Djoko-coach channels Djoko-fans, claims 2017 is Peak Fedr

RF-18

Talk Tennis Guru
my-job-here-is-done.jpg
 

AnOctorokForDinner

Talk Tennis Guru
Tennis stats will go nowhere until someone with a lot of time figures an accurate way to model playing level based on in-match statistics and taking into account opponent and opponent level. Something like the prior-based ridge regression approach NBA stats guys take for RAPM and stuff, except in tennis there's way more variables than just +/-. Problem in tennis is that I don't think large datasets are easily accessible or scrapable, i.e. complete match reports/stats for every match.

Even match stats/charts, even though quite helpful when discussing specific matches, are of course still approximations - not all winners or errors are equal. Such an all-encompassing model as you speak of won't happen until there is a way to chart the shots themselves (spin, depth, placement), which requires the resources and technology only available to the ATP.
 

metsman

G.O.A.T.
Even match stats/charts, even though quite helpful when discussing specific matches, are of course still approximations - not all winners or errors are equal. Such an all-encompassing model as you speak of won't happen until there is a way to chart the shots themselves (spin, depth, placement), which requires the resources and technology only available to the ATP.
Well that would be the dream (and definitely doable considering that they throw the spin/pace numbers at us all the time, so they must track everything) but I think even just looking at the traditional match stats and forced errors and trying to adjust for opponent quality would still be better than just blindly considering games/points stats.
 

Gazelle

G.O.A.T.
The statement in the OP is correct. Also I'd like to add that Djokovic 2017 was peak Djoko and Nadal 2017 was peak Nadal.

But I won't go that far to claim that Murray 2017 was peak Murray, hobbling around with that broken hip of his.
 

AnOctorokForDinner

Talk Tennis Guru
Well that would be the dream (and definitely doable considering that they throw the spin/pace numbers at us all the time, so they must track everything) but I think even just looking at the traditional match stats and forced errors and trying to adjust for opponent quality would still be better than just blindly considering games/points stats.

It is, but there will never be quite an agreement on opponent quality since we know form trumps name, so to make the best judgment you'd still want to watch the match, but even those who have actually watched it aren't all going to have the same opinion. And there isn't agreement on what exactly a forced/unforced error is, either. Match-game-point stat is a good help for general season analysis - if one season is much better in that regard than another, that says something about the player's general level throughout the year, e.g. Federer being consistently much more dangerous on return at his actual peak than now, even if back then he still had an occasional match where he was content to bot it out in tiebreaks, but not in late rounds unless his opponent was Karlobot.

In short, discussing individual matches is of course more precise, but it's only good when all parties not only perused the stats, but also watched the actual match with an open mind, as NatF correctly said. I think it's best not to have strong convictions about it, moreso on the negative side, since you don't have an eidetic memory or knowledge and hence could have missed something.
 

Gary Duane

G.O.A.T.
Even the scoring system in tennis presents some real hurdles and allows for huge margin of error and as such you can see a champion like Sampras with such atypical patterns not match up to more usual expectations. But even for a single player, such increases of the resolution may be revealing across years; though findings would still float in a sea of confusion until the fields are relatively judged statistically, and then one would probably need to figure out how that then matches up to routes taken to Championships and to winning. Year to year would see big swings in fortune (as you suggested earlier in this thread by comparing some of Fed's years).
Yes, your post and this thread plainly shows that not only did Federer not exceed his prime years in terms of match stats, he also out-performed what his 2017 stats would expect of him.

Hey, side Q: what do you make of Pete's game w-l in majors for his career? I know we've discussed this before (serve-dominant players can get away with lower %'s, Fed and Pete coast a little on the return, etc) but I still find it fascinating that Pete is only at 57% while Agassi is at 58.7%. Further emphasizes that, however useful our stats are, they have unfortunate limitations.
One of the first things I did with games was to get a count of all games at all majors, for the winners. From that I know that there is around a 5% difference between clay and grass, with HCs in between. That's a good ball park figure to start out with.

This is what I have for Fed and Sampras, for majors

Fed: 59.45
Sampras: 57.23

Agassi:
58.84

That just does not make sense to me.

Djokovic:
59.60

Djokovic and Fed are very close, Agassi a bit lower, but Sampras?

Even on grass, majors:

Fed: 59.89
Sampras: 59.20

But now look at grass, the last 3 rounds at majors.

Fed:
56.1126
Sampras:
58.2649

It's closer on HCs, but even there Sampras has a slight edge, and if you check only the USO, Sampras is even higher.

I think it's safe to say that Pete coasted in earlier rounds, went to entirely a different gear at the end of big events.
 

AnOctorokForDinner

Talk Tennis Guru
I think it's safe to say that Pete coasted in earlier rounds, went to entirely a different gear at the end of big events.

That has always been obvious. See also the 'times broken' stat per round at Wimbledon (considering 1992-2001 for prime):

R1 - 9 times in 10 matches (4 of them coming in 1996 R1 vs Reneberg)
R2 - 11 times in 10 matches
R3 - 4 times in 10 matches
R4 - 6 times in 10 matches (3 of them in 2001 R4 vs Federer)
QF - 10 times in 9 matches
SF - 12 times in 8 matches
F - 4 times in 7 matches (unbroken in 5 of the 7 finals he played and won)

Often coasting in R1 & R2, getting serious since R3, nice curve up to the semis, and then on the final day GOATpras appears and is super unbreakable. Point win % per round, per TA stats, bears this out as well: 56,7 - 54,9 - 60,1 - 56,3 - 53,3 - 52,3 - 54,6.
 

Gary Duane

G.O.A.T.
That has always been obvious. See also the 'times broken' stat per round at Wimbledon (considering 1992-2001 for prime):

R1 - 9 times in 10 matches (4 of them coming in 1996 R1 vs Reneberg)
R2 - 11 times in 10 matches
R3 - 4 times in 10 matches
R4 - 6 times in 10 matches (3 of them in 2001 R4 vs Federer)
QF - 10 times in 9 matches
SF - 12 times in 8 matches
F - 4 times in 7 matches (unbroken in 5 of the 7 finals he played and won)

Often coasting in R1 & R2, getting serious since R3, nice curve up to the semis, and then on the final day GOATpras appears and is super unbreakable. Point win % per round, per TA stats, bears this out as well: 56,7 - 54,9 - 60,1 - 56,3 - 53,3 - 52,3 - 54,6.
Those are very cool stats I knew nothing about.

You have to remember that there are a lot of stats that seems very logical, yet they aren't accessible.

To me the most logical thing in the world is to track game%, but when you go to the ATP site, you have to combine them yourself, and the ATP only gives us those figures rounded off. There are whole pages devoted to service games won, and return games won, but no page anyone with games won. Even so, they are there for year, on each surface, but not broken down by season. You can't check this out by event.

Then you go to a site like Tennis Abstract and see finally they added Brk%, which is the same as return games. This time it is to one decimal. But they only added that this year. Before 1990 there is no way to track games won on serve and return unless you watch matches and do it yourself. So we are missing that kind of data for older ATGs.

I got used to tracking as much of this as I could before I could find it online.
 

Gary Duane

G.O.A.T.
That's not to say that stats have rendered the 'eye-test' obsolete in baseball, far from it, every sport still has that human element regardless of how deep the stats go...nor does the cumbersome nature of devising good statistical models in tennis make such an endeavour worthless but...**** it you're intelligent enough to fill in the blanks, and I'm tired a.f. from a long work week. :p
First of all, people are either going to understand what you are saying - and I do - or they aren't going to understand any of it. So when you make intelligent points, be prepared to be understood by other intelligent people, and be content with that. ;)

I think the biggest variable in tennis is the lack of a control.

If you are thinking about bowling - and by the way I know next to nothing about it - I would assume you can rate players against the simple fact that 300 is a perfect score, and you can't better that. So if player A wins a tournament with less points than player B gets in another tournament, who loses, we can at least guess that the competition at one event is higher.

If it's straight pool, we can look at consecutive balls pocketed. We have to make sure that when comparing we are talking about pool tables that are the same size, with pockets that are the same width and shape, but other than that I don't expect things to change too much.

But in tennis the only thing we could theoretically compare would be the serve - the speed of the serve, the amount of spin, but then we are hit with the problem of equipment.

At least in baseball it pretty comes down to what a pitcher does with the ball he throws. (Baseball experts may think of something I'm missing.

For a marathon I suppose maybe better shoes make a difference, and I don't know what other factors are there, but doesn't it pretty much come down to how little time you complete the marathon in?

So if you want to argue about how much better today's athletes are in the marathon, you can simply look at the records and how the time has come down.

But what do we have in tennis?

We have comparisons between Laver, with a wood racket, and Federer, with completely different equipment. Everything is different. And yet people still use the "eye test" to make claims about who is better, and why.

Even right now there are arguments about the playing level in any given year, saying that peak Fed had easy competition, or Nadal's USO last year was next to worthless because of such an easy draw, or that somehow Novak's non-calendar-year GS only happened because competition was getting weaker. And all these conclusions are based on "eye tests".

So no matter how problematical tennis stats are - and there are indeed HUGE problems - I think they are at least a help in combating stubborn adherence to the eye-test, which claims to be more accurate than any set of facts.
 
Top