For each team stat I calculated an average between the team's stat and their opponent's opponent stat. This year's Lakers are a good example of why I did this. If you look purely at their stats, they would be predicted to lose every match up because their lower pace tends to lead to lower stats but when you take into account how teams perform against their defense, they look dominant.
I feel that this new approach is way better, however, according to the math, it is actually worse. The adjusted R squared still suggests that the model only explains less than 1/3 of the variation. I probably could have achieved the same R squared as last year by including GM control variables again but I ran out of time and patience this week so we'll save that for next season.
Looking at the predicting variables this year, they are of the same weird variety as last year, point margin (obvious), blocks (probably skewed by Kemp and Zo), and Free Throw Percentage. Free throw percentage doesn't seem to make sense but it may explain why Shaq teams have never been able to go all the way or vice versa. So, while I'm not super excited about the predictive power of this model, let's see what it has to say about the upcoming playoffs.
- 2001 playoffs.GIF (28.59 KiB) Viewed 523 times
-The model seems to think this year's postseason lineup is pretty top heavy and is predicting few upsets BUT. . .
-The entire first round of the Western Conference playoffs is a tossup (with the exception of the Lakers) so these predictions are very sketchy
-The Hawks are still predicted to easily walk to the championship. From my two years of looking at the data, I feel like teams that can score big are favored over defensive teams. . . unless you're coached by Balls, whose game planning skills are even more prominent in the playoffs
So there you have it. I'll post some updates as we go along. Best of luck to all the playoff teams.
