Every metric and acronym, in plain English

Anchor Quality IndexAQI

How a team plays with a player on the floor, scaled by how much of the offense he carries and how efficiently he shoots.

The Anchor Quality Index grades a player by how his team performs with him on the floor, scaled by his usage and his shooting efficiency. A team's best AQI is its anchor. Every champion's anchor since 2000-01 has cleared 1.73, which makes 1.73 the entry line; the lowest champion anchor was 1.85, on the 2004 Pistons. Across 776 team-seasons the anchor correlates 0.351 with winning the title and 0.902 with win percentage. Net rating here is always the on-court figure, never Box Plus/Minus. About 11% of teams that clear the line win the title, so 1.73 is a ticket in, not a pick.

Key number: 1.73: every champion's best player since 2000-01 has cleared it (lowest champion anchor 1.85)

Full explainer → · Related: AQI₂, Ticket (1.73 Player), Four Contender Tests, NetRtg, USG%, TS%, RQS

Research findings and source essay →

Clutch Paradox

Is closing a real, repeatable player skill? An open question waiting on clutch-split data.

The Clutch Paradox asks whether a small group of elite shot creators are truly better in the last minutes of close games than shot-quality models expect, and whether that holds year after year. At the team level the answer is in: clutch results do not persist (see CEV). The player-level version has a written spec: a player's clutch true shooting minus his regulation true shooting, over a minimum number of clutch attempts, across several seasons, judged against a persistence bar of 0.50. It has not been run, because it needs league clutch splits that are still being sourced.

Related: CEV, TS%, Sovereign Premium

Ghost Points IndexGPI

Empty volume: scoring that piles up without turning into wins. Built for role players.

The Ghost Points Index flags scoring volume that does not convert into winning. It multiplies a player's volume by three penalties, for inefficiency, for losing and for low PER, so a high GPI means a lot of points that did not help much. Among role players it correlates −0.656 with team win percentage, a strong signal. It is built and tested for role players; applying it to stars is outside what it has been shown to do. The next test checks whether it adds information beyond true shooting, usage and box plus-minus.

Key number: r = −0.656 with win% among role players

Related: TS%, USG%, Efficiency Tax

Load Efficiency ScoreLES

Who is the star? Efficiency above break-even, times usage, times on-court impact.

The Load Efficiency Score identifies stars. It takes a player's true shooting above a 0.55 break-even and multiplies it by his usage and his on-court net rating, so a high score needs all three: efficient, heavily used and winning his minutes. Stephen Curry's 2015-16 season holds the highest mark at 0.697. LES describes who carries a team and makes no claim about titles. It sits close to AQI, which uses the same three ingredients, and the open question is whether it adds anything AQI does not already capture.

Key number: Stephen Curry, 2015-16: 0.697, the highest on record

Related: AQI, TS%, USG%

Net Playoff Survival ScoreNPSS

Which star bigs, and which guards, get schemed out of the playoffs.

The Net Playoff Survival Score looks for players whose value shrinks when playoff opponents game-plan against them. It has two archetypes: the Schemable Big, a star big whose regular-season impact fades once playoff defenses and offenses scheme around him, and the Hunt-Exposed Guard, a guard opponents hunt on defense. It is the subject of a research paper in progress. Its two implementations are being reproduced side by side, with precision and calibration reported for each, before any accuracy figure is published.

Related: PDR, DTI, AQI, Two-Way-Wing Translation

Playoff Decay RatePDR

Who loses least when the playoffs tighten. Everyone declines; the question is by how much.

Playoff Decay Rate measures the change from regular season to playoffs in true shooting, PER and free-throw rate, grouped by player archetype. The headline is that everyone declines: the league-wide average PER drop is 1.83 across 1,392 player-seasons from 2016 to 2025. Foul-drawers and rim finishers lose the least true shooting, about 1.5 points, while catch-and-shoot players and three-heavy creators lose the most, about 3.4 to 3.5 points. No archetype rises in the playoffs. The next version adds controls for opponent, role, minutes and regression to the mean.

Key number: League mean ΔPER −1.83, regular season to playoffs (1,392 player-seasons, 2016-25)

Related: Two-Way-Wing Translation, TS%, NPSS

Second Star PremiumAQI₂

AQI for a team's second-best player. It reads the regular-season standings even better than the star's number does.

The Second Star Premium is AQI computed for the roster's second-best player. Across 775 team-seasons it correlates 0.923 with win percentage, a touch higher than the best player's 0.902. Because on-court net rating is shared with teammates, the second number works best as a reading on the team around the star. June still leans toward the first star: champions' best players ranked in the league's top five in 25 of 26 seasons. The sidekick matters too, since 23 of 26 champions had a second player at 1.73 or higher.

Key number: r = 0.923 with win% across 775 team-seasons (best player: 0.902)

Related: AQI, Ticket (1.73 Player), Four Contender Tests, RQS

Research findings and source essay →

Sunrise/Sunset RatioSSR

The shape of a career: early-prime production against late-career production.

The Sunrise/Sunset Ratio divides a player's PER from ages 22 to 26 by his PER from ages 30 to 34. A ratio under 1.00 marks a graceful ager who held or grew his production; ratios of 1.60 and above mark the steepest falls from an early peak. It pairs with the Shelf Life Score as a way to think about how long a player's value lasts. It is directional so far, built on 86 career arcs, and the next version adds controls for survival bias, since only durable players are still in the league at 34.

Related: SLS, VCR

Ticket (1.73 Player)

Shorthand for a player at 1.73 AQI or higher, the line every champion's best player has cleared.

A ticket is a player whose AQI is 1.73 or higher. The name comes from the champion record: every champion's best player since 2000-01 has been at or above that line, so having one is the price of entry to a title run. It is not rare enough to pick a winner, since about two in five teams carry at least one, but a team without one is outside the history. The first two of the four contender tests count tickets: one, and then a second.

Key number: 24 of the last 25 champions had a ticket on the roster going into the season

Full explainer → · Related: AQI, Four Contender Tests, AQI₂

True Shooting PercentageTS%

Scoring efficiency that counts threes and free throws for what they are worth.

True shooting percentage measures how many points a player scores per shooting possession, with three-pointers and free throws counted for their full value. It is the standard efficiency stat and the efficiency term in AQI, where a player's TS% is divided by a 0.550 league anchor: an above-anchor shooter scales his AQI up and a below-anchor shooter scales it down. Playoff Decay Rate tracks how much TS% each type of player loses between the regular season and the playoffs.

Key number: 0.550: the league anchor AQI divides TS% by

Related: AQI, LES, PDR, USG%

Two-Way-Wing Translation

Do switchable two-way wings hold up in the playoffs? Yes: they decline less than any other archetype.

Two-Way-Wing Translation tracks how two-way wings carry their game into the postseason. Defined strictly, as shooting guards and small forwards with a three-point attempt rate of .30 or more, a non-negative defensive box plus-minus, usage under 24% and at least 100 playoff minutes (187 players), wings drop 1.45 PER from regular season to playoffs against 1.73 for everyone else. Under a looser definition the gap is 1.28 against 1.80. Everyone declines, but wings decline least, which makes them the most playoff-resistant archetype in the data.

Key number: ΔPER −1.45 for two-way wings vs −1.73 for everyone else

Related: PDR, NPSS

Usage RateUSG%

The share of team possessions a player finishes, with a shot, a trip to the line or a turnover, while he is on the floor.

Usage rate estimates how much of the offense runs through a player: the percentage of team possessions he ends with a field-goal attempt, free throws or a turnover while he plays. In AQI it is entered as a fraction (0.356, not 35.6). It is also the part of a player's profile that travels best when he changes teams: for movers, usage carries over at r = 0.76, against 0.25 for on-court net rating. Star Dependence divides the busiest player's usage by the roster average to show how top-heavy an offense is.

Related: AQI, Star Dependence, TS%, LES

Championship Code

A two-star threshold for Finals teams: one player at 15 or more player-level WEV and a second at 11 or more.

The Championship Code describes the two-star structure behind Finals teams. It looks for one player at 15 or higher on player-level WEV and a second at 11 or higher. The thresholds are frozen, and the next step is to run them through the same held-out test WEV passed, which turns a structural pattern into a tested one. Until that runs, read the Code as a description of how Finals rosters have been built, and pair it with the AQI and second-star readings for the current season.

Related: WEV, AQI, AQI₂, Star Architecture / Commune Champion

Championship Probability ScoreCPS

A points-based contender dashboard built from banded AQI, RQS, WEV, CPV and related readings. A display, not a probability.

The Championship Probability Score gathers several team readings into one dashboard: banded points for AQI, RQS, WEV, CPV, payroll concentration and the era identity measure CIS, up to a maximum of 12. Despite the name, it is not a probability and should never be read as one, because the points are not calibrated against outcomes. The plan is to rebuild it as a calibrated model, merged with CORE, and keep the dashboard as the display layer on top.

Related: CORE, AQI, RQS, WEV, CPV, CIS

Clutch Expected ValueCEV

WEV's clutch term. Built partly from playoff results, it contains the outcome, so new scoring leaves it out.

Clutch Expected Value was meant to capture whether a team executes with the game on the line. As built, two of its three inputs, playoff win percentage and Finals MVP scoring, come from after the title is decided, so it contains the outcome it is trying to explain. A rebuilt version using only regular-season close games settled the bigger question: team clutch results barely repeat from one season to the next, at 0.16 to 0.22, against 0.60 for net rating. Clutch is not a team trait. New WEV scoring drops CEV and uses (0.30 OEV + 0.60 DEV) / 0.90.

Key number: Year-to-year persistence of team clutch results: 0.16 to 0.22 (net rating: 0.60)

Related: WEV, Clutch Paradox, NetRtg

Complete Possession ValueCPV

Offense times defense, not offense plus defense. A team composite that rewards being good at both ends.

Complete Possession Value multiplies a team's offensive composite by its defensive composite, so strength at one end amplifies the other instead of simply adding to it. It is marginally the strongest team composite in the stack. Frozen on 2000-15 data and tested on 2016-23, its correlation with the champion rose from 0.327 to 0.348, and it put the champion in its top five in 7 of 8 held-out seasons. All 26 champions in the panel had a positive CPV. It runs on the same defensive engine as WEV, so read the two as one family. CPV is a team measure; it does not grade individual shots.

Key number: Held-out r = 0.348 with the champion (2016-23); champion in its top five in 7 of 8 seasons

Related: WEV, OEV, DEV

CORE Contender DimensionsCORE

A contender-risk checklist across five dimensions: availability, anchor, depth, clutch and identity.

CORE is a diagnostic that looks at a contender from five angles: availability (is the payroll on the floor), anchor (does the best player clear 1.73), depth, clutch and identity. Each dimension is scored with its own framework: Cap Space Ghost for availability, AQI for the anchor, Floor% for depth, the rebuilt clutch measure for clutch and CIS for identity. The scores are read side by side rather than added into one verdict, and summing them waits until a dimension is calibrated against outcomes.

Related: CPS, CSG, AQI, Floor%, CEV, CIS

Defensive Expected ValueDEV

A composite of team defense and the strongest piece of WEV. Offense gets you in the room; defense wins it.

Defensive Expected Value blends defensive rating, net rating and the Simple Rating System into one normalized score. It is the separator in the team stack: its correlation with winning the title is 0.301, close to three times the offensive composite's 0.112, and it is the most stable part of WEV. When WEV's weights were refit on seasons the model had never seen, the defensive weight came back at about 0.58, right beside the 0.60 it carries. That is why DataDunkNBA reads defense first.

Key number: r = 0.301 with winning the title; its weight re-derives at about 0.58 out of sample

Related: WEV, OEV, DRtg, Positive-Defense Law, Rim Protection to Team Defense Bridge

Floor PercentageFloor%

Rotation depth: the share of a roster that is actually winning its minutes.

Floor% counts the share of a roster with an on-court net rating above +2, a simple read on how many players are positive rather than how good the best one is. It tracks the standings closely, correlating 0.863 with win percentage, while its link to titles is weaker at 0.265, and last season's Floor% adds nothing to a forecast. Depth explains a season after the fact. Floor% works best as an input next to AQI and the second-star reading, not as a rule on its own.

Key number: r = 0.863 with win%

Related: AQI₂, RQS, Star Dependence, CORE

Four Contender Tests

Four preseason questions for a contender: a 1.73 player, a second one, a top-10 net rating and a top-10 defense.

The four contender tests use only what a team had going into the season. One: a player at 1.73 AQI or higher. Two: a second such player. Three: a top-10 net rating the year before. Four: a top-10 defensive rating the year before. Since 2001-02, the 105 teams that passed all four won 16 of 25 titles and filled 25 of 50 Finals spots, while the 345 teams that passed none won one. Most champions that missed a test changed through a trade or a star's return that last season's numbers could not see.

Key number: Teams passing all four won 16 of the last 25 titles (105 teams, 2001-02 to 2025-26)

Full explainer → · Related: AQI, Ticket (1.73 Player), AQI₂, NetRtg, DRtg, WEV

Research findings and source essay →

Horry Density

How many role players a champion had who could hit the shot that mattered.

Horry Density counts, for each champion, the role players who could make the big shot when defenses loaded up on the stars. It is roster-construction context: it describes how title teams were built around their stars rather than ranking anyone. A market test using March title-futures prices has been designed, but it needs an archive of those prices that does not yet exist, so the measure stays descriptive for now. It sits naturally beside the Draymond Effect and Star Architecture as a way to read a champion's supporting cast.

Related: Draymond Effect, Star Architecture / Commune Champion

Hybrid WEV

WEV with the top two players' AQI added on. A standing challenger that has not beaten plain WEV.

Hybrid WEV asks whether adding star quality to the team composite sharpens it. So far the answer is no. With the best player's AQI and the second-best player's AQI added to an earlier WEV build, its correlation with the champion is 0.301, below WEV's 0.351, and it only ties WEV on top-one rank. It stays in the stack as a challenger. The next steps are to score it on the same frozen hold-out WEV passed and to recompute it with the rebuilt AQI panel.

Key number: r = 0.301 with the champion, against WEV's 0.351

Related: WEV, AQI, AQI₂

Net RatingNetRtg

Points scored minus points allowed per 100 possessions, for a team or for a player's minutes on the floor.

Net rating is the scoring margin per 100 possessions. For a team it is offensive rating minus defensive rating. For a player, DataDunkNBA uses on-court net rating: the team's margin per 100 possessions while he is playing. That on-court number is the base of AQI, which is never computed from Box Plus/Minus or on/off splits. Because teammates share the floor, a player's on-court net rating partly reflects the lineup around him, which is why the second-best player's AQI tracks team record so closely.

Key number: Team net rating repeats year to year at 0.60

Related: AQI, AQI₂, DRtg, Floor%, Four Contender Tests

Offensive Expected ValueOEV

A composite of a team's offense. Champions clear its floor, but it barely separates them from the pack.

Offensive Expected Value blends five offensive measures, led by offensive rating and effective field-goal percentage, into one normalized score. Its finding is that offense works as a floor, not a separator. OEV correlates 0.112 with winning the title, against 0.301 for the defensive composite, and the post-2017 champions all scored 13 or higher. Good offense gets a team into the conversation; it rarely decides who wins it. That is why OEV carries 30% of WEV while defense carries 60%.

Key number: r = 0.112 with winning the title (DEV: 0.301)

Related: WEV, DEV, CPV

Roster Quality ScoreRQS

Star power plus roster structure in one team number. It describes a season closely and does not forecast the next one.

The Roster Quality Score weights the best player's AQI four times and the second-best player's twice, then adds credit for an interior anchor and for elite late-draft players. It correlates 0.873 with win percentage across 677 team-seasons and explains about three quarters of the variation in records within a season. It is a descriptor, not a forecaster: built from last season's inputs, it adds almost nothing to a win projection beyond last season's net rating. RQS moves nearly in lockstep with the best player's AQI, so treat it as part of the AQI family rather than independent confirmation.

Key number: r = 0.873 with win% (677 team-seasons)

Related: AQI, AQI₂, Interior Anchor Rule, Draymond Effect, Floor%

Star Architecture / Commune Champion

Roster shape as a matchup variable: star-deep, hybrid and no-star teams, and how they fare against each other.

Star Architecture sorts rosters by shape: teams with three or more stars, hybrids, and committees without a true star. Head to head, teams with three or more deep stars went 29-2 against no-star teams. The lens also names the exception the star metrics miss: the commune champion, a team like the 2004 Pistons or the 2026 Knicks that wins through depth and defense without a dominant star. The categories are being preregistered so they can be tested against continuous star-depth measures in held-out series.

Key number: Teams with three or more stars went 29-2 head to head against no-star teams

Related: AQI, AQI₂, Star Dependence, Championship Code

Star Dependence

How top-heavy a team's offense is: the busiest player's usage divided by the roster average.

Star Dependence divides the highest usage rate on a roster by the roster's average usage. A high number means one player runs the offense; a low number means the ball moves through a committee. It is a descriptive lens on team shape and an input to the record model alongside Floor%. Read it as context about how a team plays rather than as a lever for winning. The 2004 Pistons and the 2026 Knicks are the reminder that a committee can win the title, the commune archetype that star-first numbers struggle to see.

Related: USG%, Floor%, Star Architecture / Commune Champion

Winning Expected ValueWEV

A defense-weighted team rating that sorts the league into real contenders and everyone else. A filter for champions, never a picker.

Winning Expected Value rates a whole team, and it weights defense twice as heavily as offense. Its job is to filter. Across 26 seasons, 25 champions ranked in WEV's top ten and 22 ranked in its top five, so a title team almost always comes from inside that group. WEV does not pick the champion from within it. The published weights are 0.30 offense (OEV), 0.60 defense (DEV) and 0.10 clutch (CEV). Because CEV contains the outcome it is meant to explain, new scoring drops it and rescales the other two: (0.30 OEV + 0.60 DEV) / 0.90. Dropping CEV leaves the record essentially unchanged.

Key number: 25 of 26 champions ranked in its top ten; 22 of 26 in its top five

Full explainer → · Related: OEV, DEV, CEV, CPV, Four Contender Tests

Research findings and source essay →

Defender Targeting IndexDTI

Defender identity as a measurable input: which defenders offenses hunt, and which ones make hunting them a bad idea.

The Defender Targeting Index turns the identity of the defender into data. For each defender it asks what an offense gains when it targets him, in points per targeted possession compared with the league baseline. A higher value means a defender who gives up more when attacked; a lower value means one who holds up. Its versions cover how a defender performs when targeted, how often he is hunted and possession context. DTI is an explanatory input, the piece that lets possession models know who is guarding whom. Individual defender values are being rebuilt from source data and are not published.

Related: Holding-Math Theorem, Lineup Hardness, PEI, Rim Protection (Defender Frontier / Suppression)

Defensive RatingDRtg

Points a team allows per 100 possessions. Lower is better.

Defensive rating is the number of points a team allows per 100 possessions, the basic measure of team defense with pace removed. It carries half the weight inside DEV, the defensive composite that drives WEV. It also anchors two plain regularities in the champion record: 25 of 26 champions had a better-than-average defensive rating, and the fourth contender test asks for a top-10 defense the season before. Lower is better.

Key number: 25 of 26 champions had a better-than-average defensive rating

Related: DEV, Positive-Defense Law, Four Contender Tests, NetRtg

Interior Anchor Rule

Does the team have an interior defender who moves the needle? Every modern champion has had one.

The Interior Anchor Rule checks for one player with a defensive rebound percentage above 18, an on-court net rating above +3 and at least 20 games played. All 24 modern champions in the panel had one. That makes it an entry condition rather than a separator, since most deep playoff teams pass too. The mechanism behind it is rim protection that works by contesting shots, which scales up into team defense. The next step is to replace the yes-or-no rule with a continuous measure of the anchor's defensive quality.

Key number: 24 of 24 modern champions had one

Related: Rim Protection (Defender Frontier / Suppression), Rim Protection to Team Defense Bridge, DEV, RQS

Lineup Hardness

How hard a defensive lineup is to break. Against soft lineups keep forcing decisions; against hard ones, take the shot.

Lineup hardness rates the five defenders on the floor together, using each one's chance of holding up when a possession forces him to decide. Sorting lineups into quartiles by hardness shows a clear gap: in the 2024-25 playoffs, offenses scored 0.928 points per possession against the softest quartile and 0.866 against the hardest, a 6.7% difference across 13,736 lineup-aware possessions. The practical rule flips with it. Against soft lineups, hold the ball and keep forcing decisions; against hard lineups, take the good shot when it comes.

Key number: 6.7% PPP gap, softest vs hardest quartile (0.928 vs 0.866), 2024-25 playoffs, n = 13,736

Full explainer → · Related: Holding-Math Theorem, DTI, Possibility Cost, PPP

Playoff Elevation IndexPEI

How a defender's targeting profile changes from the regular season to the playoffs.

The Playoff Elevation Index tracks each defender's change in Defender Targeting Index from the regular season to the playoffs, asking who holds up and who gets found once opponents have days to prepare. It is the playoff companion to DTI. Because the underlying defender values are being rebuilt and playoff samples are small, the next steps are shrinkage toward the mean, checks on repeat-season stability and an adjustment for opponent mix. Individual PEI values are not published while that work runs.

Related: DTI, NPSS, Lineup Hardness

Positive-Defense Law

Champions defend. Nearly every title team has had a better-than-average defensive rating.

The positive-defense law is a plain count: 25 of 26 champions finished with a defensive rating better than the league average. It is context for the defensive composite, DEV, rather than a tool on its own, because plenty of non-champions defend well too. What it rules out is the useful part: a below-average defense winning the title has happened once in 26 seasons. That history is why the four contender tests include a top-10 defense.

Key number: 25 of 26 champions had a better-than-average defensive rating

Related: DEV, DRtg, Four Contender Tests

Rim Protection (Defender Frontier / Suppression)

Rim protection pays off in the contest, not in scaring drives away: opponents keep attacking the rim and score less when they get there.

Rim suppression measures how much a defender lowers the make rate on shots at the rim compared with what those shooters normally hit. The finding is contest, not deterrence. Across all 30 teams in 2024-25, rim suppression had no detectable link with how often opponents attacked the rim (r = +0.01, p = 0.95), while it tracked opponent points per possession at r = -0.45 (p = 0.013). Suppression carries from season to season at about 0.40, including for defenders who change teams; the mover-versus-stayer difference is not yet resolved.

Key number: r = +0.01 (p = 0.95) between rim suppression and opponents' rim-attempt frequency, 30 teams, 2024-25

Full explainer → · Related: Rim Protection to Team Defense Bridge, Interior Anchor Rule, DTI, DEV

Rim Protection to Team Defense Bridge

Individual rim protection scales up: teams that suppress shots at the rim give up fewer points per possession.

The bridge connects one defender's skill to a team's results. Across the 30 teams in 2024-25, team rim suppression correlated −0.45 with opponent points per possession (permutation p = 0.013). Rim protection is a separator in team defense, not a guarantee on its own: a strong rim protector still needs a defense around him. This link is the middle step of the chain the site follows from the contested rim shot, to team defense, to the defensive composite DEV, to championships.

Key number: r = −0.45 with opponent points per possession (p = 0.013), 30 teams, 2024-25

Full explainer → · Related: Rim Protection (Defender Frontier / Suppression), DEV, Interior Anchor Rule, PPP

Championship Identity ShiftCIS

How the league's shot identity has shifted toward the three, adjusted for era.

Championship Identity Shift tracks the share of field-goal attempts that are threes, adjusted for era, to show how the league's shot identity has moved. As an era diagnostic it is excellent, following the league's trend at r = 0.966. As a title signal it carries nothing, at r = 0.042 with champions. That split is the lesson. The three-point shift changed how everyone plays, champions included, so shooting more threes than your era does not mark a title team. Next up is detecting the turning points in the trend.

Key number: Era trend r = 0.966; champion r = 0.042

Related: xPTS, H3′, CORE

Efficiency Tax

The points a player leaves on the table by choosing lower-value shots when better ones were available.

The Efficiency Tax measures the cost of shot selection. Each shot a player takes is compared with the better shot he could have taken, using league-average values by shot type. A contested mid-range jumper is worth about 0.86 points and a rim attempt about 1.26, so choosing the jumper costs about 0.40 points on that possession. Add those choices up and you have the tax. Some stars beat league-average values on difficult shots, which is what the Sovereign Premium measures. The name once also covered a payroll measure; that lens is now Payroll Fragility.

Related: xPTS, Sovereign Premium, Payroll Fragility, Possibility Cost

Possibility-Adjusted Shot ValuePASV

A per-shot grade: what a shot is worth minus what taking it gives up.

Possibility-Adjusted Shot Value grades each shot as its expected points minus the value of continuing the possession instead. Version one estimated that continuation value from how the league actually plays, so it ended up tracking ordinary shot quality closely; its lasting product is the shot-value ladder, from the dunk at 1.81 down to mid-range at 0.86. The reading that matters is that continuing buys little for offenses that do not force decisions. Version two, pre-registered, measures what a rim-first approach is worth by conditioning on the five defenders on the floor.

Related: xPTS, Possibility Cost, Holding-Math Theorem, DTI

Shot Value (Expected Points)xPTS

What a shot is worth on average, in points. The ladder runs from the dunk at 1.81 down to mid-range at 0.86.

Shot value is the average number of points a shot type produces per attempt, also called expected points or xPTS. Measured on 219,527 attempts from 2024-25, the ladder reads: dunk 1.81, two free throws about 1.52, rim attempt 1.26, corner three 1.17, above-the-break three 1.07, paint shot away from the rim 0.86 and mid-range 0.86. Two things stand out. A shooting foul that sends a player to the line for two is worth more to the offense than the rim attempt itself, and short paint shots are worth no more than mid-range jumpers. Within a zone, shot value does not change with the shot clock.

Key number: Dunk 1.81, rim 1.26, corner three 1.17, mid-range 0.86 points per attempt (219,527 attempts, 2024-25)

Full explainer → · Related: PASV, Efficiency Tax, Possibility Cost, PPP

Research findings and source essay →

Shot-Diet BreadthH3′

Does a varied shot diet travel better into the playoffs? A single-season test said no; a multi-year version is next.

Shot-Diet Breadth asks whether scoring from many zones holds up better in the playoffs than leaning on one or two shot types. Tested one season at a time, breadth showed no consistent link to playoff translation. The live version, H3′, is preregistered: it uses several seasons of shot diet and residualizes them, so one year of noise cannot drive the answer. Until it runs, breadth is a description of style rather than a playoff advantage, and the shot-value ladder remains the better guide to which shots to take.

Related: xPTS, PDR, CIS

Sovereign Premium

Who has earned the right to break the shot chart: players who beat expected points on shots the math says to avoid.

The Sovereign Premium, also called the Sovereign Exception, measures what a player scores on shots the shot-value ladder calls poor choices, compared with what those shots are normally worth. A positive premium means the player turns bad shots into good ones often enough to justify taking them. Nikola Jokić, Jalen Brunson, Devin Booker, Stephen Curry and Kevin Durant all come out positive, and the result holds under two different methods. The spec is not frozen yet; the next steps are to lock the definition and test whether the premium repeats season to season.

Related: xPTS, Efficiency Tax, PSS

Holding-Math Theorem

The payoff for making a defense keep deciding: every forced decision is another chance for it to break.

The Holding-Math Theorem treats each defender as having a chance to hold on every decision a possession forces him to make, so the more decisions an offense forces, the more chances the defense has to crack. The canon version gives each defender his own hold probability from his targeting profile, weighting the on-ball defender 2.5 times and off-ball defenders half. It holds out of sample: fit on three playoffs and tested on two others, t = 4.28 on 8,007 possessions. On 66,006 playoff possessions, a model that knows the matchups is significant (β = +0.945, p = 0.010) while its matchup-blind twin is not (p = 0.78).

Key number: Held out on two playoffs after fitting on three: t = 4.28 (8,007 possessions)

Related: Lineup Hardness, DTI, Possibility Cost, PASV

Option Preservation CoefficientOPC

How long a possession's decision tree stays alive.

The Option Preservation Coefficient measures how long a possession keeps its options open: whether the ball-handler's choices keep several outcomes in play or close them down early. For now it is approximated by assist percentage, which rewards players who keep teammates involved, and Nikola Jokić is the model case. The full version will be written against defender identity, because keeping options alive matters most when the defenders on the floor are the ones being forced to choose.

Related: PSS, Possibility Cost, Holding-Math Theorem

Points Per PossessionPPP

The basic unit of the game: how many points an offense scores each time it has the ball.

Points per possession is the scoring rate of an offense, or the rate a defense allows, with pace removed. It is the common currency across DataDunkNBA. Shot values are points per attempt on the same scale, lineup hardness compares PPP against soft and hard lineups, and the rim-protection bridge links team rim suppression to the PPP opponents score. Multiply PPP by 100 and you have offensive or defensive rating, the building blocks of net rating.

Related: xPTS, Lineup Hardness, Rim Protection to Team Defense Bridge, NetRtg

Possession Sovereignty ScorePSS

Who controls what a possession becomes, on a 0 to 100 scale.

The Possession Sovereignty Score rates how much a player dictates the shape of a possession, on a scale from 0 to 100. Nikola Jokić sits at the top of the current build. The composite's inputs and weights are still being set, and the next step is to freeze them and validate the score on the defender-identity data behind DTI. Until then it is a lens on who runs the floor rather than a measured result, and it pairs naturally with the Option Preservation Coefficient and the Sovereign Premium.

Related: OPC, Sovereign Premium, DTI

Possibility Cost

The value of the options a decision closes off. The idea underneath every DataDunkNBA metric.

Possibility Cost is the founding idea of the stack: every shot or decision carries a price equal to the better options it forecloses. A rim attack forces three or four defensive decisions in about two seconds, and each one is a chance for the defense to break; a pull-up three forces none. That cost is set by the defenders and by the offense's approach, not by the shot clock: within a shot zone, a shot is worth the same early or late. The working doctrine is to force a rim decision every time and cash whichever door opens.

Related: PASV, Holding-Math Theorem, xPTS, Lineup Hardness, Powerball Basketball

Powerball Basketball

A tactical design hypothesis: a set of offensive constraints built around forcing defensive breakdowns.

Powerball Basketball is a design idea for offense: a constraint system built around forcing defensive breakdowns, in the spirit of Possibility Cost and the Holding-Math Theorem. It is a framework, not a measured result. Any breakdown rate attached to it has to be tested on real possessions before it is stated, and that test has not been run. Read Powerball as a coaching hypothesis that the rest of the stack gives reasons to try, with the system kept and every percentage held back until it is measured.

Related: Possibility Cost, Holding-Math Theorem, Lineup Hardness

Availability Haircut

Discount a star's projected wins by how many games he tends to miss.

The Availability Haircut lowers a star's projected win contribution in proportion to his chronic games-missed risk. It is the player-scale version of the payroll finding: a team's wins track how often its highest-paid players are actually on the floor, the strongest single link in the payroll study. The haircut curve itself is the next thing to calibrate, using games missed against realized wins from 2011 to 2025. The lens is being merged with Cap Space Ghost and Payroll Fragility into one roster-economics view.

Related: Payroll Fragility, CSG

Cap Space GhostCSG

The payroll a team spent on players who were not on the floor, as a share of the cap.

Cap Space Ghost turns missed games into dollars: games missed divided by 82, times salary, summed across the roster and expressed as a share of the cap. It is accounting, the bill for availability. The 2026 finalists show what a light bill looks like: the Knicks lost 21% of the cap to missed games and the Spurs 24%, the second- and third-healthiest rosters in the league. Cap Space Ghost is being merged with the Availability Haircut and Payroll Fragility into one roster-economics lens.

Key number: 2026 finalists: Knicks 21% and Spurs 24% of cap, the 2nd- and 3rd-healthiest rosters

Related: Payroll Fragility, Availability Haircut, CORE

Payroll Fragility

Teams win less when their biggest contracts sit out. Availability, not payroll concentration, carries the link.

Payroll Fragility measures how exposed a team's wins are to its highest-paid players missing games. Across 480 team-seasons from 2010-11 through 2025-26, fragility correlated −0.394 with regular-season win percentage, and the sign was negative in all 16 seasons. The driver is top-earner availability, which correlates +0.47 with win percentage. Paying stars is not the problem: payroll concentration correlates +0.28, positive, because good teams pay good players. The cost arrives in games missed. This lens was previously called the payroll Efficiency Tax; that name now belongs to the shot-selection metric.

Key number: r = −0.394 with regular-season win%, negative in 16 of 16 seasons (480 team-seasons)

Full explainer → · Related: CSG, Availability Haircut, Efficiency Tax, CORE

Research findings and source essay →

Shelf Life ScoreSLS

How long a max contract will stay worth paying: a weighted read on archetype, age, efficiency and mileage.

The Shelf Life Score estimates how long a big contract will keep paying off. It blends four things: the player's archetype (30%), age (25%), efficiency (30%) and career mileage (15%). The idea is that some player types age gracefully while others fall off sharply, and a front office should price that before signing, not after. Whether the score does that is an open question. Its first test used too few contracts to say anything, so the next step is to widen it to every max deal from 2011 to 2025 before attaching any result to it.

Related: SSR, VCR, Availability Haircut

Value-to-Contract RatioVCR

Production per dollar: winning value divided by salary.

The Value-to-Contract Ratio divides winning value by salary in millions, so it shows who returns the most for what he costs. Its clearest lesson is that rookie-scale deals are the best value in the league, which is why teams prize them. The numerator is being pinned to a single WEV version, and the ratio will then be tested forward on 2026-27 contracts. Treat current readings as a way to frame a contract rather than a verdict on it.

Related: WEV, SLS, Payroll Fragility

Draft Power Law

Draft value is extremely top-heavy: a handful of picks produce most of it, and most picks never clear replacement level.

The Draft Power Law describes how value is spread across the draft. The top 5% of picks produce 49.3% of all draft value, the Gini coefficient is 0.845, and 60% of picks never clear replacement level. Prediction has a ceiling too: what is known before the draft accounts for only about 27 to 30% of the variation in careers. The lesson for front offices is that the draft is a hunt for rare outliers, and landing one is worth more than everything else in a typical class.

Key number: Top 5% of picks = 49.3% of all draft value; Gini 0.845

Related: DVOE, Draymond Effect, Steal Tell

Draft Value Over ExpectedDVOE

Career value above what a player's draft slot would predict.

Draft Value Over Expected compares a player's career value over replacement (VORP) with the value a typical player taken at the same slot produces. A positive DVOE means a team found more than the slot usually yields; a negative one means less. It turns the Draft Power Law's distribution into a grade for individual picks. The spec is clean, and the first real grades arrive as the 2026 class matures, since careers take years to show their value.

Related: Draft Power Law, Draymond Effect, Steal Tell

Draymond Effect

Champions keep finding winning rotation players who were drafted 20th or later, or not drafted at all.

The Draymond Effect flags a rotation player drafted 20th or later, or undrafted, who posts an on-court net rating above +3. Twenty-two of 24 modern champions had one. Stars come overwhelmingly from the top of the draft, with a 41% star rate for picks 1 to 5 against 3% for picks 31 to 60, so finding a winning player late is how champions fill out a roster around them. Most good teams pass this test too, which makes it a roster-construction signal rather than a separator. The rotation-minutes threshold is the next piece to pin down.

Key number: 22 of 24 modern champions had one

Related: Draft Power Law, DVOE, RQS

Steal Tell

Steal rate as a draft signal.

The Steal Tell uses a prospect's steal percentage, the share of opponent possessions he ends by taking the ball away, as a draft signal. It was applied in draft writing on the 2026 class, where it pointed against parts of the consensus at the top of the board. It is an applied lens so far and does not yet make a rule. The next step is a written spec: what steal-rate threshold, against what level of competition, and tied to which career outcome.

Related: Draft Power Law, DVOE