Over the course of the past week, there have been another 58 NBA games that have been played, bringing the total number of games played to 151; over ten percent (12.28%) of the season complete. To date, we still have one undefeated team (Golden State, 11-0) and one winless team (Philadelphia, 0-10).
Using the beta regression model, we are able to build an updated power rankings. The pseudo r-squared value is an exceptionally high 0.7223, indicating a relatively strong fit for modeling outcomes of games.
We see the beta model fits relatively well. Retroactively fitting the model to the 151 games, if the model is above 0.50, then the leading team (alphabetical order) in the data matrix is predicted to win. The blue value is the observed result of the game, relative to the first team. Here, we predicted the winner of every game correct. This is fairly expected due to the high r-square value. However, since the r-square value is only about 0.7, we cannot predict the actual scores relatively well.
The resulting rankings from the regression models gives the Golden State Warriors the top spot, but there are a lot of massive changes in the top ten. We see San Antonio and Boston jump into the top ten after a couple key wins: San Antonio went 3-0 for the week with a key win against free-falling Portland; Boston went 3-1 but scored two big wins over Atlanta and Oklahoma City.
We see the high volatility of the rankings due to the small percentage of games, combined with the fact that not every team has had a chance to play each other. As the next few weeks play out, we expect these rankings to begin to stabilize.
Reblogged this on Mon site officiel / My official website.
LikeLike