Another look at what NTRP rating combinations works best in Adult 55 & Over? More Interesting Tennis League Stats

A few days ago I took at look at what ratings combinations win the most in 55 & Over Adult league play.  What I found was consistent with earlier analysis I've done and that is that unbalanced pairs, e.g. a 4.0/3.0 pair in 7.0, beat balanced pairs, e.g. a 3.5/3.5, more than they lose.

What some pointed out, and I knew I wanted to do more research into, is that while the unbalanced pairs win more, that may be because of a selection bias of sorts, where it is possible that only stronger players at their level opt to play as an unbalanced pair, and the fact they win more often may be expected.  It is even possible that while they win more than lose, they may not win as often as expected.

For example, at 7.0, it may be that it is the strong 3.0s that are attracted to play or perhaps are recruited to play.  That means when they (say a 2.90 3.0) partner with an average 4.0 (3.75) vs two average 3.5s (3.25), the rating comparison is 6.65 vs 6.50, and the unbalanced pair has the higher combined ratings and thus is arguably expected to win.  Even if the 3.5s are a bit better (3.30) it is still 6.65 vs 6.60.

With that in mind, I went about digging deeper.  Specifically, I narrowed the matches analyzed to only those where I had an established rating for all 4 players and then looked at not just how often unbalanced pairs won, but how often they were expected to win.  The results were certainly interesting, if not a bit surprising.

Note, the actual winning percentages are slightly different from the earlier post because some matches where the players didn't all have ratings have been filtered out.

For each level/gender below, I'm using 2022 55 & Over Adult league data again and showing the percentage of the time the unbalanced pair won, followed by the percentage of the time the unbalanced pair had the higher combined ratings.  When the actual is higher than the expected, the unbalanced is overachieving, and when the actual is lower than the expected, they are underachieving.

The ratings used for this analysis are my Estimated Dynamic NTRP Ratings I use for doing the various individual, team, and flight reports I offer.

For the women:

• 6.0 - 3.5/2.5 vs 3.0/3.0 - 48.7% actual, 54.7% expected - underachieved
• 7.0 - 4.0/3.0 vs 3.5/3.5 - 54.6% actual, 54.5% expected - nearly perfectly as expected
• 8.0 - 4.5/3.5 vs 4.0/4.0 - 56.2% actual, 62.3% expected - underachieved
• 9.0 - 5.0/4.0 vs 4.5/4.5 - 66.9% actual, 73.7% expected - underachieved

Here we see the selection of matches change the actual percentages slightly, and at 6.0 the unbalanced actually lost slightly more than they won, but at three of the four levels, while the unbalanced won more, they were expected to win more than they did, so arguably underachieved.

For the men:

• 6.0 - 3.5/2.5 vs 3.0/3.0 - 58.6% actual, 58.6% expected - exactly as expected
• 7.0 - 4.0/3.0 vs 3.5/3.5 - 55.6% actual, 51.1% expected - overachieved
• 8.0 - 4.5/3.5 vs 4.0/4.0 - 55.4% actual, 56.3% expected - slightly underachieved
• 9.0 - 5.0/4.0 vs 4.5/4.5 - 50.9% actual, 55.2% expected - underachieved

Here the actual percentages vary a bit more with the selected matches, especially at 9.0.  And we also see that at 6.0 it is exactly as expected, the unbalanced overachieve at 7.0, but 8.0 has a slight underachieve and 9.0 a significant underachieve.

So, it may not be quite as clear cut as the earlier analysis showed.  While unbalanced pairs do win more than they lose, it appears it is because it is the right unbalanced pairs are playing and they are expected to win because they are higher rated overall, not just because they are an unbalanced pair.  And they may not even win as often as expected.

What do you think?  Is this more in-line with your experience than the initial analysis was?