Tuesday, February 17, 2015

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

I've been doing sub-flight previews that show how teams stack up by comparing the average Estimated Dynamic NTRP Rating of the players on their roster.  While using average rating doesn't necessarily tell you who will win a sub-flight, for those that I've looked at after matches have been played, for the most part the higher average rating teams are at or near the top.

From this we could reasonably support the notion that higher rated players will generally win more often over lower rated players.  This shouldn't be a big surprise, but it gets more interesting if we dig deeper.

One way to look at how accurate dynamic ratings are is to see how well they predict individual matches.  I did this a little bit looking at some Nationals matches last year, but I've looked at more matches and here is how they do.

Here is what percent of the time the higher rated player/doubles team won a match grouped by the gap between teams looking at matches during the 2014 USTA year.

GapWinning %
0.00 - 0.0553%
0.05 - 0.1563%
0.15 - 0.2575%
0.25 - 0.3584%
0.35 - 0.4590%
0.45 - 0.5593%
0.55 - 0.6595%
0.65 - 0.7596%

No big surprise here.  If players are rated about the same, the match is just about a toss-up.  The bigger the gap, the more likely it is that the higher rated player(s) win.

What may be surprising is how quickly the winning percentage goes up.  With a gap of just 0.2 the higher rated player(s) win 75% of the time.  And by the time the gap gets to 0.4 the winning percentage is 90%.

The above stats are from 2014, but I'll be looking at 2015 soon.  And one can imagine that it is important to get the right match-ups in matches as well and I'll be writing about that.

Stay tuned!