Just curious: what exactly is the criteria to judge a good serve if you throw out ace counts and first serve %? # points won on the first serve? please elaborate.
Nobody's saying the ace counts and 1st-serve %'s don't count. They're in fact the exact reasons why Lendl (and Borg, by extension) isn't included and someone like Tanner is several places below Roddick even though their actual top-end speeds are probably roughly the same. But the numbers aren't everything. The important thing is that these serves should be viewed in the context of their time.
The relatively low ace counts of the pre-'90s are a case in point. Back then, it was standard for S&Vers and attacking players to take off a few mph from their 1st serves so that they can set up their follow-up volleys better. In other words, they often used their 1st serves as set-up shots, and naturally they had fewer aces. In fact, by the (admittedly) few accounts I've read, Laver used to be criticized for taking too many risks on his serves, though he probably had more aces per match than, say, Newcombe. That would be considered "good" today, but in his era such an approach was frowned upon. (It's to Laver's credit that the risky tactics worked for him.)
Still, when you look at the stats posted by bricks, you can see that all-time great servers like McEnroe, Noah and Curren have the edge over Lendl, however slight it is. (I've covered this topic in more detail upthread.) You need to understand that there's actually not a whole lot separating these servers, but the gap is still there. That gap can mean a few places' difference or, in Lendl's case, the difference between inclusion and exclusion.
Which brings me to the increasingly higher ace counts and 1st-serve %'s over the years. What's telling here is that this increase is due not only to the newcomers, but also some of the same players who thrived in the '90s. That tells us that perhaps something more fundamental is at work. As to what brought about this change, there are a few possible factors. Maybe the racquet technology factor is bigger than we realize (though, for the record, I doubt it helps speed up the serves all that much--that may be true of the journeymen and others with imperfect mechanics, but not of the best servers). Or maybe the S&Vers learned how to put volleys away while firing bombs constantly. But again I'd say the most important factor was the players' mindset. By the '00s the prevailing MO was to win the point outright on each and every 1st serve, hence the higher ace counts and, perhaps thanks in part to the bigger racquets, the higher %'s.
I'm usually skeptical of any claims about the supposedly big changes due to the surface, and this one is no exception. For comparison, Krajicek at '02 Wimbledon served 21 aces on average per match, as opposed to just under 20 at '96 Wimbledon, his best showing at a major. This was on the supposedly "clay-courted" grass, and just after he had come back from extended time off due to surgery. (In fact '02 Wimbledon was just his 2nd event he'd played that year, the 1st being 's-Hertogenbosch where he lost to Federer in the 1st round.) And Sampras averaged just over 20 aces at the '02 USO, his very last event. Without digging up previous stats I think it safe to say this was one of his higher yearly ace counts, or at the very least not much lower than his average. Likewise this was after the USTA supposedly slowed down the surface in '01. Also, while I don't have the stats handy, I do know that Sampras averaged about 15 aces per match in '02. If the surfaces were indeed slower, these changes were not enough counter whatever other effects that allowed these players to post such high numbers. (You also seem unaware that before moving to Ashe Stadium in '97 the USO used to be in fact slower, but I'll save that topic for another discussion.)
But, again, the numbers aren't everything. Karlovic's stats from '07 look impressive on paper and very impressive they are, but what they don't tell us is that back in '07 Ivo had yet to fully mature as a player (by his standards). That was when he had yet to make the 4th round at any one of the majors, so he faced a limited pool of top players back then. I hasten to add that this isn't just a matter of facing the best returners, though much of it is. When you play a Murray, Djokovic, Federer or Davydenko, they not only make you hold serve harder due to their excellent return of serve, but they also put more pressure on your return and in turn on your serve because they can hold serves easily as well. The game then becomes more psychological and tactical, and you might hit fewer 1st serves and more DFs, or in some cases even hold back on purpose.
Here's what I mean. If you watched the '09 USO final, you should've seen that as the match progressed del Potro went for increasingly conservative 1st serves. Now, people have said he was probably tired and I'm sure fatigue had something to do with it, but IMO that was a conscious and smart decision. He probably figured that instead of wearing himself out further with each and every big 1st-serve attempt, he could conserve his energy and also get more 1st serves in and still be competitive. In other words, he sacrificed a few additional aces for a higher % of 1st serves in and points won. And we all know how that turned out.
Karlovic hasn't needed to go to such lengths all that often, and especially not in '07. Ivanisevic, on the other hand, often did, and one can imagine what that could do to such a headcase like him. Also, when you look at the years when a player served 1000 or more aces, you'll see that Goran is the leader with 4 years as opposed to just 1 for Sampras' '93, Roddick's '04 and Karlovic's '07. And what's interesting here is that Ivo played more matches in '07 than any of the years listed for Goran except '96. So even in stats alone Goran isn't far behind, and might actually be ahead from a consistency standpoint.
And where Goran also has the edge over Ivo is in disguise and the 2nd serve. Karlovic likes to serve flat, and to his credit he can still maintain a high % because his height allows a large margin for error. But it's not a particularly well-disguised serve. By contrast, Goran's serve was exceptionally hard to read, as his service motion was so swift and his toss very low. Plus he had that lefty advantage. These things come in handy when you're playing the best players.
As for the 2nd serve, when you study the stats from Karlovic's matches against Davydenko, Federer and especially Murray (at least the ones they won), one thing that sticks out is his low % of 2nd-serve points won, which sometimes hover barely above 40%, if not lower. No doubt this is due to Ivo's limited set of tools, but to me the most important factor is his relatively weak 2nd serve. My guess is that if you gave him the 2nd serve of Isner or Roddick, he'd probably win more points on 2nd serves. (Remember, we're trying to rate these serves as stand-alone shots.) And of course Sampras' 2nd serve wouldn't be a bad replacement, either. (FYI that 2nd serve is why Pete retains the 3rd place. If you go back and look you'll see that I'd initially put Pete at No. 4, below Krajicek, but people responded that I was not giving the 2nd serve enough weight and I made the adjustment that still stands.)
And speaking of Sampras, what made his serve so feared and formidable wasn't just its speed and high ace counts--though they were clearly above average--but its spin. John Yandell wrote an article a couple of years ago illustrating that his research showed the Sampras serve having 50% more spin than that of other serves with similar velocities, and also a much higher topspin component than other serves he measured. As a result the ball off the Sampras serve was "significantly higher and heavier at the time of the return" (I quote from the original article), and a few of his contemporaries have said as much, noting how hard it was to return his serve. So while Pete did not score as many aces as Goran or Karlovic, his serves often drew weak returns that he could put away outright or use to take immediate control of the point. I've read that Newcombe's serve was "heavy" like that.
I could also go on about the variety of McEnroe, who had every serve in the book, and so on. My point is that you can't get this kind of info from the stats. For that you need to be knowledgeable and have seen these players yourself, or at least consult those who have and who are well versed in the game to give an informed opinion. That's why I've been asking people to weigh in on some of the old-timers. To judge these servers by the stats alone is to miss the forest for the trees.
And let me repeat that there's actually not much that separates good servers from great ones. But small differences matter in a comparison like this, and you might have seen that even such ace dispensers as Rosset, Querry and Ljubicic failed to make this list (though I'm leaning toward adding Ivan right now). Federer's serve is indeed a very good stroke, sometimes nearly great, but not quite at the elite level. That's nothing to be ashamed of.
, and it's not even close. 
