Stats glossary

On Courtside

Record (W-L)

Wins and losses in regular-season games. Playoff games are counted separately.

Win %

Wins divided by games played. A 50-32 team wins .610 of its games.

Points per game (PTS)

Average points a team scores in a game.

Allowed per game (OPP)

Average points a team gives up. Lower is better.

Point differential (Diff)

Average scoring margin: points scored minus points allowed, per game. A team at +8.0 wins by eight on average. It predicts future results better than win-loss record, because close wins and losses are partly luck.

Last 10 / Last 5

The same numbers over a team's most recent 10 or 5 finals. Good for spotting hot and cold streaks.

Streak

Consecutive wins (W) or losses (L) up to the latest game. W4 means four straight wins.

Home / Away

Record in home games and in road games.

Season series

Head-to-head results between two teams in one season.

PlayersPTS, REB, AST, STL, BLK, MIN

Per-game averages over games a player actually played. Games he sat out don't count.

PlayersTrue shooting % (TS%)

Scoring efficiency that counts threes and free throws fairly: points ÷ (2 × (shots + 0.44 × free throws)). 30 points on 20 shots and 8 free throws is 63.8%. League average usually sits in the high 50s.

PlayersEffective FG % (eFG%)

Field goal percentage with a made three worth 1.5 twos. Ignores free throws.

PlayersUsage rate (USG%)

Share of his team's plays a player ends with a shot, free throws or a turnover while he's on the floor. 20% is average for five players; stars sit near 30%.

PlayersAssist % (AST%)

Share of teammates' made baskets a player assisted while on the floor.

PlayersRebound % (REB%)

Share of available rebounds a player grabbed while on the floor.

PlayersTurnover % (TOV%)

Turnovers per 100 plays a player ends. Lower is better.

PlayersPer 36 minutes

What a player's numbers would be over 36 minutes. Lets bench players and starters compare on rate.

PlayersGame Score (GmSc)

One number for a box score line, from John Hollinger. 10 is an average game, 40 is a great one.

PlayersPlus-minus (+/-)

Team point margin while a player was on the court. Noisy game to game, since it depends on teammates too.

How Courtside computes these

Advanced numbers come from box scores with Basketball-Reference's formulas. Possession-based stats can differ slightly from NBA.com, which counts possessions from tracking data.

TeamsFour factors

Dean Oliver's four things that decide most games: shooting (eFG%), turnovers (TOV%, turnovers per 100 plays), offensive rebounding (ORB%, share of your own misses you grab back) and free throws (FT rate, free throws made per shot). Each has a defensive twin: opponent eFG%, forced turnover %, defensive rebound % and opponent FT rate.

TeamsScouting report (teams)

All 30 teams are ranked on the eight four-factor numbers. A top-5 rank is a strength, bottom 5 a weakness, up to three of each. If fewer than three qualify, we widen to top or bottom 8. Under 10 games the report is marked Early read. Tendencies (pace, share of threes, share of assisted baskets, top scorer's share of points) describe style, not quality.

PlayersScouting report (players)

Players averaging 15+ minutes who played in at least 20% of their team's games are compared with others at their position (guards, forwards, centers). The 85th percentile or higher is a strength, the 15th or lower a weakness. Low usage is never called a weakness, since it's a role, not a flaw.

PlayersPercentile

The share of the comparison group a player beats. 90th percentile in true shooting means he's more efficient than 90% of his peers. For turnovers the scale is flipped, so a high percentile always means better.

PlayersRole tags

First rule that fits wins. High-usage creator: usage 28%+ and assist % 25+. Primary scorer: usage 28%+. Floor general: assist % 30+. Rim protector: a center at the 80th+ percentile in blocks per 36. Stretch big: a forward or center taking 35%+ of shots from three and rebounding at least at the median. 3-and-D: 45%+ of shots from three and 60th+ percentile in steals or blocks per 36. Floor spacer: 45%+ of shots from three. Glass cleaner: 85th+ percentile in rebound %. Everyone else: Role player.

Players3PA rate, FT rate

3PA rate is the share of a player's shots that are threes. FT rate is free throws made per 100 shots, a measure of how often he gets to the line and converts.

Stats table shading

Green cells are in the top quarter of the players currently shown (after position and minimum-games filters), darker green the top 10%. Red is the bottom quarter. Games and minutes aren't shaded.

GamesMatchup edges

Each game preview sets the four factors against each other: one team's offense against the other's defense on shooting, turnovers, rebounding and free throws. An edge is where one team ranks top 5 and the other ranks bottom 8 on the same battle, like a ball-secure offense (2nd in turnover %) against a defense that rarely forces turnovers (27th). Strength against strength (both top 5) is listed as a key battle. Until both teams have played 10 games, previews use last season's numbers.

PlayersPlayer vs player

Two players' regular seasons side by side. Blue marks who leads each row, and "leads 6 of 8" counts those rows. Usage is shown but not scored, because taking more shots isn't better or worse by itself. The small numbers are percentiles against players at the same position who play 15+ minutes. Before a player has 10 games in the new season, his numbers come from last season and say so.

PlayersForm and Hot / Cold

Form compares a player's last 1 to 20 games with his season average. Hot means his last 5 games are 15% or more above his season in points or true shooting; Cold means 15% or more below. Mixed signals (one up, one down) get no tag. Tags need 10+ games and only show while the season is on. Example: Tyrese Maxey's last 5 regular-season games of 2025-26 averaged 23.0 points vs 28.3 for the season, 19% below, which would have tagged him Cold.

ShotsShot zones

Shot charts split the half court into the zones NBA.com uses: the restricted area (within 4 feet of the rim), the rest of the paint (the 16-foot-wide lane up to the free-throw line), midrange (every other 2), the left and right corner 3s (the straight part of the line, up to 14 feet from the baseline), above-the-break 3s, and the backcourt. A shot is a 3 when ESPN scored it as one. Each zone is colored by how the player's FG% compares with the league average in that zone.

ShotsRestricted area

The half circle 4 feet from the center of the rim. It's where layups, dunks and tip-ins happen, and the league makes about 66% there. On our shot charts it is its own zone, separate from the rest of the paint.

ShotsCorner 3

A three from the straight part of the line along the baseline, up to 14 feet out. It's 22 feet from the rim instead of 23.75 at the top of the arc, so it's the shortest three. The league shot about 39% on corner 3s in 2025-26 vs 35% above the break. Left and right are split on the chart.

ShotsShot types

ESPN names every shot (Driving Layup Shot, Pullup Jump Shot, Turnaround Hook Shot and about 50 more). We group them: Layup/Dunk (layups, dunks, finger rolls and tip-ins), Jumper (every jump shot that isn't a floater), Floater/Hook (floaters and hooks). End-of-quarter heaves are left out.

ShotsHex chart

The Hex view groups shots into small hexagons about 3 feet across. Bigger hexagons mean more shots from that spot. The color compares his FG% there with what the league shoots from the same distance (2s and 3s separately), so a 45% corner 3 spot shows hot even though 45% would be poor at the rim.

ShotsShot clips

Tap a dot on a shot chart and choose "Watch this shot" to see the video of that exact shot. On a computer it opens in a small NBA.com window over the page (a new tab on phones). The video plays on NBA.com; Courtside only links to it. We match each ESPN shot to NBA.com's shot log by game, quarter, clock, shooter and result (99.4% of 2025-26 shots matched). Example: SGA's 29-foot step-back 3 at the Clippers on April 8. New-season games show a link to the game's ESPN play-by-play until NBA.com's shot log for that season is published.

ShotsGame shot chart

Every shot from both teams in one game, dots in each team's color (filled = made, x = missed). Filter by team, player, quarter or result, or tap a name in the shooting list. Example: open Lakers game 32 (Jan 2 vs Memphis), pick LeBron James, and tap any of his shots to watch it. Open it from the Shot chart button on a game page, or pick One game on a player's shot chart.

PlayersCourtside score

A 0-100 rating that averages a player's percentile ranks against others at his position (guards vs guards, forwards vs forwards, centers vs centers) in five areas: Scoring (points per 36, true shooting), Playmaking (assist rate, turnover rate with fewer turnovers scoring higher), Rebounding (rebound rate), Defense (steals and blocks per 36) and Shooting (3-point percentage and 3-point attempt rate). Overall is the average of the five. Only players with 15+ minutes a game in at least 20% of their team's games are ranked. Example: Victor Wembanyama's 2025-26 categories are 89.5, 70.5, 98.0, 81.5 and 34.5, so his overall score is 74.8, first in the league. It is a box-score composite, not an impact metric.

ShotsShot twin

A current player whose shot diet is closest to another player's this regular season. The diet has six numbers: share of shots at the rim, in the rest of the paint, from midrange, from the corner 3 and from above the break 3, plus free throws per 100 shots. The match score is 100 minus 1.5 times the straight-line distance across the six numbers, so identical diets score 100. Most players' closest twin scores between 83 and 94. Only players with 300 or more shots count. Example: Tyrese Maxey and Donovan Mitchell are 8.7 points apart, a match score of 87. It measures where shots come from, not how well they go in.

PlayersTier list

Your own ranking of current players into five tiers (S best to D), shared as a link. Starter lists fill the tiers by the numbers: scorers by points per game, positions and team rotations by game score. Each player shows one stat you pick: points, rebounds or assists per game, or true shooting, from the latest regular season.

PlayersFollow

Tap the star on any team or player page to follow them. They appear at the top of the home page under Your Courtside: teams show their next game (or the live score), last result and injuries; players show their last game, their Hot or Cold tag during the season and their team's next game. There is no account: follows are saved on this device only, up to 10 teams and 20 players. Use Manage on the home page to unfollow.

TeamsOn/off and WOWY

On/off compares how a team does with a player on the floor versus on the bench, using net rating (points scored minus points allowed per 100 possessions). Example: OKC is +17.1 with Shai Gilgeous-Alexander on the floor and +4.3 with him sitting, so his on-off is +12.7. WOWY (with or without you) splits a team's minutes for two teammates into four parts: both on, only one of them, or both sitting. Built from every lineup in ESPN play-by-play, checked against box-score plus-minus (98.9% exact). Small samples swing a lot.

PlayersImpact rating (RAPM)

The points per 100 possessions a player adds to his team's scoring margin compared with an average player, after accounting for the other nine players on the floor in every stretch of the season. It splits into offense (points added) and defense (points saved). A ridge regression pulls every player toward average, with the strength chosen by testing on games the model did not see; the 80% range comes from refitting on reshuffled games. It uses no box score. Example: Shai Gilgeous-Alexander leads 2025-26 at +7.0 (range +5.3 to +8.5). One season is noisy, so read the ranges, not the exact ranks.

TeamsLineup net rating

Points a group of players scores per 100 possessions minus the points it allows per 100 opponent possessions while they share the floor. A 5-man unit is five teammates on the court together; a 2- or 3-man combo counts every minute those players shared, whoever else was on. Example: OKC's Gilgeous-Alexander, Dort, Williams, Holmgren and Hartenstein were +32.9 in 113 minutes. 50 minutes is only about 100 possessions each way, so small samples swing a lot.

PlayersShot quality (expected points)

Every shot gets a make chance from a model that knows only where and how it was taken: distance, two or three, corner, paint or midrange, shot type, 4th quarter, last 24 seconds of a period and home. Expected points = that chance x 2 or 3. Shot selection = a player's expected points per shot minus the league's, so +0.05 means his shots are worth 0.05 more points each than a typical shot. Shot making = actual minus expected points per 100 shots, so +5 means he scores 5 more per 100 shots than an average shooter on the same shots. 200+ shots to be ranked. The model cannot see defender distance, play type or the shot clock, and over 300 shots luck alone moves shot making by about 6 points per 100. See /shot-quality.

PlayersIs it real? (stabilization)

Every stat is part skill, part luck. The stabilization point is the sample size where a player's number is half skill and half luck. Measured on 2025-26: 3P% about 500 attempts, FT% about 26 attempts, TS% about 220 shooting possessions, rim FG% about 86 shots, points streaks about 10 games. Likely true level = (made + M x his usual rate) / (attempts + M), shown with an 80% range. A hot 3-point week on 40 attempts is mostly noise; a new free throw stroke shows up fast. See /stabilization.

GamesGame breakdown and runs

A one-page recap of a finished game: the score by quarter, a chart of how the lead moved, the four factors, a short 'why they won' list with rough point values, the top performers by Game Score, and where each team shot from compared with its regular season. The point values are rules: turnovers and offensive rebounds x 1.1, 3s made x 3, 2s x 2, free throws x 1 (a fit on 2025-26 gives about 1.0 per turnover, so they can run 5 to 10% high). A run is a stretch where one team scores at least 8 points while the other scores 2 or fewer, for example '14-0 run, Q3 8:12-4:40'. See /breakdowns.

GamesWin probability and Elo

The model's pregame chance that a team wins. It combines team Elo ratings (points won and lost after every game, more for upsets and blowouts), home court, days of rest and back-to-backs. Replaying 2025-26 one day at a time, using only earlier games each day, it picked 68.0% of winners, versus 55.4% for always picking the home team. Elo: 1500 is average; a 100-point edge is about a 64% chance on a neutral court. Brier score: the average squared miss of a probability forecast; 0 is perfect and 0.25 is a coin flip. It does not know about injuries or trades. See /predictions.

GamesPlayoff odds

The chance a team makes the playoffs, from playing out the rest of the regular season 10,000 times. Each game is decided by a weighted coin flip using our pregame win probability. Every simulation also nudges each team's rating a little at random, because ratings are estimates, while keeping each game's average chance equal to the model's. Playoffs means a top-6 finish or winning a play-in spot. The play-in is seeds 7 to 10: 7 hosts 8 and the winner is the 7 seed, 9 hosts 10, and the loser of 7 vs 8 hosts the winner of 9 vs 10 for the 8 seed. Top 4 means home court in the first round. Projected record is the average final record, and the range runs from the 10th to the 90th percentile of wins. Magic number: own wins plus losses by the closest chasing team that clinch finishing ahead of it (82 + 1 minus own wins minus the chaser's losses). Ties in the standings are broken by a random draw, not the NBA's tiebreak rules. Updated every morning.

GamesLive win probability

Each team's chance to win at any moment of a game, from the score margin, the time left and the pregame win chance. A lead counts for more as the clock runs down, and the pregame edge fades as the game goes on. Tested on 264 held-out games from 2025-26: log loss 0.453 and Brier 0.150, vs 0.487 and 0.166 for a version that only knows the score and the clock. Biggest swing: the largest change in win chance inside any three minutes of game time, including a shot at the buzzer, e.g. "Lakers 89% after the 7-0 run in Q4 (from 47%)". It can't see possession, fouls, timeouts or who is on the floor.

PlayersPlayer projections

Our estimate of a player's minutes, points, rebounds, assists and threes in his next game. It starts from his minutes and per-minute production this season, over his last 10 games and over his last 5, then adjusts for how much the opponent allows to his position (guards, forwards or centers) and for the pace of both teams. Each number comes with an 80% range: in testing on 2025-26, about 8 in 10 real results landed inside it. Until a player has 5 games in a new season, his projection is based on last season. Projections are frozen at tip-off and graded after the game.

ModelsModel report card

The page where every Courtside model is graded in public. Each model is scored against a simple baseline. For example, win probability is compared with always picking the home team at its usual win rate, and shot quality is compared with a zone-only lookup. The grade word says how much smaller the model's error is: Strong at 10% or more, Solid at 3% or more, Even within 3% either way, Behind otherwise (3% and 1% for AUC). Forecasts go into a timestamped log before tip-off, and the log can never be edited. Brier score: the average squared miss of a percentage forecast. 0 is perfect and 0.25 is calling every game 50/50. A 70% favorite that wins scores 0.09; one that loses scores 0.49.

PlayersWatch playlists

Courtside does not host video. Every play in a playlist opens NBA.com's own clip for that shot, matched by NBA.com game id and event number (99.4% of 2025-26 regular-season shots). The rest link to ESPN's play-by-play. So do playoff games and 2026-27 games until NBA.com's shot file for that season is published and matched. One small window plays the clips. Tap Next or press the right arrow when a clip ends, because nothing plays by itself. A playlist holds up to 200 plays. Clutch: the last 5 minutes of the 4th quarter or overtime with the score within 5 points before the shot. Go-ahead or tying shot: a make in the last minute of the 4th or overtime that tied the game or put the shooter's team ahead. Only field goal attempts have clips: free throws, turnovers, rebounds, assists and defense are not in playlists.

GamesGame plan

A short list of rules showing how one team beats a given opponent, made from both teams' season numbers. Each item has numbers and ranks out of 30: Attack: a zone where the team is good and the opponent defends badly. Take away: the opponent's favorite zone where they are also efficient. Target: the opponent's weakest defensive lineup among their most-used units. Key scorer: the opponent's top scorer's best and worst zones and any left or right lean. Tempo and glass: pace, rebounding and turnover edges. Until a team has played 10 games, its plan uses last season's numbers with the current roster. There is no tracking data, so the plans can't see defenders, individual matchups or play types.

PlayersWhat changed?

A nightly check for players whose game really moved: more minutes, a bigger role, a new shot diet, better or worse shooting. We compare a recent stretch with the rest of his season. The stretch is his games since a trade, since his minutes or usage shifted, or his last 10 games (last 25 when we are still showing last season). A change must be big (for example 4+ points a game or 5+ minutes). It must also pass the "is it real?" test: both stretches are pulled toward his season level by how noisy the stat is, and at least half the change must survive. "Probably real" keeps 60% or more, "Partly real" 50 to 59%. Each change lists the likely causes with numbers: minutes, shot volume, efficiency, shot diet vs shot-making, starting role, teammates out or new, and opponents.

TeamsTrade and lineup lab

A tool that estimates how good teams or lineups would be after you move players. A team's roster rating adds up each player's impact rating (RAPM) weighted by the minutes he is expected to play. Those minutes come from last season's minutes per game, counted over at least 41 games and capped at 36. Minutes nobody covers are filled at replacement level (-2.0). One point of roster rating has been worth 1.22 points of real net rating, and one point of net rating has been worth about 2.1 wins over 82 games. For a lineup, offense is league average plus the five offensive ratings times 1.22, and defense is league average minus the five defensive ratings times 1.22. It shows impact only. It does not check salary, cap, apron or roster rules, and it cannot see fit, chemistry or injuries.

ShotsHeaves not counted

ESPN logs the long buzzer-beater attempts players throw at the end of a quarter. Official box scores don't count a missed heave as a field goal attempt, so we leave them out too and note how many there were. That keeps made plus missed equal to the FGA in the box score. Example: in 20 games checked against box scores, 10 of 12 differences were missed heaves.

PlayersShot profile

Six percentiles on the scouting report from shot locations: share of shots at the rim, from midrange and from the corners, and FG% in each of those zones (FG% needs 50+ shots in the zone). Compared with players at the same position who average 15+ minutes. Shares describe a player's shot diet, so they are never listed as a weakness, except rarely taking corner 3s for players who shoot a lot of 3s.

TeamsPer 100 possessions

Box stats scaled to 100 possessions, so fast and slow teams compare fairly. A team scoring 115 per game at a fast pace may score fewer per 100 than a slow team scoring 110.

TeamsLast 30 / 14 days

Team stats counted over games in the 30 or 14 days before the season's latest game, so they still mean something in the offseason.

Around the web

Advanced stats that need play-by-play or tracking data. We plan to add some as our data grows.

TeamOffensive / defensive rating

Points scored or allowed per 100 possessions. Removes pace so fast and slow teams compare fairly.

TeamNet rating

Offensive rating minus defensive rating. The per-possession version of point differential.

TeamPace

Possessions per 48 minutes. How fast a team plays.

LineupsOn/off

How a team performs with a player on the court versus on the bench. Noisy on its own, useful with context.

ImpactRAPM

Regularized adjusted plus-minus. Estimates a player's effect on the score while accounting for teammates and opponents.

ImpactDPM / EPM / BPM

Impact estimates that blend box-score stats with plus-minus data. Different sites use different recipes.

Box scorePER

Player efficiency rating. An older all-in-one number built from box-score stats, where 15 is league average.