How accurate are Dynamic NTRP ratings at predicting tennis matches? 2022 edition - Interesting League Stats

It has been awhile since I took a look at how accurate my Estimated Dynamic Ratings are at predicting match results so I thought I'd take a look at how they did in 2022.

As a reminder, if you are interested in getting a detailed report using my ratings, see this post.

For this analysis, I'm looking at all matches played in 2022 in the 18 & Over, 40 & over, and 55 & Over divisions.  Unlike before, I'm going to slice the data a few different ways.

Note, these are my dynamic ratings, that largely mirror what the USTA does with their ratings but are not the USTA's ratings, so any accuracy or claims regarding predicting matches is for my ratings and not the USTA's.

First, here are the winning percentages by the higher rated player/pair across all the divisions for all matches, and then broken down by the difference between the ratings as well as singles and doubles.

Rating Gap	Overall	Singles	Doubles
Overall	72%	73%	72%
<= 0.05	54%	53%	54%
0.05 - 0.15	66%	64%	66%
0.15 - 0.25	79%	76%	80%
0.25 - 0.35	88%	85%	89%
0.35 - 0.45	93%	89%	94%
> 0.45	96%	94%	97%

These percentages are remarkably consistent with the past analysis I did.  The matches between closely rated players are close to 50/50 but the higher rated player/pair does win a little more often.  As the gap grows, the higher rated player/pair wins increasingly often.

Interestingly, the doubles percentages are generally slightly higher.  Perhaps that is due to more players on the court making it less likely that one player being really on or off has as much effect.

Taking a look at just the 18 & Over division:

Rating Gap	Overall	Singles	Doubles
Overall	73%	73%	72%
<= 0.05	54%	53%	54%
0.05 - 0.15	66%	63%	66%
0.15 - 0.25	79%	76%	80%
0.25 - 0.35	87%	85%	89%
0.35 - 0.45	92%	89%	94%
> 0.45	96%	95%	97%

These percentages are nearly identical to the overall, which makes sense since 18 & Over is the largest division and most matches.

Then just the 40 & Over division:

Rating Gap	Overall	Singles	Doubles
Overall	72%	73%	72%
<= 0.05	54%	54%	54%
0.05 - 0.15	66%	65%	66%
0.15 - 0.25	80%	76%	80%
0.25 - 0.35	88%	85%	89%
0.35 - 0.45	93%	90%	94%
> 0.45	96%	94%	97%

Very similar percentages here too, just slightly higher at the larger gaps.

And last, just the 55 & Over division:

Rating Gap	Overall
Overall	72%
<= 0.05	54%
0.05 - 0.15	66%
0.15 - 0.25	81%
0.25 - 0.35	91%
0.35 - 0.45	95%
> 0.45	98%

This division is doubles only, and these percentages are basically the same as the doubles stats for the other divisions.

So it appears the ratings predict match results very consistently regardless of division, which should be the case for a good rating system.  So perhaps my ratings are sound and a good rating system!

What do you think?

Dissecting the USTA"s 2022 tennis participation press release

Each year, the USTA has a press release or post discussing a variety of participation numbers.  This year was no different with the headline being that participation grew for a third straight year.  There are many ways to look at tennis participation, so this deserves a closer look to see what exactly grew.

Those that are regular readers of this blog know that I take a look at USTA League participation each year and there has been a general decline in the number of players playing in the main 18 & Over, 40 & Over, and 55 & Over leagues.  However, depending on how you look at 2020/2021 since league play was suspended and there were no Nationals in 2020, there has been some growth in league participation after a big drop due to COVID.  Regardless of how you look at it though, even if participation in 2022 was up from 2021, 2022 was down from 2019 (258K down to 242K) and on a similar or even quicker trajectory to prior years.

So how can the USTA claim participation grew for three straight years?  The answer is that they are using a different metric than league participation.  The number they are using is from the Tennis Industry Association (TIA) that is estimating the number of people that play tennis in the US irrespective of being part of any organized league or tournament.  The TIA is saying that 23.6 million people played tennis in 2022, up 5.9 million from the start of 2020.  I believe the TIA bases their estimate on reported sales of tennis balls or other similar sales metrics, and much of the increase is attributed to COVID and how tennis was a safe sport to play and the population has flocked to it.

You'd think that an additional 5.9 million players (33% growth) over three years would result in a larger group of players that would be interested in USTA League play, but my numbers show a decrease from 2019 of 6.2%.  Now, if the TIA's numbers are correct, the percentage of the playing population that plays USTA League is barely over 1%, so one wouldn't necessarily expect a strong correlation.  Clearly, the vast majority of tennis is not played as a part of USTA League.

That said, the USTA stated that 284K players played USTA League in 2022 and that is different from my number of 242K.  Why is that?  My number is looking at just the main Adult leagues as that is something I can do consistently from year to year.  I'm assuming the USTA's number is using all leagues and so it is larger.  By my count the number of unique players in all leagues is just over 284K as well, so I have no argument with that number and it appears accurate.

The USTA also touts 2022 having an 11% increase over 2021 and that too seems accurate by my numbers as my main Adult league count went up just over 10% and these are very close.  And 2021 was up over 2020 which is no surprise as 2020 was way down due to COVID.  It is likely that at least some of the growth in 2022 was due to 2021 still being down due to COVID and more players coming back.

The key thing is that 2022 is still short of 2019.  Whether that is because the general decline in USTA League participation has continued, or we are still suffering from the effects of COVID remains to be seen.  One would think that 2022 didn't have many players still staying away from league play due to COVID but perhaps there were some.  The league participation numbers for 2023 will go a long ways to answering the question of if the pre-2019 decline is continuing, or if the USTA can capitalize on the additional tennis playing population and convert some to league play.

While writing this blog, it occurred to me that there is another metric that could be tracked, and that is the number of players that receive a year-end rating each year.  This will include more players than my main Adult league metric as it will include Mixed and Tournament players as well as players that play in alternate leagues that count for ratings.  It won't count players that play Combo or other leagues that don't count for ratings, won't count tournament players that play non-sanctioned or non-NTRP tournaments, and won't include players that don't get the minimum three matches played.

The result of this is a metric the USTA doesn't want to publicize at is falls well short of some of the aforementioned metrics.  I show just 231K players that had a 2022 year-end rating and this was down 8.0% from the 2021 number of 251K.  This is well below the USTA's 284K and below my 242K as well.  This means that just over 50 thousand league players play leagues that don't count for ratings, or didn't play enough matches to get a year-end rating.

Note that 2021 was up 2.7% over 2019's 245K, but keep in mind that 2021 covered a 2 year period which would inherently inflate the number.

Here is the full chart for the past 10 years.

Here we see the slow but steady decline with a slight bump up in 2019 and another in the 2-year period 2021, but a drop to a level lower than any other year in 2022.  And the drop is a similar percentage for both women and men.

So, was there growth in general tennis participation?  Yes, if you believe the TIA.  But has that transferred to adult play in USTA League and those players getting year-end ratings?  It doesn't appear so, at least yet.

Will the significant growth the TIA reports in the tennis playing population find its way to USTA League play in 2023?

A summary of how USTA League 40 & Over 4-court format is encountering 2-2 ties

The 40 & Over 4-court format has been around for a few years and I've written about it numerous times, but a friend played a match last night ending in a 2-2 tie that went all the way to the court 1 winner to decide the team win, which was a healthy reminder of the format and how it works.

With that in mind, I thought I'd summarize what we've observed from league play using the format the past few years.

First, in league play, in the 2021 championship year, of the 27,795 team matches played, 7,215 (26%) were tied 2-2 with 2,934 (41%) of those being tied on sets and of those, 249 (8.4%) were tied on games and went to the winner of court 1 doubles.

In league play, in the 2022 championship year, of the 33,694 team matches played, 8.766 (26%) were tied 2-2 with 3,443 (39%) of those being tied on sets and of those, 315 (9.1%) were tied on games and went to the winner of court 1 doubles.

Nationals is where it can really come into play where the tie-breakers decide important matches.  While the format was put in place for 2020, there were no Nationals that year so we start this summary with 2021.

At 2021 40 & Over Nationals, there were 292 team matches played with 84 (29%) of them ending in a 2-2 tie.  Of those, 31 (37%) were tied on sets, and of those, 3 (9.7%) were tied on games and went to the winner of court 1 doubles.

At 2022 40 & Over Nationals, there were 298 team matches played with 72 (24%) of them ending in a 2-2 tie.  Of those, 25 (35%) were tied on sets, and of those, 4 (16%) were tied on games and went to the winner of court 1 doubles.

If we summarize it all then, we get:

Season	Tied 2-2	Sets Tied	Games Tied
2021 Regular Season	26%	41% (11%)	8.4% (0.90%)
2021 Nationals	29%	37% (11%)	9.7% (1.0%)
2022 Regular Season	26%	39% (10%)	9.1% (0.93%)
2022 Nationals	24%	35% (8.4%)	16% (1.3%)

In the above I'm showing the same percentages from the paragraph, plus in parenthesis I'm showing the percentage of all team matches played as well.

This shows a pretty consistent trend of a quarter or a bit more of team matches ending in 2-2 ties, around 10% of all of them being tied on sets, and around 1% of all matches being tied on games and being decided by the court 1 winner.  But 2022 Nationals showed a bit of an uptick in those decided by the court 1 winner, but that is a pretty small sample size so may be due to that.

And for what it is worth, while 2023 is still young, so same percentages seem to be holding.

What do you think?  Are these percentages of matches that come down to tie-breakers significant enough that you think it warrants going back to a 5-court format?