Men's Elo Ratings Open Era

[Wuornos]

The following are the peak ELO ratings for male players in the open era. ELO ratings measure dominanace with an adjustment for standard of opposition. I have been asked whether it would be possible to post the top 100 using this system, devised by Dr Arpad Elo, in one of my previous posts.


1 Roger Federer 2796
2 Ivan Lendl 2789
3 Rod Laver 2783
4 Pete Sampras 2769
5 Björn Borg 2768
6 Mats Wilander 2759
7 John McEnroe 2756
8 Stefan Edberg 2751
9 Boris Becker 2745
10 Jim Courier 2739
11 Ken Rosewall 2726
12 Jimmy Connors 2723
13 Arthur Ashe 2711
14 Andre Agassi 2707
15 Tony Roche 2701
16 John Newcombe 2693
17 Rafael Nadal 2690
18 Guillermo Vilas 2685
19 Stan Smith 2682
20 Jan Kodeš 2680
21 Lleyton Hewitt 2678
22 Vitas Gerulaitis 2678
23 Andy Roddick 2676
24 Pat Cash 2674
25 Michael Chang 2672
26 Patrick Rafter 2671
27 Ilie Năstase 2670
28 Miloslav Mečíř 2669
29 Andrés Gimeno 2657
30 Novak Đoković 2654
31 Todd Martin 2654
32 Yevgeny Kafelnikov 2653
33 Marat Safin 2653
34 Juan Carlos Ferrero 2652
35 Henri Leconte 2650
36 Tom Okker 2650
37 Roscoe Tanner 2645
38 Michael Stich 2645
39 Nikolay Davydenko 2641
40 Goran Ivanišević 2639
41 David Nalbandian 2637
42 Johan Kriek 2635
43 Manuel Orantes 2634
44 Carlos Moyà 2633
45 Richard Krajicek 2633
46 Kevin Curren 2631
47 Sergi Bruguera 2627
48 Alex Metreveli 2626
49 Cédric Pioline 2624
50 Clark Graebner 2623
51 Yannick Noah 2623
52 Gustavo Kuerten 2621
53 Thomas Muster 2619
54 Marcos Baghdatis 2618
55 Mark Philippoussis 2618
56 Petr Korda 2618
57 Andrés Gómez 2616
58 Fernando González 2616
59 Tim Henman 2615
60 Thomas Gorman 2614
61 Patrick Proisy 2614
62 Brian Gottfried 2610
63 Raúl Ramírez 2608
64 Víctor Pecci 2608
65 Harold Solomon 2608
66 Mark Edmondson 2606
67 Dick Crealy 2605
68 Cliff Richey 2605
69 Roger Taylor 2605
70 Roy Emerson 2605
71 Mikael Pernfors 2605
72 Željko Franulović 2605
73 Greg Rusedski 2604
74 Chris Lewis 2604
75 Thomas Johansson 2604
76 Dennis Ralston 2604
77 Sjeng Schalken 2603
78 Sébastien Grosjean 2603
79 MaliVai Washington 2602
80 Darren Cahill 2602
81 José Luis Clerc 2602
82 Corrado Barazzutti 2601
83 Marcelo Ríos 2601
84 Pancho Gonzales 2600
85 Adriano Panatta 2600
86 Malcolm Anderson 2599
87 Aaron Krickstein 2599
88 Ray Ruffels 2598
89 Cliff Drysdale 2598
90 Patrick McEnroe 2598
91 Rod Frawley 2598
92 Onny Parun 2597
93 Emilio Sánchez 2597
94 Jonas Björkman 2597
95 Magnus Norman 2596
96 Slobodan Živojinović 2595
97 Richard Gasquet 2594
98 Arnaud Clément 2594
99 Brian Teacher 2594
100 Fred Stolle 2593

For a fuller explanation of how Elo ratings are calculated see the following:

http://en.wikipedia.org/wiki/Elo_rating

For further discussion on Tennis Elo ratings see the following:

http://tt.tennis-warehouse.com/showthread.php?t=162345

Many evalutae tennis players by the number of majors won, but Elo adjusts for dominance and quality of opposition therefore the list will not be purely representative of achievements in majors but we would expect to see some correlation between the two.

In the men's game :

the top 5 have won 50 majors in the open era
6-10 won 30
11-15 won 23
16-20 won17


Just thought people might be interested in these statistics which are generated from my own Elo database and are purely evidentially based without interference of human opinion.

I am not recommending ELO as a substitute to the official ranking methodology, but am posting the results purely for information purposes and to provide a different but independent perspective on the evaluation of players.

Take care

Tim

 

[pabloJD]

Great! Some surprises for me there, though

52 Gustavo Kuerten 2621
53 Thomas Muster 2619
54 Marcos Baghdatis 2618 --- how is Baghdatis' peak so high??

or

80 Darren Cahill 2602
83 Marcelo Ríos 2601 --- really? Cahill above Rios? That IS surprising

Thanks Tim!

 

[urban]

Speaking of high level of competition. In the top 20 of this list ,Federer has 1 (one) contemporary, Laver has 6 (six).

 

[justineheninhoogenbandfan]

Speaking of high level of competition. In the top 20 of this list ,Federer has 1 (one) contemporary, Laver has 6 (six).

Yeah somewhat more likely a long retired player whose contemparies are long retired would have more people make the list, then a current player who has many opponents who have yet to finish their careers. You are a genius arent you.

 

[urban]

Now, Mr. Genius, i don't understand this rating at all, but it seems, as it is said, to be peak ratings. Has nothing to do with career finishes or so. Or do you think, that Roddick will reach his peak in the next 3-4 years.

 

[justineheninhoogenbandfan]

Now, Mr. Genius, i don't understand this rating at all, but it seems, as it is said, to be peak ratings. Has nothing to do with career finishes or so. Or do you think, that Roddick will reach his peak in the next 3-4 years.

I dont think Roddick will, but it isnt impossible, crazier things have happened. More likely though Nadal, Djokovic, and Murray all could in that time.

Anyway what are you defining as needing be considered an opponent. The funny thing is Rosewall as a rival to Laver in his grand slam year in 1969 is trumpeted, yet Agassi as an opponent for Federer in 2004-2005 is dismissed due to his age, yet they were almost exactly the same age. Also if Agassi were really hypotheticaly still "past it" at that age, and Rosewall was not, then that spells even more how inferior Agassi must be to Rosewall, since Rosewall maintained his whole career at a high level, and if Agassi with all those tank job years still is past it at an age Rosewall was still apparently strong, then that speaks even worse to him in comparision. After all the longevity arguments has come into play for both.

 

[urban]

Now, Laver and Rosewall have 3-4 years of age in between. How many has Agassi against Federer, 10 or 12 or so. OK, give Federer Agassi, then you have to give Laver Borg, Vilas or some more of the 70s, all guys, whom Laver played and beat. Thing is, if someone makes a list, to support Federer, and some other reads the same list in another way, than it has been meant, some of the Fedfans begin to howl. Going by this somewhat strange list, made by ELO and some human opinion kit in between, all the 60s, 70s, 80s and 9os, have more people in the top twenty than the actual contemporaries. Its not my fault, i am only a reader, not the originator of this list.

 

[chaognosis]

For this to be in any way credible, you are going to need to provide some details about precisely what statistics you are using and how you are fitting them into this formula. Obviously accurate numbers for Laver, Rosewall, etc., aren't nearly as accessible as those for current players--so where are you getting them? And why, then, can't you extend this back further to look at pre-open era players? Also, wasn't Elo's original system based on five-year performance rather than an individual peak? And didn't he explicitly state that this system could only be used to rate players against their contemporaries--that comparing players of different generations was impossible?

 

[Wuornos]

Speaking of high level of competition. In the top 20 of this list ,Federer has 1 (one) contemporary, Laver has 6 (six).

Yes but most of Laver's contempraries achieved their peak rating after his. There is still time for Federer's contemoraries to do this. Also even if this were the case I'm unclear of what your point is.

I think I made the point in a previous post that Federer's rating suffers from lower than normal oppostion (ratings). As did Sampras.

See post #28 in http://tt.tennis-warehouse.com/showthread.php?p=1823483#post1823483

Take care and thanks for your interest and query.

Tim

 

[Wuornos]

Yeah somewhat more likely a long retired player whose contemparies are long retired would have more people make the list, then a current player who has many opponents who have yet to finish their careers. You are a genius arent you.

Oh yes! That's what I just said !

 

[Wuornos]

Now, Mr. Genius, i don't understand this rating at all, but it seems, as it is said, to be peak ratings. Has nothing to do with career finishes or so. Or do you think, that Roddick will reach his peak in the next 3-4 years.

Yes that's right peak rating. Roddick's end of year ratings for the past 4 years have been as follows:

2004 2646
2005 2658
2006 2664
2007 2667 (his peak was also during this year at 2676)

He does seem to be making slow but steady progress but is being left behind by both Federer and Nadal that makes him appear less succesful, and some interpret that as a decline. E.g At the end of 2005 when Roddick was 2658, Federer was 2721 (63 points ahead of Roddick) and Nadal was at 2589 (69 points behind Roddick) having just exploded onto the scene. Now Roddick is at 2667 and Federer is up at 2796 (129 points clear) and Nadal is at 2690 (23 points ahead).

He might hit a higher peak in the next few years yes but I doubt this will be particularly noticable given the fact he plays in the shadow of Federer.

Had he played at a different point in history I believe his achievements would have been much greater.

Take care

Tim

PS I'm not a genius, that was Arpad Elo. I'm just standing on his shoulders and utilising his formulae.

 

[Wuornos]

For this to be in any way credible, you are going to need to provide some details about precisely what statistics you are using and how you are fitting them into this formula. Obviously accurate numbers for Laver, Rosewall, etc., aren't nearly as accessible as those for current players--so where are you getting them? And why, then, can't you extend this back further to look at pre-open era players? Also, wasn't Elo's original system based on five-year performance rather than an individual peak? And didn't he explicitly state that this system could only be used to rate players against their contemporaries--that comparing players of different generations was impossible?

Great criticism Chaog. Some good points. I always respect your posts relating to tennis history. I'll answer them in turn.

1. What statistics am I using. The only statistic required for Elo's formula is the names of the two players in a match, the date the match was played in order to achieve correct chronolgy and the result. I make no allowance for if the score was a double bagel or whatever, just as the original formula didn't.

2. Availavble numbers(?) for Laver and Rosewall etc. If you mean available results, then yes they are more difficult to get the further I get back in time. I'm not going to lie about that. This may remove a certain element of accuracy of the players in the early years but would not systematically inflate or deflate their ratings as a missing result from the database is as likely to improve a rating as decrease it. What it does mean is that while we might have confidence in the numbers calculated to be accurate as true representation of standard to an accuracy of +/- 5 points for later players, this accuracy might increase to +/- 20 points for earlier years.

3. Your final point, Elo's statement regarding how his formula can only be used in comparing contemporaries. Yes Elo did only guarantee the ratings to be valid for contemporaries and this was because it was impossible to guarantee that the overall population was not improving or worsening overtime. These ratings make an assumption that the standard of the population as a whole is constant. Now this may be true or it may not, but most posts on this site evaluate 'greatness' of players on their achievement or dominance of their contemporaries and do not say that 'the general population is better now so Laver was a nobody'. I would go along with this as an assumption to be used in assesing greatness and if we are to take this assumption as being read then the ELO rating as a measure of dominance adjusted for dominance of opposition faced is valid providing the population median is monitored and maintained as a constant. Elo was only driving at the absolute values of ratings not being comparable, that is it does not prove that Borg would beat Federer if they met at their peak. However, I do not agree with Elo when he said that the rating of players cross era was impossible. To do this we would have to adopt an additional adjuster for slope of line of improvement, e.g. 20 points per year, but this is something I would not be willing to do as it would not make interesting reading and would fly in the face of what people mean by 'greatness'. If you were wanting to do this it can be achieved by looking at the world record holders across a large range of sports assuming that tennis is typical in its rate of improvement and adjusting the slope of the improvement to achieve a comparable level of world record holders. Crude I know, but I can't think of any other way of doing that. Do you have any ideas? I have done some work on this but to be honest it doesn't make interesting reading, it just means the top 100 players are dominated to a great extent by players from the last 10 years. Not what people mean by greatness at all.

Strangely enough despite Elo's stating that his formula could not be relied upon to compare cross era performance because of the moving population standard he then went on to compose a list of the highest ELO rated chess players of all time based on a 5 year moving horizon in his book 'the rating of chess players past and present'. As I no longer have a copy of this book, I lost it some 15 years ago in a house move, I am unable to check whether he qualified this list in the way I would expect by stating that it assumes a constant population median.


4, Now we will return to your third point. I would love to take things back prior to the open era. However the problem is that there are two distinct populations before this point, the amateur and the professional. As highlighted in point 3 for ratings to be a valid indicator of adjusted domination it is important to ensure the population median remains constant. If we have two populations we would also need to make sure the population medians are in harmony with the relative standards of each group, and of course this relationship is something that would also change over time. For me this is a step to far to have confidence in the outputs and despite giving a great deal of thought to the problem I can't find a way of calculating this without human opinion creeping in. The crossover data i.e when one player switches from amateur to professional is just to small to be sufficiently robust and where we do have a larger sample of data between the two populations, i.e. when the open era started, this is only a single point in time and can be considered as representative for only two years at best. Let me know if you think of a way around this problem.

5. Finally your penultimate point. Yes Arpad Elo in his book 'The rating of chess players, past and present' did make calculations based on a rolling 5 year average. However this was adapted to have a more gradual fading out of past results by the use of the k factor. This avoids the problem of a result from 5 years ago being given the same value as a result from last week in current ratings, it also enables players with fewer matches who are normally assigned a higher K factor to achieve a representatitive rating. Without this Laver would not qualify for a rating having less than 5 years results in the open era and players with less than 5 years experience on the tour would fail to appear on the rating list. I'm also sure you don't need me to point out that this original methodology was not really fair as players who may have played 5 games in a 5 year average would have far greater variation than those who had played 500 games. The reducing K factor however, allows for a reasonable ratings estimate for players with less than 5 year results or limited numbers of games/matches. I believe Elo was part of this implementation of the K factor for the USCF rather than the five year average.

Finally to re-iterate what is a very good point being made by Chaog. These ratings do not reflect who would have beaten whom had they met at their peak. They just show a slightly new take on domination of contemporaries allowing for the domination levels of the opposition faced for the player being calculated. A second iteration of domination if you will.

Thanks Chaog. I always look forward to your posts. Some great questions and points made.

Regards

Tim

 

[Wuornos]

Now, Laver and Rosewall have 3-4 years of age in between. How many has Agassi against Federer, 10 or 12 or so. OK, give Federer Agassi, then you have to give Laver Borg, Vilas or some more of the 70s, all guys, whom Laver played and beat. Thing is, if someone makes a list, to support Federer, and some other reads the same list in another way, than it has been meant, some of the Fedfans begin to howl. Going by this somewhat strange list, made by ELO and some human opinion kit in between, all the 60s, 70s, 80s and 9os, have more people in the top twenty than the actual contemporaries. Its not my fault, i am only a reader, not the originator of this list.

I'm not sure I understand the points about age etc. Sorry.

I'm not sure what you mean by 'made by elo and some human opinion kit in between'. Sorry again.

I understand your last point and while this is true now it may not be true in the future as the current contempraries may yet hit a higher peak taking them into the top 20. Players currently climbing fast and who may be future open era top 20 players i.e in excess of 2680 at peak are as follows:

1. Novak Đoković 2654
2. Nikolay Davydenko 2641
3. Richard Gasquet 2594
4. David Ferrer 2590
5. Tommy Robredo 2578
6. Tomáš Berdych 2553

I'm not saying they will make the top 20 I'm just saying they were all big climbers in 2007 and if only 2 break the 2680 barrier then they will join Federer and Nadal in the top 20.

Actually this type of period is not unusual in the sport. I'm not sure why it happens but I have read about it being referred to in other sports as the changing of the guard. I suppose occasionally you will get all the top players drop out of the game in quick succesion and when this happens it takes time for the void to be filled. A similar type of thing happened in the mid 1970s when suddenly the aging greats seemd to be surpassed by Jimmy Connors and then Bjorn Borg. At that point in time it would have been possible to say that there were only 2 contemoraries in the top 10 and that it was disproportional. It took a while but we eventually got Guilermo Vilas, John McEnroe and Vitas Gerulaitis and suddenly the era was again represented at the top of the game.

I just think we're at that stage again now and we will see players emerge as Federer / Nadal top 20 contemporaries. Incidentally only 2 of the top 10 Elo rated players at the end of 2000 are still active now, i.e. Carlos Moya and Marat Safin. This sort of bears out what I was saying about 'the changing of the guard'.

Thanks for the interest

Tim

 

[chaognosis]

Thanks for your thorough reply--what you are doing makes a lot more sense to me, now. I am still curious how you actually make all these computations; it seems like it would be an awful lot of work, tracking down every match ever played in the open era and doing the necessary math in each case... I am sure you have some sort of computerized method, but I am probably not knowledgeable enough to really understand the technological side of things.

As for devising an improvement slope, I have talked about this very problem with SgtJohn before and I think both of us are a bit befuddled by it. I agree though that ultimately greatness has to be understood in the context of its own time--a player stands or falls based on his performance against his peers--and I would not be interested in any "all-time" ranking dominated by players from the past decade.

 

[Wuornos]

Thanks for your thorough reply--what you are doing makes a lot more sense to me, now. I am still curious how you actually make all these computations; it seems like it would be an awful lot of work, tracking down every match ever played in the open era and doing the necessary math in each case... I am sure you have some sort of computerized method, but I am probably not knowledgeable enough to really understand the technological side of things.

As for devising an improvement slope, I have talked about this very problem with SgtJohn before and I think both of us are a bit befuddled by it. I agree though that ultimately greatness has to be understood in the context of its own time--a player stands or falls based on his performance against his peers--and I would not be interested in any "all-time" ranking dominated by players from the past decade.

Thanks Chaog. Yes I do utilise a computer as the there are literally hunderds of thousand of calculations necessary to cover the whole open era.

The database i have contains results from varying sources and I'm constantly on the look out for errors or anomolies within the records. Some of these results I have keyed manually, good god that was time consuming. Some have been given to me by tennis enthusiasts etc and some I have bought from people who had compiled their wown databases.

I am always in the lookout for good quality data both new and as a cross reference to check data that is already included in my system, so if you hear of anyone with a large database of results in access or excel or even csv that they wish to sell, let me know. I'm always interested.

Regards

Tim

 

[racquetfreak]

no system is perfect

interesting but no cigar. too many variables are left out because they really aren't quantifiable in an objective way. no system will ever be acceptable to all. this one is pretty good at the very top but quickly breaks down as you go below the top ten and further back in time. good example is Pancho Gonzales.

 

[hoodjem]

The following are the peak ELO ratings for male players in the open era.

I too had an issue with this ranking until I read the fine print, that is the first line. Notice it says "open era."

So leaving off Pancho, or placing Laver after Lendl (not on my GOAT list, thank you) makes more sense under these circumstances.

Indeed, I would offer that ranking Laver so high based ONLY on his open era performances is further testimony to his greatness.

 

[chaognosis]

Indeed, I would offer that ranking Laver so high based ONLY on his open era performances is further testimony to his greatness.

This impressed me too. Laver attained a higher rating after the age of 30 than Sampras or Borg ever reached even at their absolute peaks. Imagine how good Laver was during his actual prime in the mid-1960s!

 

[hoodjem]

My thoughts exactly!

And if one could add his pre-Open Era (amateur and professional) achievements (not inconsiderable) to his Open Era record, then he might top such a list by a considerable margin.

 

[Moose Malloy]

ineresting post Wuornos. There used to be a site 'setratings.com' that used your system(its also used with chess correct?) as well & he came out with different rankings than you did(Laver did not come out that well)

He used to post here & explained that losses hurt a players 'peak' performance ranking. Laver lost a lot in his best years of the open era, while Fed, Mac, & Borg did not. They were the top 3 in his system, by a rather large margin. Sampras was far back as well.

 

[urban]

Yes, i remember this setratings. The poster had Mac as Nr.1 with the best peak rating. If i remember it correctly, he had his data base put together only since 1970, with more results availabe on the ATP webside. On pure match percentage over a couple of years, Borg is still unsurpassed in open era.

 

[Wuornos]

interesting but no cigar. too many variables are left out because they really aren't quantifiable in an objective way. no system will ever be acceptable to all. this one is pretty good at the very top but quickly breaks down as you go below the top ten and further back in time. good example is Pancho Gonzales.

Yes, but Pancho Gonzalez was pre open era and these ratings only include performances of the open era and so his peak rating will only reflect his play during the open era and well past his peak.

Thanks for the interest.

Tim

 

[Wuornos]

ineresting post Wuornos. There used to be a site 'setratings.com' that used your system(its also used with chess correct?) as well & he came out with different rankings than you did(Laver did not come out that well)

He used to post here & explained that losses hurt a players 'peak' performance ranking. Laver lost a lot in his best years of the open era, while Fed, Mac, & Borg did not. They were the top 3 in his system, by a rather large margin. Sampras was far back as well.

Thank's Moose it would be interesting to contact the guy and be able to discuss where the systems differ and why. I suspect either his or mydatabase will have more or less results included. E.g. I never include data from exhibition matches as I do not see them as truly reflective of player performance.

I can understand Sampras' rating not being as high as people expect due to the limited competition when he was at his peak i.e. other players in excess of 2700 points using my base, but I doubt he should be outside the top 5.

Thanks for the feedback on SetRatings though.

Appreciated.

Tim

 

[Wuornos]

My thoughts exactly!

And if one could add his pre-Open Era (amateur and professional) achievements (not inconsiderable) to his Open Era record, then he might top such a list by a considerable margin.

I have no doubts given the rapid rise of Laver's Elo points in the early open era and given this was at the tail end of his career that he would easily top the list if I could find some way to extend the calculations in order to cover the pre open era. Unfortunately I haven't found a way yet that is sufficiently robust to have confidence in the data without introducing human opinion into the calculations and that is something I will not do under any circumstances.

Thanks for the positive feedback.

Tim

 

[Wuornos]

Yes, i remember this setratings. The poster had Mac as Nr.1 with the best peak rating. If i remember it correctly, he had his data base put together only since 1970, with more results availabe on the ATP webside. On pure match percentage over a couple of years, Borg is still unsurpassed in open era.

Thanks Urban

I agree on pure match percentage over several years Borg is hard to beat.

If SetRatings didn't have data prior to 1970 this would go a long way to explaing why Laver was lower than on my list. My list begins in 1968 and by excluding these two years in Setratings it excludes Laver's peak performance period of the open era.

Take care and thanks again.

Tim

 

[shakes1975]

very interesting analysis. i would offer just one angle of thought. when using this rating, it's hard to account for a player's attitude towards some tournaments. In essence, it's useful to measure how consistent a player is/was, but not his peak abilities.

for eg., it's well known that sampras played at only about 70-75 % of his full ability in most of the non-slam tournaments, and that could've prevented him from dominating more than he did. But his slam results are clearly superior to every one else. So, his rankign doesn't actually reflect his peak achievements. Rather, it just shows that he wasn't as consistent, week-in and week-out, as lendl or fed.

similarly, another example would be becker, edberg. edberg ranking above becker is truly surprising. Becker won more number of total tournaments, has a leading h2h against edberg, has a superior davis cup record, and won 2 slams in a yr at least once, finished 11 yrs in top-10 to edberg's 10, and both finished 9 yrs in top-5. Also, the following link shows that becker had a better win % in all slams except the AO: http://www.tennis28.com/slams/winpct_slam.html

 

[Wuornos]

very interesting analysis. i would offer just one angle of thought. when using this rating, it's hard to account for a player's attitude towards some tournaments. In essence, it's useful to measure how consistent a player is/was, but not his peak abilities.

for eg., it's well known that sampras played at only about 70-75 % of his full ability in most of the non-slam tournaments, and that could've prevented him from dominating more than he did. But his slam results are clearly superior to every one else. So, his rankign doesn't actually reflect his peak achievements. Rather, it just shows that he wasn't as consistent, week-in and week-out, as lendl or fed.

similarly, another example would be becker, edberg. edberg ranking above becker is truly surprising. Becker won more number of total tournaments, has a leading h2h against edberg, has a superior davis cup record, and won 2 slams in a yr at least once, finished 11 yrs in top-10 to edberg's 10, and both finished 9 yrs in top-5. Also, the following link shows that becker had a better win % in all slams except the AO: http://www.tennis28.com/slams/winpct_slam.html

Hi there Shakes

Your first point is very good. I think your correct that Sampras did play better in the slam events and didn't give his full attention to other lesser tournaments. As you quite rightly identify this will have lowered his rating to a certain extent. But then I'm not sure what can be done about this. Any rating system has to evaluate results and results only and if Sampras lost some matches then that needs to be reflected. Perhaps you could explain how measuring consistency can help this as I am a bit unclear. I'm interpreting consistency as meaning a measure of results over a long period, but this would also include his losses at lesser tournaments. Also perhaps you are misunderstanding what is meant by Peak Elo. Elo rating is a measure of consistency, I am just showing what the peak elo standard was during his career. The peak elo standard would reflect results over sevral years with more distant results having less weighting than more recent. The point about his lack of emphasis on lesser events though is well made and will have an effect on his rating.

Your second point regarding edberg and becker is a little more nebulous to quantify and therefore more difficult to disprove or prove. First of all Elo is a measure of dominance over time. So whether Becker won more tournaments in the open era isn't really relevant. The question is how dominant was he in winning them, that is what was his strike rate and over how maqny events. If pure tournament count in the open era was significant, Laver would be outside the top 10. Next, in my opinion Head to head counts mean very little. By definition they are a very small sample and therefore are not sufficiently robust to draw many conclusions from. Most of the matches are played when one player is at their peak and the other isn't and finally there's the rock paper scissors effect. 2 Slams in a year as opposed to one is again a sample of just 4 tournaments, albeit big tournaments. Therefore not really significant in isolation . If you throw a double six once does it mean the dice are loaded or is it a small sample? Again your coiunt of years in the top 10 is selective and would lead to some leading players dropping down the rankings e.g. does this prove Borg was not as good as becker?

I could go on but the fact is that your first point was absolutely correct. The only measure of a players worth is to monitor a large number of matches over a protracted period of time and look at the quality of opposition. This is what Elo ratings do.

Thanks for the interest and sorry we can't see eye to eye on your second point.

Regards

Tim

 

[Wuornos]

very interesting analysis. i would offer just one angle of thought. when using this rating, it's hard to account for a player's attitude towards some tournaments. In essence, it's useful to measure how consistent a player is/was, but not his peak abilities.

for eg., it's well known that sampras played at only about 70-75 % of his full ability in most of the non-slam tournaments, and that could've prevented him from dominating more than he did. But his slam results are clearly superior to every one else. So, his rankign doesn't actually reflect his peak achievements. Rather, it just shows that he wasn't as consistent, week-in and week-out, as lendl or fed.

similarly, another example would be becker, edberg. edberg ranking above becker is truly surprising. Becker won more number of total tournaments, has a leading h2h against edberg, has a superior davis cup record, and won 2 slams in a yr at least once, finished 11 yrs in top-10 to edberg's 10, and both finished 9 yrs in top-5. Also, the following link shows that becker had a better win % in all slams except the AO: http://www.tennis28.com/slams/winpct_slam.html

Hi there Shakes

Your first point is very good. I think your correct that Sampras did play better in the slam events and didn't give his fullattention to other lesser tournaments. As you quite rightly identify this will have lowered his rating to a certain extent. But then I'm not sure what can be done about this. Any rating system has to evaluate results and results only and if Sampras ;lost some matches then that needs to be reflected. Perhaps you could explain how measuring consistency canb help this as I am a bit unclear. I'm interpreting consistency as meaning a measure of results over a long period, but this would also include his losses at lesser tournaments. Also perhaps you are misunderstanding what is meant by Peak Elo. Elo rating is a measure of consistency, I am just showing what ythe peak elo standard was during his career. The peak elo standard would reflect results over sevral years with more distant results having less weighting than more recent. The point about his lack of emphasison lesser events though is well made and will have an effect on his rating.

Your second point regarding edberh and becker is a little more nebulous to quantify and therefore more difficult to disprove or prove. Fist of all Elo is a measure of dominance over time. So whether Becker won more tournaments in the open era isn't rreally relevant. The question is how dominant was he in winning them, that is what was his strike rate and over how short a timescale. If pure tournament count in the open era was significant, Laver would be outside the top 10. Next, in my opinion Head to head counts mean very little. By definition they are a very small sample and therefore are not sufficiently robust t draw many conclusions from. Most of the matches are played when one player is at their peak and the other isn't and finally there;s the rock paper scissors effect. 2 Slams in a year as opposed to one is agian a sample of 4 tournaments. Not really significant. If you throw a double six once does it mean the dice are loaded or is it a small sample? Again your coiunt of years in the top 10 is selective and would lead yto some leading players dropping down the rankings e.g. does this prove Borg was not as good as becker?

I could go on but the fact is that your first point was absolutely correct. The only measure of a players worth is to monitor a large number of matches over a protracted period of time and look at the quality of opposition. This is what Elo ratings do.

Thanks for the interest and sorry we can't see eye to eye on your scond point.

Regards

Tim

 

[Steve132]

very interesting analysis. i would offer just one angle of thought. when using this rating, it's hard to account for a player's attitude towards some tournaments. In essence, it's useful to measure how consistent a player is/was, but not his peak abilities.

for eg., it's well known that sampras played at only about 70-75 % of his full ability in most of the non-slam tournaments, and that could've prevented him from dominating more than he did. But his slam results are clearly superior to every one else. So, his rankign doesn't actually reflect his peak achievements. Rather, it just shows that he wasn't as consistent, week-in and week-out, as lendl or fed. QUOTE]

Um.... no, unless you interpret "slam results" to mean simply the total number of Slams won. Sampras holds that record primarily because (a) Laver was banned as a pro from participating in Slams for a number of his peak years; (b) Borg retired at age 26 (and played in the Australian Open only once); and (c) Federer is just 26 years old.

Borg has the best record of all Open era players in terms of Slam performance in relation to the number of Slams played. This is true whether we measure performance by wins only or by his overall record. Federer has won 3 Slams in a year in three of the past 4 years. Sampras never achieved this feat. Federer is the only player in tennis history (male or female) to win 11 Slams in 4 years, and the only man in history to reach 10 consecutive Slam finals (the runner up is Jack Crawford with 7 in the 1930's, while the Open era runner-up mark is just 4). These records seem to me to be far more impressive achievements than a career total of 14 Slams.

 

[urban]

In Tennis mag recently was shown another statistical approach to peak performance. It was based on the new ATP points system, and showed the highest points total, a player had achieved. The date line was 1973 and the inventing of the computer ranking.Federer was leading quite clearly here, ahead of Sampras. The disadvantage of the 70s players was the including of all the majors, with the AO problem 1977-1982. For the pre 1973 era and the 70s, you have to find equivalents for Masters or Super Nine events, to get a comparable picture.

 

[Wuornos]

In Tennis mag recently was shown another statistical approach to peak performance. It was based on the new ATP points system, and showed the highest points total, a player had achieved. The date line was 1973 and the inventing of the computer ranking.Federer was leading quite clearly here, ahead of Sampras. The disadvantage of the 70s players was the including of all the majors, with the AO problem 1977-1982. For the pre 1973 era and the 70s, you have to find equivalents for Masters or Super Nine events, to get a comparable picture.

Thanks Urban. Sounds like the system is a volume based system again otherwise the failure to play the AO wouldn't be a problem. Elo gets round this by only rating a player on the events he actually plays in and his results against the standard of opposition faced.

You know me now though, I'm always interested in any statistical rating system for tennis. Can you let me know what magazine and which issue this was published in as I would like to try to get a copy.

Thanks for the info much appreciated.

Take care.

Tim

 

[Wuornos]

Just a minor point but I had a look at who the biggest Elo gainers were in the top 50 men over the past 12 months.

I thought this might be an indicator as to the players who were improving fastest.

The top 5 in order were

1. Novak Đoković (Improving from 2555 to 2654)
2. Richard Gasquet
3. Tommy Robredo
4. Fernando González
5. Tomáš Berdych

I suppose its difficult to improve by to great an extent if your name is Federer or Nadal! Incidentally Federer went from 2761 to 2796, while Nadal went from 2649 to 2690.

I usually define the greats as the 2700 players of which only Federer is playing at this standard at the moment. However, given the gains seen above I wouldn't be surprised if we see both Rafael Nadal and Novak Đoković past this milestone next year. If so this would take the numebr of players exceeding the 2700 point mark in the open era to 17. I'm also interested to see if Federer can become the first player in the open era to exceed the 2800 point barrier.

Regards

Tim

 

[Wuornos]

Mean Elo Ratings for top 10 players for each year of the open era.

The following are the mean elo ratings for the top 10 players for the end of each year of the open era.

1968 2620
1969 2656
1970 2661
1971 2666
1972 2662
1973 2654
1974 2642
1975 2641
1976 2638
1977 2638
1978 2638
1979 2643
1980 2650
1981 2653
1982 2656
1983 2651
1984 2653
1985 2658
1986 2663
1987 2672
1988 2675
1989 2672
1990 2669
1991 2667
1992 2669
1993 2665
1994 2669
1995 2665
1996 2663
1997 2647
1998 2640
1999 2637
2000 2635
2001 2641
2002 2634
2003 2630
2004 2637
2005 2639
2006 2644
2007 2650 (current)


I thought this might be of interest in showing how some periods are more competitive than others and therefore how simple measures of achievement and dominance can have varying values depending on what period they relate to.

Take care

Tim

 

[Moose Malloy]

Wuornos, it was from the Tennis website, not the magazine. Here is the link:

http://tennisworld.typepad.com/tennisworld/2007/10/the-elite---par.html#comments

 

[Wuornos]

Wuornos, it was from the Tennis website, not the magazine. Here is the link:

http://tennisworld.typepad.com/tennisworld/2007/10/the-elite---par.html#comments

Thanks Moose.

Appreciated.

Tim

 

[urban]

What i don't understand in this tennis magazine online ranking system, is the quite low rating of McEnroe 1984 and Connors 1974. Connors won 3 majors, and i think 3 events, equivalent to Super Nine events: Johannesburg, Wembley and US pro. And he won at least 4 minor events, he didn't play the Masters Cup, won by Vilas. He should have been in the same region as Federer in the last 2-3 years. Mac won 2 majors, 1 final, (no AO), Masters Cup and some Super Nine, too. He must have been at Federer's level, too.

 

[Wuornos]

What i don't understand in this tennis magazine online ranking system, is the quite low rating of McEnroe 1984 and Connors 1974. Connors won 3 majors, and i think 3 events, equivalent to Super Nine events: Johannesburg, Wembley and US pro. And he won at least 4 minor events, he didn't play the Masters Cup, won by Vilas. He should have been in the same region as Federer in the last 2-3 years. Mac won 2 majors, 1 final, (no AO), Masters Cup and some Super Nine, too. He must have been at Federer's level, too.

Yes Urban I was wondering about that too.

1. I suppose at the end of the day the system proposed is replicating the ATP points system so is essentially it is reward based system based on level of activity. I think Andrew states that 'I've estimated the maximum number of ranking points they might have won had the tournaments of the day been awarding ranking points as they do now'. For me the significant word here is estimated and not calculated, I think the answer to your question would probably lie in how the estimate was made.

2. I'm very unsure about the awarding of points to events as a constant. The value of events change over time and the best system should take this into account. In a volume based system its no good awarding the same number of points to the Australian Open in the 1970s as is awarded today. It just wasn't as strong then. When I put together the Elo rankings this was one of the uppermost thoughts in my mind and Elo adjusts according the rating of the players present. That means no, Connors, no Borg, No McEnroe and there's less points floating around at that event to be picked up by the players competing and to gravitate towards the pplayers in the final stages. Undjusting for changes in the value of events deflates the achievements of these 3 great players in particular.

3. No account is made of the proximity of other great players to an individuals achievements. E.g. Winning a major when Connors was the only dominant force in the game receives the same credit as winning when there was Borg, Connors and McEnroe slugging it out. This would devalue the achievements of Lendl, Edberg, Wilander and Becker amongst others. It would alos devalue the achievements of lower ranked players who were active at that time like Pat Cash.

4. The coefficient of correlation was interestig though at 0.8 between peak year and total majors won. I did the same calculation for the ELO rating system and it came out at 0.85. I even toyed with the idea of using the regression model to convert the Elo into major equivalent as I felt that if the data were presented in this format people would find the ratings easier to understand but in the end decided against it as I felt it would create more questions than it answered for casual observer.

5. I think the idea of looking at peak performance is correct but doubt a single year is enough time to eliminate the 'luck' aspect.

6. I was thinking perhaps of sending in an article myself to either this site or a magazine. What do people think?

Regards

Tim

 

[djones]

The following are the peak ELO ratings for male players in the open era. ELO ratings measure dominanace with an adjustment for standard of opposition. I have been asked whether it would be possible to post the top 100 using this system, devised by Dr Arpad Elo, in one of my previous posts.


1 Roger Federer 2796
2 Ivan Lendl 2789
3 Rod Laver 2783
4 Pete Sampras 2769
5 Björn Borg 2768
6 Mats Wilander 2759
7 John McEnroe 2756
8 Stefan Edberg 2751
9 Boris Becker 2745
10 Jim Courier 2739
11 Ken Rosewall 2726
12 Jimmy Connors 2723
13 Arthur Ashe 2711
14 Andre Agassi 2707
15 Tony Roche 2701
16 John Newcombe 2693
17 Rafael Nadal 2690
18 Guillermo Vilas 2685
19 Stan Smith 2682
20 Jan Kodeš 2680
21 Lleyton Hewitt 2678
22 Vitas Gerulaitis 2678
23 Andy Roddick 2676
24 Pat Cash 2674
25 Michael Chang 2672
26 Patrick Rafter 2671
27 Ilie Năstase 2670
28 Miloslav Mečíř 2669
29 Andrés Gimeno 2657
30 Novak Đoković 2654
31 Todd Martin 2654
32 Yevgeny Kafelnikov 2653
33 Marat Safin 2653
34 Juan Carlos Ferrero 2652
35 Henri Leconte 2650
36 Tom Okker 2650
37 Roscoe Tanner 2645
38 Michael Stich 2645
39 Nikolay Davydenko 2641
40 Goran Ivanišević 2639
41 David Nalbandian 2637
42 Johan Kriek 2635
43 Manuel Orantes 2634
44 Carlos Moyà 2633
45 Richard Krajicek 2633
46 Kevin Curren 2631
47 Sergi Bruguera 2627
48 Alex Metreveli 2626
49 Cédric Pioline 2624
50 Clark Graebner 2623
51 Yannick Noah 2623
52 Gustavo Kuerten 2621
53 Thomas Muster 2619
54 Marcos Baghdatis 2618
55 Mark Philippoussis 2618
56 Petr Korda 2618
57 Andrés Gómez 2616
58 Fernando González 2616
59 Tim Henman 2615
60 Thomas Gorman 2614
61 Patrick Proisy 2614
62 Brian Gottfried 2610
63 Raúl Ramírez 2608
64 Víctor Pecci 2608
65 Harold Solomon 2608
66 Mark Edmondson 2606
67 Dick Crealy 2605
68 Cliff Richey 2605
69 Roger Taylor 2605
70 Roy Emerson 2605
71 Mikael Pernfors 2605
72 Željko Franulović 2605
73 Greg Rusedski 2604
74 Chris Lewis 2604
75 Thomas Johansson 2604
76 Dennis Ralston 2604
77 Sjeng Schalken 2603
78 Sébastien Grosjean 2603
79 MaliVai Washington 2602
80 Darren Cahill 2602
81 José Luis Clerc 2602
82 Corrado Barazzutti 2601
83 Marcelo Ríos 2601
84 Pancho Gonzales 2600
85 Adriano Panatta 2600
86 Malcolm Anderson 2599
87 Aaron Krickstein 2599
88 Ray Ruffels 2598
89 Cliff Drysdale 2598
90 Patrick McEnroe 2598
91 Rod Frawley 2598
92 Onny Parun 2597
93 Emilio Sánchez 2597
94 Jonas Björkman 2597
95 Magnus Norman 2596
96 Slobodan Živojinović 2595
97 Richard Gasquet 2594
98 Arnaud Clément 2594
99 Brian Teacher 2594
100 Fred Stolle 2593

For a fuller explanation of how Elo ratings are calculated see the following:

http://en.wikipedia.org/wiki/Elo_rating

For further discussion on Tennis Elo ratings see the following:

http://tt.tennis-warehouse.com/showthread.php?t=162345

Many evalutae tennis players by the number of majors won, but Elo adjusts for dominance and quality of opposition therefore the list will not be purely representative of achievements in majors but we would expect to see some correlation between the two.

In the men's game :

the top 5 have won 50 majors in the open era
6-10 won 30
11-15 won 23
16-20 won17


Just thought people might be interested in these statistics which are generated from my own Elo database and are purely evidentially based without interference of human opinion.

I am not recommending ELO as a substitute to the official ranking methodology, but am posting the results purely for information purposes and to provide a different but independent perspective on the evaluation of players.

Take care

Tim

Let's do the list again
I'm curious to see what it looks like now.

 

[Limpinhitter]

Electric Light Orchestra ratings?

 

[BorisBeckerFan]

Comes as no surprise at all that there is so little between the top 5 players on this list and is about what I would have expected from this type of formula. I have to agree with others that as the list goes further and further down there does seem to be a huge breakdown in what one would expect the results to be and what they actually are. Any reconfiguration in the order of the top 5 would have been totally believable to me. Matts and Mac totally make sense and so does Edberg followed by Becker. It's from ten on that all hell brakes loose. This formula has a wicked deviation from what I think most people would have as there top 10-25.