Djokovic vs. Federer: A decade at Wimbledon

weakera

Talk Tennis Guru
Most people would probably guess offhand that Federer had the superior decade at Wimbledon, but did the Serbian Superstar in fact outdo the Swiss Maestro?

Federer 2000-2009
52-4 (92.8%) record
6 Wimbledon titles, 1 final, 0 semifinals
Opponents defeated in finals: Philippoussis, Roddick, Roddick, Nadal, Nadal, Roddick
4.16 average # of sets needed to win finals

Djokovic 2010-2019
57-4 (93.4%) record, 1 injury retirement
5 Wimbledon titles, 1 final, 2 semifinals
Opponents defeated in finals: Nadal, Federer, Federer, Anderson, Federer
4.2 average # of sets needed to win finals

Mean Opponent Elo
2014 Wimbledon Djokovic 2098.6
2015 Wimbledon Djokovic 2095.3
2018 Wimbledon Djokovic 2068.7
2005 Wimbledon Federer 2068.3
2006 Wimbledon Federer 2060.3
2009 Wimbledon Federer 2045.4
2011 Wimbledon Djokovic 2040.6
2007 Wimbledon Federer 2034.2
2019 Wimbledon Djokovic 2025.6
2003 Wimbledon Federer 1985.7
2004 Wimbledon Federer 1978.0
 

blablavla

G.O.A.T.
Most people would probably guess offhand that Federer had the superior decade at Wimbledon, but did the Serbian Superstar in fact outdo the Swiss Maestro?

Federer 2000-2009
52-4 (92.8%) record
6 Wimbledon titles, 1 final, 0 semifinals
Opponents defeated in finals: Philippoussis, Roddick, Roddick, Nadal, Nadal, Roddick
4.16 average # of sets needed to win finals

Djokovic 2010-2019
57-4 (93.4%) record, 1 injury retirement
5 Wimbledon titles, 1 final, 2 semifinals
Opponents defeated in finals: Nadal, Federer, Federer, Anderson, Federer
4.2 average # of sets needed to win finals

Mean Opponent Elo
2014 Wimbledon Djokovic 2098.6
2015 Wimbledon Djokovic 2095.3
2018 Wimbledon Djokovic 2068.7
2005 Wimbledon Federer 2068.3
2006 Wimbledon Federer 2060.3
2009 Wimbledon Federer 2045.4
2011 Wimbledon Djokovic 2040.6
2007 Wimbledon Federer 2034.2
2019 Wimbledon Djokovic 2025.6
2003 Wimbledon Federer 1985.7
2004 Wimbledon Federer 1978.0

what about Nadal decade at Wimbledon?
 

Third Serve

Talk Tennis Guru
1. Fair point on the win percentage. Fed's 2000-2002 numbers bring this down a bit (his best overall decade at Wimbledon was actually 2003-2012, but apparently we're only doing decades from 0-9). In that regard, that's the best point you have to bring up.

2. Nothing much to say on the tournament stats themselves. They are what they are, I guess.

3. The opponents in the finals are actually a bit more similar in pure level than the names would indicate. 2007 Nadal is easily the best final opponent from both lists. Fed 2014 / 2015 and Roddick 2004 / 2009 would proceed after that, in whatever order you want. Then 2019 Fed / 2006 Nadal and then the rest. The competition both faced at Wimbledon is not that far apart.

4. Sets dropped in the finals??? Kind of an odd stat to bring up. I guess it makes sense since you brought up the finals opponents, but I would prefer sets dropped across the whole tournament. You may have a different point of view, but I value consistent high level across a whole tournament. Given the inclusion of Fed's 2000-2002 tournaments, I don't think these numbers would be more flattering, though.

5. ELO? Let's not go there, please.
 

Got Aced

Rookie
2006 Nadal > 2008 Nadal

1177gr.jpg
 

weakera

Talk Tennis Guru
Thank you for starting a thread that has nothing to do with the French Open. :)

Who cares about any of this stuff? The only thing that matters ultimately is that Roger Federer has 8 Wimbledon titles and Djokovic has 5. Anything else is ludicrous cherry picking or just boredom-inducing.

It's worth discussing how strongly Djokovic's decade at Wimbledon compares to the much celebrated Federer run.
 

mike danny

Bionic Poster
1. Fair point on the win percentage. Fed's 2000-2002 numbers bring this down a bit (his best overall decade at Wimbledon was actually 2003-2012, but apparently we're only doing decades from 0-9). In that regard, that's the best point you have to bring up.

2. Nothing much to say on the tournament stats themselves. They are what they are, I guess.

3. The opponents in the finals are actually a bit more similar in pure level than the names would indicate. 2007 Nadal is easily the best final opponent from both lists. Fed 2014 / 2015 and Roddick 2004 / 2009 would proceed after that, in whatever order you want. Then 2019 Fed / 2006 Nadal and then the rest. The competition both faced at Wimbledon is not that far apart.

4. Sets dropped in the finals??? Kind of an odd stat to bring up. I guess it makes sense since you brought up the finals opponents, but I would prefer sets dropped across the whole tournament. You may have a different point of view, but I value consistent high level across a whole tournament. Given the inclusion of Fed's 2000-2002 tournaments, I don't think these numbers would be more flattering, though.

5. ELO? Let's not go there, please.
Picking 2000-2009 is stupid anyway. Of course Djoker did perhaps better in his decade since he was closer to prime tennis in early 2010's than Federer in 2000-2002.
 

ND-13

Hall of Fame
Most people would probably guess offhand that Federer had the superior decade at Wimbledon, but did the Serbian Superstar in fact outdo the Swiss Maestro?

Federer 2000-2009
52-4 (92.8%) record
6 Wimbledon titles, 1 final, 0 semifinals
Opponents defeated in finals: Philippoussis, Roddick, Roddick, Nadal, Nadal, Roddick
4.16 average # of sets needed to win finals

Djokovic 2010-2019
57-4 (93.4%) record, 1 injury retirement
5 Wimbledon titles, 1 final, 2 semifinals
Opponents defeated in finals: Nadal, Federer, Federer, Anderson, Federer
4.2 average # of sets needed to win finals

Mean Opponent Elo
2014 Wimbledon Djokovic 2098.6
2015 Wimbledon Djokovic 2095.3
2018 Wimbledon Djokovic 2068.7
2005 Wimbledon Federer 2068.3
2006 Wimbledon Federer 2060.3
2009 Wimbledon Federer 2045.4
2011 Wimbledon Djokovic 2040.6
2007 Wimbledon Federer 2034.2
2019 Wimbledon Djokovic 2025.6
2003 Wimbledon Federer 1985.7
2004 Wimbledon Federer 1978.0

Djokovic has been very impressive at Wimbledon but one thing that you remind us is that Federer is seen greater because even during Djokovic's regime he won 2 more Wimbledon and 3 more finals with a MP in one, 5 setter in the next and a 4 setter in the last.

Also Federer won 5 in a row and also did the FO-Wimb Combination which makes it more special.
 
Last edited:

weakera

Talk Tennis Guru
Weakera vacillates between some good posting and not-so-good postings, like all of us. Peak Weakera is actually funny and adroit. But the recent spate of French Open threads has been a bit much.
tenor.gif

Thank you for the relatively kind words!
 

Third Serve

Talk Tennis Guru
Picking 2000-2009 is stupid anyway. Of course Djoker did perhaps better in his decade since he was closer to prime tennis in early 2010's than Federer in 2000-2002.

Similar stats for 2003-2012 for those who want to know:

Federer 2003-2012
7 titles, 1 final, 2 quarterfinals
Opponents defeated in finals: Philippoussis, Roddick, Roddick, Nadal, Nadal, Roddick, Murray
Win percentage: 63-3 (95.4%)
Sets dropped on average (different stat): 2.9
 

daggerman

Hall of Fame
I personally think Federer is clearly a (slightly) better player at Wimbledon than Djokovic, but I do want to mention that Elo rating gets a bad rap on this forum, and I'm not sure why. It's a useful metric when used properly, and that it's dismissed as completely useless is beyond me.

(And I'm hoping somebody references that critique of Elo somebody posted a while ago because I want to address it in a thread but am too lazy to look it up.)
 

mike danny

Bionic Poster
Weakera vacillates between some good posting and not-so-good postings, like all of us. Peak Weakera is actually funny and adroit. But the recent spate of French Open threads has been a bit much.
tenor.gif
He is condescending and a bully who never apologizes for personal insult. The worst kind on this planet.
 
(And I'm hoping somebody references that critique of Elo somebody posted a while ago because I want to address it in a thread but am too lazy to look it up.)
You mean this?
 

daggerman

Hall of Fame
You mean this?

Thanks! I actually have never given it a proper read. Maybe it'll also convince me Elo is worthless :-D
 

NoleFam

Bionic Poster
I personally think Federer is clearly a (slightly) better player at Wimbledon than Djokovic, but I do want to mention that Elo rating gets a bad rap on this forum, and I'm not sure why. It's a useful metric when used properly, and that it's dismissed as completely useless is beyond me.

(And I'm hoping somebody references that critique of Elo somebody posted a while ago because I want to address it in a thread but am too lazy to look it up.)

It's pretty obvious why it gets a bad wrap. :D It's only the official system used to calculate the ratings for the most popular sport in the world but according to TTW, it's completely useless. Lol.
 

Knovax

Rookie
You mean this?
Thanks for the link. I have to see what's up with this ELO thingy.
 

Rosstour

G.O.A.T.
What’s interesting is that when you break it down that way, Novak has 0 Wimbledon titles and 0 Finalists during Fed’s decade, but Fed has 2 titles and 3 Finalists during Novak’s.

So these numbers don’t quite embarrass Fed the way some might have hoped.
 

AnOctorokForDinner

Talk Tennis Guru
I personally think Federer is clearly a (slightly) better player at Wimbledon than Djokovic, but I do want to mention that Elo rating gets a bad rap on this forum, and I'm not sure why. It's a useful metric when used properly, and that it's dismissed as completely useless is beyond me.

(And I'm hoping somebody references that critique of Elo somebody posted a while ago because I want to address it in a thread but am too lazy to look it up.)

Because it's misused all the time (lol). Also is slow to adjust and overvalues consistency, which undermines its applicability to a sport like tennis where level of play is normally quite volatile and affected by conditions & match-ups as well.
 

daggerman

Hall of Fame
Because it's misused all the time (lol). Also is slow to adjust and overvalues consistency, which undermines its applicability to a sport like tennis where level of play is normally quite volatile and affected by conditions & match-ups as well.

If Elo were as resilient to volatility as you think, it probably wouldn't be as good a predictive tool as it is. But yes, I agree that those are some of Elo's shortcomings. I would nonetheless argue that the extent to which it is dismissed is disproportionate to the severity of its shortcomings.
 

AnOctorokForDinner

Talk Tennis Guru
If Elo were as resilient to volatility as you think, it probably wouldn't be as good a predictive tool as it is. But yes, I agree that those are some of Elo's shortcomings. I would nonetheless argue that the extent to which it is dismissed is disproportionate to the severity of its shortcomings.

Up to 70% prediction rate may be the best there is in tennis, but IRL you'd hardly be content to trust a decision maker that had a 30% error rate, eh? And I recall the UTS guy said that predictive success in tennis (across the whole ATP tour as a sufficient sample) must be capped around 75% due to inherent volatility. In the end, ELO is merely a somewhat better predictor than official rankings, and the same considerations apply: it's no surprise that higher-ranked players beat lower-ranked players more often than not, but not nearly often enough to make a specific assumption about a higher-ranked player playing better than a lower-ranked player in any specific match or tournament. Especially not when we have eyes to see and brains to analyse.
 

daggerman

Hall of Fame
Up to 70% prediction rate may be the best there is in tennis, but IRL you'd hardly be content to trust a decision maker that had a 30% error rate, eh? And I recall the UTS guy said that predictive success in tennis (across the whole ATP tour as a sufficient sample) must be capped around 75% due to inherent volatility. In the end, ELO is merely a somewhat better predictor than official rankings, and the same considerations apply: it's no surprise that higher-ranked players beat lower-ranked players more often than not, but not nearly often enough to make a specific assumption about a higher-ranked player playing better than a lower-ranked player in any specific match or tournament. Especially not when we have eyes to see and brains to analyse.

Yeah, I agree with most of this.

I think Elo does a much better job than 70%, but I'd have to run the numbers to confirm. I'll report results if I get around to it today.
 

Mike Sams

G.O.A.T.
Federer and Serena Williams love blue clay while Nadal and Djokovic whine and moan and cry about it.
Blue clay seperated the gazelles/runners from the shotmakers.
It's why Tomas Berdych the shotmaker made the final while Djokovic the jackrabbit was falling all over the place because he couldn't hop all over the court doing his silly splits while defending.
 

daggerman

Hall of Fame
Up to 70% prediction rate may be the best there is in tennis, but IRL you'd hardly be content to trust a decision maker that had a 30% error rate, eh? And I recall the UTS guy said that predictive success in tennis (across the whole ATP tour as a sufficient sample) must be capped around 75% due to inherent volatility. In the end, ELO is merely a somewhat better predictor than official rankings, and the same considerations apply: it's no surprise that higher-ranked players beat lower-ranked players more often than not, but not nearly often enough to make a specific assumption about a higher-ranked player playing better than a lower-ranked player in any specific match or tournament. Especially not when we have eyes to see and brains to analyse.

I said:
Yeah, I agree with most of this.

I think Elo does a much better job than 70%, but I'd have to run the numbers to confirm. I'll report results if I get around to it today.

I just ran the numbers on all matches from 1968 - 2018 (using my own Elo, which shouldn't be too different from UTS). Got 70.1% accuracy on Elo predictions. I can admit when I'm wrong! Fair play to you. :)

I still think Elo gets unfairly dismissed on this site, though. It has value when used properly. And just because it's used doesn't mean it's used improperly.
 

AnOctorokForDinner

Talk Tennis Guru
I just ran the numbers on all matches from 1968 - 2018 (using my own Elo, which shouldn't be too different from UTS). Got 70.1% accuracy on Elo predictions. I can admit when I'm wrong! Fair play to you. :)

I still think Elo gets unfairly dismissed on this site, though. It has value when used properly. And just because it's used doesn't mean it's used improperly.

I just recalled what Cekovic of UTS said (he posts here occasionally). Sackmann of TA said something similar too in one of his many articles, that the prediction rate of his best algorithms is slightly above 70%.

That's how it should be given the complexity of tennis, wouldn't you think? If it ever becomes significantly more predictable, something will have to have been ruined for that to happen. A highly stratified tour isn't very interesting when most matches aren't even competitive. Take Ferrer, who reached a very respectable ELO peak of 2348 per UTS in early 2013, on the basis of a strongly consistent year, all the while being the absolute Big 3 bunny: from 2012 AO to 2013 Acapulco (peak Ferrer ELO), poor Daveed went 1-24 in sets against them, and the lone set he won happened in abnormal hurricane conditions at the USO against Djokovic, who promptly restored order the following day anyway. In comparison, Wawrinka's peak UTS ELO is 2291 - a whole 57 points worse - but we all know which of the two was the far, far better competitor when it mattered most at the business end of slams. This is my go-to example of why tennis ELO isn't a very precise tool at all when applied to measure competition.
 

daggerman

Hall of Fame
Take Ferrer, who reached a very respectable ELO peak of 2348 per UTS in early 2013, on the basis of a strongly consistent year, all the while being the absolute Big 3 bunny: from 2012 AO to 2013 Acapulco (peak Ferrer ELO), poor Daveed went 1-24 in sets against them, and the lone set he won happened in abnormal hurricane conditions at the USO against Djokovic, who promptly restored order the following day anyway. In comparison, Wawrinka's peak UTS ELO is 2291 - a whole 57 points worse - but we all know which of the two was the far, far better competitor when it mattered most at the business end of slams. This is my go-to example of why tennis ELO isn't a very precise tool at all when applied to measure competition.

That's the thing. Even if all of that is fair enough, I don't think the conclusion to draw from those criticisms is in alignment with the conclusions that are often drawn about Elo. A few thoughts:
  • It's true that Wawrinka was less consistent overall than Ferrer. That fact is captured by their differences in peak Elo.

  • Recent Elo rating is a better tool than vanilla Elo for evaluating a player's form at a given time. This is my main response to the criticisms of Elo posed by Towny in that seminal post. It's probably a better predictive tool in general, but I'm only assuming that's the case.

  • While it's also true that Elo values consistency, citing Ferrer should also be accompanied by the caveat that Ferrer is probably the most extreme case of high consistency, low peak level in men's tennis history. I'd wager that nobody with a peak Elo as high as 2348 has a peak level as low as Ferrer. I also suspect nobody with a peak Elo as low as 2291 has a peak level as high as Wawrinka (except maybe Safin. I don't know what his peak Elo is off hand). All of that is to say Elo does better in player-to-player comparisons than the comparison between those two particular players would suggest. Still, I understand the pitfalls of using Elo for player-to-player comparisons in general. Your point is not lost on me.

  • Identifying individual cases where Elo falls short doesn't address Elo as a measure of overall competition. Two sets of opponents having an average Elo difference of 100 is more meaningful than two individuals having an Elo difference of 100. If one player plays a set of opponents with an average Elo of 2000, then arguing that they didn't face a stronger set of opponents (on average) than a player with an average opponent Elo of 1900 would require a different kind of reasoning than what I've seen so far. This is, in fact, the best use of Elo imo, but it often gets lumped in with the "bad uses" despite being a fundamentally different application of the metric.

tl;dr: Elo has flaws but is useful. I encourage users on here to not be afraid to use Elo rating. If you use it correctly and get grief for it, I'll have your back :sneaky:
 

falstaff78

Hall of Fame
Most people would probably guess offhand that Federer had the superior decade at Wimbledon, but did the Serbian Superstar in fact outdo the Swiss Maestro?

Federer 2000-2009
52-4 (92.8%) record
6 Wimbledon titles, 1 final, 0 semifinals
Opponents defeated in finals: Philippoussis, Roddick, Roddick, Nadal, Nadal, Roddick
4.16 average # of sets needed to win finals

Djokovic 2010-2019
57-4 (93.4%) record, 1 injury retirement
5 Wimbledon titles, 1 final, 2 semifinals
Opponents defeated in finals: Nadal, Federer, Federer, Anderson, Federer
4.2 average # of sets needed to win finals

Mean Opponent Elo
2014 Wimbledon Djokovic 2098.6
2015 Wimbledon Djokovic 2095.3
2018 Wimbledon Djokovic 2068.7
2005 Wimbledon Federer 2068.3
2006 Wimbledon Federer 2060.3
2009 Wimbledon Federer 2045.4
2011 Wimbledon Djokovic 2040.6
2007 Wimbledon Federer 2034.2
2019 Wimbledon Djokovic 2025.6
2003 Wimbledon Federer 1985.7
2004 Wimbledon Federer 1978.0

6>5

(Really nice analysis by the way)
 

AnOctorokForDinner

Talk Tennis Guru
That's the thing. Even if all of that is fair enough, I don't think the conclusion to draw from those criticisms is in alignment with the conclusions that are often drawn about Elo. A few thoughts:
  • It's true that Wawrinka was less consistent overall than Ferrer. That fact is captured by their differences in peak Elo.

  • Recent Elo rating is a better tool than vanilla Elo for evaluating a player's form at a given time. This is my main response to the criticisms of Elo posed by Towny in that seminal post. It's probably a better predictive tool in general, but I'm only assuming that's the case.

  • While it's also true that Elo values consistency, citing Ferrer should also be accompanied by the caveat that Ferrer is probably the most extreme case of high consistency, low peak level in men's tennis history. I'd wager that nobody with a peak Elo as high as 2348 has a peak level as low as Ferrer. I also suspect nobody with a peak Elo as low as 2291 has a peak level as high as Wawrinka (except maybe Safin. I don't know what his peak Elo is off hand). All of that is to say Elo does better in player-to-player comparisons than the comparison between those two particular players would suggest. Still, I understand the pitfalls of using Elo for player-to-player comparisons in general. Your point is not lost on me.

  • Identifying individual cases where Elo falls short doesn't address Elo as a measure of overall competition. Two sets of opponents having an average Elo difference of 100 is more meaningful than two individuals having an Elo difference of 100. If one player plays a set of opponents with an average Elo of 2000, then arguing that they didn't face a stronger set of opponents (on average) than a player with an average opponent Elo of 1900 would require a different kind of reasoning than what I've seen so far. This is, in fact, the best use of Elo imo, but it often gets lumped in with the "bad uses" despite being a fundamentally different application of the metric.

tl;dr: Elo has flaws but is useful. I encourage users on here to not be afraid to use Elo rating. If you use it correctly and get grief for it, I'll have your back :sneaky:

Comparing sets of players by ELO sure helps to remove the effect of outliers but not that it values consistency as much as peak. Since ELO is close to zero sum as every match redistributes points - the loser gains as much as the winner loses, with the sum of both players' points staying the same (not exactly zero sum since retiring players don't leave the game with exactly the same rating they started, but the inflation effect wouldn't be significant in tennis I figure) - what it really measures then is the stratification degree of the tour. Top players having high elo being consistent means lower-ranked players have lower elo being less successful. Same as the official points ranking, again really. It doesn't necessarily capture competitiveness since neither very high nor very low stratification is optimal. A highly stratified tour with low upset score means better players usually beat lesser players handily, since if more matches were close there'd more upsets (a specific player may go on a streak of winning close matches but a whole set of players constantly edging another set of players in tight encounters is exceedingly improbable so as to be practically impossible), which means lesser competitiveness. A highly upset-prone tour lacking consistency even when adjusted for surfaces suggests lower level of play, since a player who played really well to win is unlikely to drop so much in short time as to lose in early rounds a couple times afterwards - again there have been specific high-profile 'high highs, low lows' players like that, but when it's a general trend it's rather likely the winning level isn't so high overall, since an alternative explanation would require several genuine Safin/Wawrinka types to be present at the same time, which is highly unlikely at the top level, or a narrower gap between lesser players like #30 and top players like #1, which itself suggests that top players are worse as the alternative that it's #30s getting better closer to #1s is historically implausible. It's best to see a happy medium, even if we have no exact calculations for it as of yet.
 
Top