Discussion in 'Former Pro Player Talk' started by Wuornos, Apr 9, 2008.
Please see http://tt.tennis-warehouse.com/showthread.php?t=244754.
hi, quick question : does the ELO ratings include Federation Cup and Davis Cup results?
Unfortunately not 1970. No. Most seniour tournaments, majors, EYC etc.
It will surprise some to see Lendl rated this highly, but it shouldn't. In fact, he was right up there with Laver and was objectively great on all surfaces. If we look the stretch from somewhere around summer '85 to the end of '87, it's truly a stretch to rival any in the open era. ELO reflects his dominance.
Tim - is there a basic way to explain McEnroe being below Wilander on the peak ELO list? This is an interesting surprise, considering what we know about Mac at his best in 1984. Thoughts?
Looks like Wilander's 1988 may add up to a higher peak. Wilander was the man to beat that year, and his three slams + QF at Wimbledon do seem better than Mac's 2 + one final.
I remember the 1987 Master's final between Lendl and Wilander. An extraordinary match in terms of low errors on both sides. Lendl won it, but he had to play perfect; it was like he had to pull every point out of Wilander with pliers. An exercise in long-term concentration on both sides. Their 1988 US Open final was like that too. It was sometime in 1988 that I heard Cliff Drysdale make a comment that stuck with me about Wilander's game. He said something like: "I don't think Wilander can have a long career. I see the pain he has to go through to win these matches." At the end of that 1987 master's final, during the interview, Lendl mentioned that he felt *mentally* exhausted after his win. That's what beating Wilander did to others, and ultimately that's what winning matches did to him too.
But if you analize his 1988 it is likely that the quality of his wins exceeds Mac's 1984; beating Edberg and Cash at the Australian, winning the French, and then his win over Lendl at the US Open final.
Then, just as Drysdale had predicted, he faded away rather quickly. But in 1988 he was way up there.
Yes I can explain why ELO puts Wilander above McEnroe objectively.
It's really down to multiple reasons which when amalgamated together add up to the higher peak rating for Mats.
Firstly it is a mistake to think of ELO ratings covering a single year. Yes matches played in the latest year will have a greater weight given to them in relation to the current rating but earlier periods will still contribute, with a diminishing effect with each subsequent match played. McEnroe did not have a great year prior to 1984 and therefore the albeit reduced significance of these results effected the peak rating he achieved.
Secondly, and as Benhur quite rightly points out, the quality of opposition faced by Wilander was higher during his peak, or at least had greater depth, than that faced by McEnroe. In such a situation ELO adjusts the ratings accordingly meaning that it was not necessary for Wilander to dominate the elite group by as much to achieve an equilvalent or superior rating as domination was more dfficult at that time.
In summary then Wilander just edges McEnroe at their respective peaks in the eyes of ELO due to greater consistency over multiple years and facing a higher quality of opposition during those years, including a close to peak Ivan Lendl.
Finally I would point out that the difference quoted between these two greats of the game is negligible at 23 ELO points. In fact ELO would predict the result of a match between them both at their peak as being closse to a coin toss at 55/45.
I hope this helps your understanding of how this unexpected result came about and gives some indication of the depth of analysis carried out by ELO as a matter of standard process.
Take care Cyborg and thanks for the interest in my work.
Thanxs Tim again!
Interesting stuff and it's hard to ignore that Mac's competition in 1984 was indeed poorer than Wilander's in 1988. A younger Lendl. An older Connors. Beat an aging Gerulaitis in the Canada final. A lot of opponents a bit too old and a lot of them a bit too young.
One has to admit though that Wilander was a bit fortunate that Lendl has his injury problems in 1988. He took advantage and ELO doesn't account for this, but no system can.
Electric Light Orchestra ratings? :]
Good to see the stats are back. Im interested to know which tournaments you used to determine the peak period for some of the players, ie did you use Laver's results in 1969 (his Grand Slam year), Connors (1974), Wilander (1988) etc when you did the mens results to work out the rankings? In 1969 I have Laver playing 32 tournaments, winning 18 and runner-up in 5.
Also interested to know if you considered the strength of the WCT Finals during the early seventies and the Masters final, as I believe they were stronger in the seventies and early eighties than most of the Australian Championships of that period.
If you watch the '87 USO final and compare it to the '88 final it doesn't appear that Lendl played any different but Wilander did. So I wouldn't place Lendl's lost due to injury issues.
Interesting that you brought up the '87 Masters final. I was watching it again the other day and was wowed at how great Lendl played. I remember after the match Wilander was quoted in saying that Lendl was at the top of his game and couldn't improve anymore while the rest of the major players (Edberg, Becker, etc..) could still improved. It turned out to be true cause '87 was the last year Lendl truly dominated.
Lendl played significantly less in 1988 than usual. After the US Open he had surgery on his right shoulder. It was definitely affected, but he still played very well.
Yes, and if I had to choose one match as representative of Lendl's tennis at his best, I think I would choose that 87 master's final against near-peak Wilander. A lesson in controlled and precise agression from the back against one of the best defenders in tennis. I would not be surprised if Lendl's peak ELO was achieved at that tournament.
Wilander was near his peak (he went on to win the Australian a few weeks later) and even though he was beat soundly by Lendl (2, 2 and 3), he never really played poorly or below his ability, which helps assess the quality of Lendl's tennis that day. There is a well known effect whereby when one player is playing extremely well, it eventually causes the opponent to start playing below his abilities. Nadal mentioned this a few days ago, when talking on how difficult he often finds it to play Blake: "...not because of me but instead because of James. He puts unbelievable pressure... it is more a mental thing than physical. I always say that normally (not always) when a player plays great it is difficult to play very well. It is like a balance. If one is too good, the other can't be that good."
That's true and it sometimes makes it difficult to assess the overall quality of the tennis played by the winner.
In this regard, Wilander may well be one of the tennis players most immune to this effect in the history of the game, almost always able to maintain a good and even level of play independently of how great the other guy was playing. He was the ultimate "you have to beat me; I won't help you at all" kind of player. And I think this was quite true in this particular match.
When I saw Sergi Bruguera with an ELO of 2607 and ranked 46 on the first list, I started to wonder where Thomas Muster was? He didn't make the 2600 list.
yet, Muster in the quarter final or better is listed at 69, Bruguera is listed at 41 on that list.
however, their head to head Muster dominates winning 12-3 (from atp site), and 4 of those 12 wins were finals, and 4 of those 12 wins were semi-finals.
so, it seemed a little odd to me.
Anyone have any thoughts?
Well I guess I can take a shot. ELO does not calculate for the bad matchups. I think this is what it is, Muster being a very bad matchup for Bruguera, but when you look at overall picture Brugera earned a better ELO ranking.
All data and ratings now up to and including Australian Open.
K Factor amended following comparison exercise which showe a higher value was required to make current ELO values act as better predictor.
Because of the smaller volume of data prior to the open era a greater K factor is used.
Methodology now updated so K factor is self adjusting depending on volume of results now available.
The change in K Factor will have some effect in the overall ranking of player and a more significant effect upon the actual rating achieved with greater weight being given to more recent results and less to past results although this greater voatility would still be less than that seen in the official single rankings.
The change has been made because I have now had sufficient opprotuniy to make an objective analysis of the results of players compared with their current elo ratings. The K factor now calculates its optimum level, thereby improving the accuracy of the ELO ratings calculated in relation to tennis.
Please let me know if you spot any errors.
Wow, interesting stuff. What I can't understand is why I can't find myself on that list. Maybe you overlooked something?
This is because Federer and Nadal have been consistently avoiding you. It's difficult to get a high ELO rating without playing the top players
I don't think we can any longer refer to these as elo ratings as they there have been so many amendments to methodology and improvements over the past year and a half that the formuale used are significaly different from those first developed by Prof Arpad ELO.
E.g. The built in decay due to inactivity.
I'm not going to go into all the differences as in the past this has often led to discussions relating to rating methodology rather than the tennis players themselves.
I shall revert to referring to these ratings as DOT ratings as they are still calculated in their most basic form from Dominance, Standard of Opposition and the quality of Tournament.
Happy to answer any questions regarding the ratings of individual players progression of ratings. Rating comparisons between players over time etc.
Deleted post as contained error in data
Tim is back! The legend is here!
Welcome back my friend!
Separate names with a comma.