Descriptive only — a measured pattern, no edge claimed.
Analytics module · as of 2024-25 (Elo end-of-season) x 2024_25 on/off slice
Star Removal: what one player is worth in win-prob points
Take Nikola Jokic off Denver and, on a neutral floor against a league-average team, the Nuggets' win probability falls from 0.634 to 0.0517 -- a 0.5822 swing, the largest of any team's top on/off player.
confirmednull (a finding)not testabledescriptivepending
Cf Star Removal
scripts/platformkit/analytics_showcase/out/cf_star_removal.json·2024-25 (Elo end-of-season) x 2024_25 on/off slice
What it means
Each team's biggest on/off swing player is mapped onto the Elo win-prob curve to ask a blunt what-if: how much of this roster's strength walks out the door with that one name. Jokic's swing is enormous because Denver's on-court net rating collapses without him and the logistic curve turns that into a near-total win-prob wipeout. The numbers are a ceiling, not a clean causal player value -- a full-lineup on-vs-off gap gets attributed to one man.
Caveats & confounds
CEILING: the on/off net-rating delta overstates marginal individual value -- it is a full-lineup swing with no control for teammates, opponent, coach trust, or garbage time (roster confound), and the net-rating-to-Elo conversion is a single frozen constant, not a fit. DESCRIPTIVE_ONLY, not a causal or edge claim.
Method.
rating to prob curve
repo Elo logistic 1/(1+10^(-d/400))
star rating removed
on_off_net_rating_delta * ELO_PER_NETRATING_POINT
elo per netrating point
25.3333
elo per netrating point anchor
elo_config.ELO_HFA=76 Elo ~ 3.0-pt home edge (76/3.0)
