It seems another statistical modeling approach is a big fan of the Boston Celtics this season, joining FiveThirtyEight in seeing Boston among the NBA’s elite teams of the 2020-21 NBA season.
The DARKO (Daily Adjusted and Regressed Kalman Optimized) projections, a self-described “machine learning-driven basketball player box-score score projection system” was created by Kostya Medvedovsky and applied by Andrew Patton according to its home website. It was recently combined with the FiveThirtyEight playoff rotations with DARKO updated player ratings by Medvedovsky, and yielded some very bullish data on the Celtics, ranking them fourth overall in net rating with a score of 9.8.
Took the 538 playoff rotations and combined them with DARKO's updated player ratings (coming online shortly!) to get DARKO's estimated strengths of each team. DARKO can't quit the Bucks (which seems wrong); also suggests selling the Nets. pic.twitter.com/UidY5A4zgk
— Kostya Medvedovsky (@kmedved) January 21, 2021
That was good enough to put Boston behind just the Milwaukee Bucks (13.9, and yes, they are also confused about that), the Los Angeles lakers (12) and Clippers (10).
Interestingly, it has the new-look Brooklyn Nets just ninth at a net rating of just 7.2. What does this all mean?
According to Medvedovsky, it highlights the “estimated strengths of each team” according to DARKO, which was created “using … classical statistical techniques and modern machine learning methods.
For a more detailed explanation, be sure to check the model’s page.
.png?w=600)
