Sunday, October 4, 2015

How well did the ratings predict individual matches at USTA League Nationals this weekend?

When Nationals comes around, it is always fun to take a look at how the my Estimated Dynamic NTRP Ratings predict matches.  So with one weekend of play at 3 levels complete, here is a first look.

For this analysis I'm taking a look at the semis and final for each level and gender and looking only at completed matches.  These are matches players should be going all out in, but by ignoring any retirements the analysis is not skewed by results where the better player retired because their team had already won.

In the Women's 3.0 level, one semi predicted 3 of the 4 completed matches, and in the other semi 5 of 5 were predicted correctly.  In the final, all 5 courts were also picked correctly.  Going 13 for 14 isn't too bad.

The Men's 3.0 level saw one semi have only 2 of 5 picked correctly, but each of the misses were on essentially pick'em matches where the rating difference was less than 0.05.  The other semi had 4 of 5 picked correctly.  The final saw all 5 courts picked correctly.  So this total 11 of 15.

The 4.0 Women's first semi, 4 of 5 were picked right, the one miss having a difference of less than 0.01, the second semi had 4 of 5 as well, and the final had 4 of 5 picked correctly to boot.  So that is 12 of 15.

For the 4.0 Men, the first semi had just 2 of 5 picked right, the other had 3 of 5, one miss being a near pick'em, and the final had 5 of 5 picked correct.  This is 10 of 15 right.

The 5.0+ Women had one semi with 2 of 3 correct, the miss having a difference of just 0.06, the other was 3 of 3 correct, and the final was 3 of 3.  This is a total of 8 of 9 correct.

For the 5.0+ Men, one semi had 2 of 3 predicted correct, the one miss having a difference of just 0.01, the other semi just 1 of 3, and the final 3 of 3.  This is 6 of 9 correct.

What does this all mean?  While ratings are not perfect at predicting matches, they do give a very good indication of who the stronger player/pair are and who is likely to win.  In the case of the semis and finals this weekend, they predicted 60 of 77, or 78%.

I'll continue to monitor this the rest of this month and report back.