Descriptive only — no edge claimed. These are career per-36 rates, not projections.

NBA players · NBA · as of 2026-04-12

Gary Harris

Descriptive card · conservative measured rates.

seasons played

Higher than 64% of 482 in pack

games

148

Higher than 49% of 482 in pack

minutes

2625.4

Higher than 42% of 482 in pack

PTS /36

8.7

Higher than 3% of 482 in pack

REB /36

2.9

Higher than 1% of 482 in pack

AST /36

2.3

Higher than 28% of 482 in pack

FG %

42.7

Higher than 25% of 482 in pack

3P %

37.4

Higher than 74% of 482 in pack

FT %

73.4

Higher than 31% of 482 in pack

Bars show where this card sits within its own pack (482 nba players), by simple rank of the measured value. Higher is only higher -- no quality judgement is implied.

Also measured elsewhere

Joined onto 330 of 482 player cards; the leaderboard badges cover far fewer.

Team win rate with/without this player

MIL: active 0.3125 vs missed 0.5000 (delta -0.1875)

n_active=48, n_missed=32, ci95_offset=[-0.2173, 0.2173]

WITH/WITHOUT = ROSTER CONFOUND: this is NOT a causal player-impact estimate. Why a player missed games is entangled with injuries clustering, rest-game scheduling, blowout pulls, and a different replacement lineup -- none controlled for. DESCRIPTIVE_ONLY, not a predictive or edge claim.

Consistency profile -- least_consistent_top15 #3

composite_cv_shrunk=0.8818

games=125

NOT a skill/value metric -- a low-volume role player can be very consistent at low output; NOT opponent/rest/role/home-away adjusted -- raw game-to-game dispersion only; NOT predictive -- describes past games, does not forecast future variance

Form curve mover -- top_movers_fallers #15

delta=-6.793

n_qual_games=131

NOT a live hot/cold streak or predictive signal (net-change over the whole span); NOT BPM/EPM/RAPM/DARKO (box-composite only, same limits as box_value_index.py); NOT age/opponent/role adjusted (raw box rates only); NOT player_id-deduped (accented names split across ids upstream -- carried through)