Originally Posted by krosero
Yes I think this is an issue with any stats that use Unforced Errors: namely methods like the Aggressive Margin, winner/error differentials, ratios, etc.
On grass you will commonly see very high winner/error differentials: +50 or more. On clay it's hard to get even a positive differential, and negative ones are common.
Perhaps I should have listed the AM's grouped by surface, because that is the factor that seems to have the greatest impact on how people play -- and therefore an impact on how UE's are scored in those matches.
On grass two SVers coming in behind every serve will have very few opportunities to make UE's. On clay two baseliners engaged in a war of attrition are going to have many such opportunities.
That's one reason that AM's are higher on grass. But leaving aside the question of playing styles, it's also just generally true that you can make more winners and forcing plays on a fast surface. That has nothing to do with styles per se, because you could have two baseliners playing each other, and they will find it to be true: they will have an easier time forcing their opponent when they're playing on a fast surface.
On clay you can hit the ball just as hard, or just as wide, but the defender gains an advantage, as compared to grasscourt matches. Any surface that slows down the ball will make it easier for the defender to catch up to it, and harder for anyone to hit winners or to make forcing plays.
And that's what the AM is about: making forcing plays while keeping your unforced errors down. That is just inherently easier on a fast court.
So that makes it really impossible to compare AM's -- or winner/error differentials and ratios -- across surfaces (except for limited purposes like I did with the Federer-Delpo matches).
Take the top AM's in my opening post as examples. McEnroe has 53% on grass, the highest AM of all. The highest AM on clay belongs to Nadal, with 35%. No one would suggest that McEnroe's level on grass is a greater level than what Nadal achieved on clay. I think it's fair to assume that both of those performances were GOAT level for their respective surfaces.
Sampras/Stich should be compared to other grasscourt matches of the time period. It was very common then for players to follow all, or nearly all, of their serves to net. Sampras/Stich could easily be compared with Sampras/Becker, for example.
But if you try comparing it to this era's matches at Wimbledon, which commonly feature two baseliners slugging it out, you're going to have problems.
At the very least you could not use the numbers in such a comparison literally, if at all.
But that all results from surface being such a large factor in tennis. I'm not sure there's any good or efficient way around this. I, for one, would not want to chuck the distinction between a forced and an unforced error; I would just restrict comparisons by surface.
But maybe we need to think more about how to score UE's.
Your point about surface affecting AM is well taken. You say, for instance, that fast/low-bouncing surfaces tend to have higher AMs. The reason you give for this seems to be twofold: (1) it's easier to execute aggressive plays on fast/low-bouncing surfaces; and (2) netplay is more common on fast/low-bouncing surfaces. Both of these two aspects (easier to force plays, more netplay) seem to positively affect AM, yet you talk as if matches played on the same surface (with the same speed) are roughly comparable with respect to AM. This would only be true if the amount of netplay is roughly constant between matches on a given surface.
It's not clear to me that their is this constancy - especially if we look across eras. Compare Borg-Vilas or Djokovic-Nadal at the FO to Laver-Rosewall. The former pairs remained almost exclusively at the baseline, while the latter approached the net in most points. I include Djokovic-Nadal because, while both camped out on the baseline, they also both go for winners when they have an opening, perhaps in contrast to the more conservative play of Borg-Vilas. This is worth mentioning because I'm going to guess Laver-Rosewall would have the highest AMs of the three pairs due to less unforced errors, but I don't want this attributed to the 'rally-all-day' style of Borg-Vilas. Rather, I think, Laver-Roswall will have the higher AMs than the other pairs because they had less opportunity to hit unforced errors due to the frequent net approaches. This would be the case even though the surface is held relatively constant.
You've talked about this yourself, and the lesson I think is that one can only use AM to compare match quality when holding the surface and level of net play relatively constant. Otherwise AMs can be easily inflated or deflated by factors unrelated to quality of play.