rewritten.