- APEX— Adjusted Per-Possession Efficiency Index
- The composite player rating produced by the APEX Hoops pipeline. APEX combines stint-level plus-minus with leverage weighting, per-player recency adjustment, and teammate/opponent strength normalization to produce a single per-36-minute number per (player, position).
- APEX/36
- APEX expressed as net points per 36 minutes of play. League average is ≈ 0. Star players cluster +3 to +7. The metric is positional, but each player is rated at exactly one position — the highest-APEX position where they have 2,000+ minutes.
- Stint
- A continuous span of game time in which the same 10 players are on the floor. APEX is built from millions of stints across every regular-season, play-in, and playoff game since 2003-04. Each stint has a duration, a point differential, and a leverage value.
- Leverage adjustment
- An additional weight applied to high-leverage minutes — close games and late shot clock. Garbage-time minutes count for less; every season type (regular season, play-in, playoff) is weighted on the same per-stint scale.
- Recency adjustment
- Per-player tapered weighting that makes recent seasons count more than older ones. The taper is calibrated per-player based on stint volume, so high-minutes veterans don't have their entire career averaged equally with their current form.
- Teammate/Opponent adjustment
- A simultaneous regression that corrects for lineup strength. A player who logged minutes against bench units has their raw plus-minus deflated; a player who consistently faced starters has it inflated.
- Per-36
- All APEX bridge metrics are normalized to a 36-minute pace so high-volume starters and low-minute role players can be compared on the same scale.
- Position assignment (body-size rule)
- Each stint is assigned positions by lineup geometry, not by NBA.com listings. Within a 5-player lineup, players are ranked by height × weight: the single smallest body = G (Guard); the single largest body = C (Center); the middle three = W (Wing). Tiebreaks on equal height × weight: taller, then heavier, then higher career RPG (so the bigger man slides toward C and the smaller toward G). This is scheme "3pos_v6_g1w3c1".
- Qualified minutes
- A player must have logged at least 2,000 career minutes at a position for that position to qualify. A player is rated at their single highest-APEX qualified position (any other positions they've played render muted for context). Positions with 2,000+ minutes are also eligible for that slot on the Optimal depth chart. Below the threshold, a player is shown italic/grayed ("Unqualified") and excluded from team weighted-average composites.
- Tier scale
- Players are grouped into 7 tiers by per-player APEX rank: First Team (1–5), Second Team (6–10), Third Team (11–15), All-Star (16–24), Starter (25–150), Rotation (151–240), Reserve (241+). Tier boundaries are recomputed daily from active qualified players.
- Bridge metrics
- The four intermediate values that decompose an APEX/36 row: raw plus-minus per 36, leverage adjustment, recency ("age") adjustment, and teammate/opponent adjustment. They sum (with weights) to the final APEX/36. Bridge metrics are shown on the leaderboard, on every player page, and in CSV exports.
- Plus-minus
- Point differential while a player is on the floor, rate-normalized to per-36. APEX uses possession-weighted plus-minus from stint data, not the box-score per-game plus-minus reported by NBA.com.
- Active player
- A player whose name appears on the current NBA.com active roster (CommonAllPlayers with is_only_current_season=1) AND who has logged at least one stint in the current season. Retired players keep their historical APEX rows but are flagged inactive and excluded from tier-boundary computation.
- DARKO (DPM)— Daily Plus-Minus
- A daily-updated, machine-learning impact metric from darko.app (Kostya Medvedovsky) that blends box-score and on/off data through a Kalman-filter model. Among the catch-all metrics most preferred by NBA front offices in recent surveys. Unlike DARKO, APEX uses no box-score prior and adds leverage weighting plus a per-player recency taper.
- EPM— Estimated Plus-Minus
- An all-in-one impact metric from Dunks & Threes (Taylor Snarr) that combines box-score, on/off, and player-tracking data under Bayesian priors. One of the public metrics NBA front offices trust most. APEX drops the box-score prior entirely and adds leverage weighting and a per-player recency taper.
- LEBRON— Luck-adjusted player Estimate using a Box prior Regularized ON-off
- An impact metric from BBall-Index built on luck-adjusted RAPM with a box-score prior and role stabilization. APEX keeps the plus-minus core but uses no box-score prior, and weights each minute by game leverage with a per-player recency taper.
- RAPTOR— Robust Algorithm using Player Tracking and On/Off Ratings
- A box-score, player-tracking, and on/off hybrid published by FiveThirtyEight. Like the other hybrids it regresses toward RAPM using box-score inputs; APEX uses no box-score prior and adds leverage weighting and a per-player recency taper.
- RPM— Real Plus-Minus
- ESPN's Bayesian impact metric (Jeremias Engelmann, Steve Ilardi) that blends box-score and on/off data with priors. APEX shares the plus-minus core but uses no box-score prior and weights minutes by game leverage.
- RAPM— Regularized Adjusted Plus-Minus
- Pure plus-minus impact estimated by ridge regression over every lineup stint, with no box-score inputs — the statistical backbone most public hybrids regress toward. APEX shares this no-box-score-prior philosophy but adds leverage weighting and a per-player recency taper.
- BPM / VORP— Box Plus-Minus / Value Over Replacement Player
- Basketball-Reference's box-score regression (Daniel Myers) trained to approximate RAPM from the stat sheet; VORP scales BPM by playing time. Both are box-bound, so they can't capture impact the box score misses. APEX is built directly from stint-level plus-minus.
- On/Off Net Rating
- A team's point differential per 100 possessions with a player on the floor versus off it (NBA.com, Cleaning the Glass, PBP Stats). It is uncontrolled for teammates and opponents, so it conflates a player with their lineup. APEX's simultaneous teammate/opponent solve isolates the individual signal.