Will we make it out of the 1st round? ESPN says...........

Did you read the article?

My original comments were directed at KnickerBloggers prediction that the Mavs would beat the Blazers - based solely on regular season point differential. He twice mentioned that the Blazers were only 13th in point differential and will, therefore, lose to the Mavs who are 8th in point differential - that was his sole basis for the Mavs beating the Blazers in a first round series. He ignored recent performance (Blazers +6.05, 5th best and Mavs +3.71, 10th best, over the last 25% of the schedule, recent head-to-head match-ups, late season acquisition of Gerald Wallace, etc.) and based his entire prediction on the fact that the Mavs have the higher regular season point differential, and would therefore win the season.

33% correlation isn't that great, and considering a single variable as a accurate predictor in a multivariate problem is fraught with problems. If he's going to use point differential, he should at least weight it toward recent performance. The way teams are performing in April should carry more weight than the way they performed in November - especially when roster changes are involved. I'd also consider recent performance against quality opponents (other playoff teams), ability to win close games (fewer blowouts in the play-offs), etc.

Is there any single predictor of playoff success? No, there isn't. That was my point and why I disagreed with KnickerBlogger's prediction based on a single variable (especially one that considers November's data every bit as important to post season success as April's).

BNM

 

Thanks for expanding on your post. I agree with what you're saying here. I was just making the point that dismissing an indicator that, by itself, has a 33% accuracy, probably isn't a wise thing to do if one can't put forth a more accurate model. It would be interesting to see how things change when an EMA is applied to the pt differential instead of a straight average.