Originally Posted by urban
Matchups like Lendl-Wilander imo show, how difficult it is, to determine level of play and to define unforced or forced errors. Both played more on long term strategy, a sort of position war, Lendl a bit too defensive for my gusto, Wilander 1988 more willing to attack on certain points. A match is not decided by the sum of games or points, but by winning the decisive quality points, there are big points and minor significant points. And still the point on the same scoring situation could be more important in different situations. For example: A 40-15 point for a server has different worth than in a 15-40 situation, if the returner is leading.
To the unforced error question. In such a match Lendl- Wilander you could bring up the hyothesis - i now make a too radical argument to get the point over - that they played no unforced errors at all. The same shank is a different error, when hit after one excange or hit after an excange of 30 strokes and more, when the player is near to collapse. So in a way, in such a long match with many extremely long exhausting rallies, many more " easy" appearing errors are "forced" (by nature, exhaustion, mental fatigue and other factors) than in other matches, where a shank is a shank, by lack of concentration, bad positioning, bad stroke production or other pure technical or form factors.
Without question there are more important points than others, and I'm glad that Moose and I have started consistently counting the number of first serves made on break points, to take one example. In tennis you win if you break in a service game, or if you get the mini-break in a tiebreak. You don't need to win the most points in a match if you're going to win it.
That said, the player who wins the most points, the vast majority of the time, is the player who wins the match. What I like about the AM method is that it always tells you who won the most points; which is the reason that the player with the higher AM almost always wins the match. There is a stronger correlation with the AM than with other common methods which often give the wrong impression about who won the match or about what the margin of victory was (methods like winner/ue differentials and ratios).
When you do come across a match in which the loser has the higher AM, you know automatically that he must have played the important points more poorly. You won't know, of course, which points those were; totals can never point to specific moments. But you will know that the loser's quality of play on important points was lower than the victor's: and that's something you would want to be reflected in any stat, even if the details are something you can only look up by taking a close look at the match.
I have some matches like this. Safin trailed Federer 194-201 in total points at the 2005 AO despite winning the match; so his AM trails Federer's, 19.0% vs 20.8%. The year before there was a similar situation at the 2004 AO, between Safin and Agassi: Safin won fewer points in total, and therefore had the lower AM, but won the match.
Another example is Nadal d. Federer in Rome (the important points lost by Federer in that match can be very simply identified, in that case: he made two UE's on the two match points that he held). Two more examples are Federer d. Agassi at the 2004 USO, and Agassi d. Medvedev in the 1999 RG final, etc.
All of this is to say that the AM, because it tracks the total points won, has a built-in way of telling you who produced the higher quality of play consistently from point to point, over the course of the match (something we would want to know, and which is a strength in any player); and who played better on the important points (something we would also want to know, if it's not the same person who won the most points).
I do think that tennis stats are growing increasingly sophisticated and can focus in on incredibly detailed moments: like telling you how far back Roddick is standing when he returns serve on break points.
Those details are unquestionably important. Tennis writers have always focused on them; it's more difficult to measure them broadly with objective stats, but I think tennis stats are getting there.
As for the argument about not all unforced errors being the same, I can only agree wholeheartedly
Two strokes may both be wide forehands, and may be executed the same way from a superficial standpoint; but one occurs on the second shot of the exchange, under little pressure, while the other occurs at the end of a wild 30-stroke rally. Unquestionably there's a huge difference.
All that has an impact on the AM's, though I think you can identify the impact and account for it.
Let's take Wilander and Lendl as an example. There you have two great defenders engaged in long rallies. Their AM's in the 1987-88 USO finals were relatively low, because two defenders facing each other will find it hard to hit through each other and pull off winners.
Next year, Wilander loses to Sampras in five sets, and Mats has an AM of 29.8%. Now, was his quality of play really higher than the previous year when he beat Lendl with an AM of 13.8%? Not a chance.
The reason his AM is so high against Sampras is that Sampras came into net relentlessly, forcing to Wilander to make errors or to hit winners: so Mats just didn't have that many chances to make unforced errors. That's the general rule about errors: if you make an error while you're opponent is at the net, your error is typically judged as forced.
The unforced errors that Wilander did make in the Sampras match, when he had a chance to make them, did not come at the end of grueling 50-shot rallies; they were more genuinely "unforced errors" in the ideal sense of that term: unnecessary, sloppy errors.
Knowing all this, you know not to jump to the conclusion that Wilander's level of play in the '89 match was superior to what he did in '88.
I think AM comparisons work best when you can judge AM's that took place on similar surfaces, between opponents of similar style. If there is some radical difference in how the unforced errors occurred, in two matches, then the AM's cannot be used in a straightforward manner, if at all.
Terrific comments, Urban. I always enjoy your comments about match dynamics.