From DARKO.app – this is the #1 stat for predicting a player’s current ability (rather than retroactively evaluating an individual season).
Pretty dramatic recent jump. +2.5 defense is to be expected, but +2.3 offense is surprisingly high
by tinkady
@San Antonio Spurs
7 Comments
And 5th in DRtg, despite how bad the team around him has been overall.
By this time next year he’ll be #1
[This thread with the creator of DARKO explains why that’s probably a mirage](https://twitter.com/ShipTheJustice/status/1747330476954493365). He will likely add a fudge factor before Victor can get to #1. Wemby is obviously better than his on/off so far and the model is capturing some of that, but he is just too much of an outlier as a player compared to everyone else in the database for the machine learning model used by DARKO for the box score component to handle properly at the moment. Normally this is counteracted by the +/- component, but the sample size for his +/- isn’t very large right now so it’s not weighted very high compared to the box score component. Failure on out of sample data is a general issue with this kind of machine learning model.
In Wemby’s case, less sophisticated models like EPM and LEBRON (and of course time decay RAPM) that use more standard regression models and simple manually justified adjustments for things like luck and context are probably more accurate currently. He is good, and *very* good for a 19-20 year old rookie, but he is not close to the best player in the league in his rookie season. In particular it’s his offense that the model is probably way too high on–all the models agree that his defense is DPOY level.
HMMMMMM I DONT SEE CHET IN THIS TOP 10 SURROUNDED BY ALL THE BEST PLAYERS IN THE LEAGUE
I am hoping he can make all nba defense this year.
With probably one of the worst perimeter defenses surrounding him. Imagine him with, say, Derrick White and Jrue Holiday funneling him half-dead meat every other possession?
This stat is broken if nikola is number 1