If this is what statisticians do, it just doesn't make sense. Cats and percentages are different things, so just because the unit stays in one case, doesn't mean it should stay in another case. It seems as though they're just using "percentage" as a synonym for "fraction." While adding fractions, the unit can stay the same. While adding percentages, it depends on what you're measuring. If you're saving up to buy a cat, and you've saved 57% of the price and then you save another 35% of the price, then, yes, you've saved 92% of the total price and the unit can stay. But it really doesn't make sense for it to stay when what you're actually measuring is an average not a total.
Anyway, fair enough to just use it as a dominance (I'd say "success," but that doesn't really matter) ratio without a unit.
The reason for not dividing by 2 is that it is an extra step and a pain.
You may not be aware of it but the way the ATP records data for us to see is very sloppy. When they say that player A has won 80% of games, that could be 80.49% or 79.50. That error is not just possible. It is quite common. What that means is that there can be a .99% wobble. When you go to the site and see a whole number of players "tied" at 80%, you assume that the one at the top is highest. But in fact the ATP sorts the % by another fied, total number of matches played.
Then when you look at % of games won on return, you have the same error. This means that two players listed as being 80% and 35% can be off by a full %.
80.49 + 35.49
or
79.50 + 34.50
That's 115.98 vs. 114.
Since it gets rounded to 116 vs 114, that's a lot of error. For the top players every % point is huge. When comparing careers that's about like saying batting 304 and 295 is the same thing.
So that's the real problem.
Other than that it's just easier looking at both figures as a sum because it stops the extra step
(80.49 + 35.49)/2
But that's all we are really doing 57.99. In the end it is +/- about 1/2%.
This can be fixed by using the ATPs actual data, because they do give the exact number of games played and won both on serve and on return.
As said, over a career and even over a season service games and return games are so close to even that it really isn't worth the extra time to worry about that.
In contrast, % of points lost on serve and won on return can't simply be added. There are always way more points played on return by any player who is dominant. Those figures only approach and equal figure when a player is winning about 50% of all points, where serve and return sort of balance out because nothing is really being gained.
