Thursday, February 19, 2015

How accurate are 2015 Dynamic NTRP ratings at predicting tennis matches? Interesting League Stats

I recently wrote about how well dynamic NTRP ratings do at predicting the winners in USTA League matches by looking at my Estimated Dynamic NTRP Ratings and match results during the 2014 USTA season.  While it is still early in the 2015 season, I thought I'd take a look to see if the percentages have changed.

Here is the same table from 2014 with the 2015 percentages thus far added in.

Winning %
Gap20142015
0.00 - 0.0553%53%
0.05 - 0.1563%62%
0.15 - 0.2575%73%
0.25 - 0.3584%83%
0.35 - 0.4590%89%
0.45 - 0.5593%93%
0.55 - 0.6595%94%
0.65 - 0.7596%96%

We can see that there are no major changes, but 2015 is a hundredth or two lower nearly across the board.  My guess is the reason is that the USTA made some significant adjustments at year-end last year, mostly bump ups, so some players may have ratings that are perhaps a little inflated which has reduced the winning percentages accordingly early in the year while the algorithm attempts to adjust players back to where they should be.

What do you think?  I'll try to update this later in the year.