The Goat Debate

[mick1303]

And anyway, if we want a numerical rating, Elo is the best, it tells us Djokovic is GOAT

I have a serious objection to applying Elo method to tennis results. Elo was created for chess and naturally treats each game as a separate event and they all have the same weight (which is true for chess tournament/match structure). This is not the case with tennis - the weight of each match is dependent on the tournament tier and the round. First round and the final cannot be treated the same. Likewise 250 series is not the same as Slam.

[ponchi101]

Either one of you please post and explanation of ELO, or a link to an explanation.
I will search it, but would like your guidance.
Edit: Ok. Looked it up. I did not know about it.
I like it. It says nothing (to me) of the tournament difficulty, only the matchups. Only thing I would look into would the "400" variable, and see if it would need to be changed. In tennis, having Novak come to a tournament with a ranking of #1, but with 8,000 points is regular, so the subtraction of 400 would mean little.

[ashkor87]

Nate Silver site fivethirtyeight has an extensive graphic detailing ELO ratings for tennis..Djokovic has the highest rating ever..Serena is a bit lower than Navratilova, and, as the article points out, the reason could be that Serena did not have opponents with high ELO for most of her career.. Sharapova, for instance, is so far below Serena that beating her gave Serena hardly any increase in Elo rating .if only Henin had played longer...this, by the way, is the real weakness of ELO- the big 3 today all have the advantage of having played many times against each other, all very high ELO, so it keeps their Elo high. Serena does not have that advantage.
At some simplistic level, the GOAT debate is about - who would have beaten whom? The ELO answers that directly. It does not weight the importance of the tournament or the round, but then, I am sure, Nadal, Djokovic and Federer don't want to lose to anyone at any round in any tournament, so they probably fit the ELO well! In the past, there have been players, even great players, who did not 'show up' for lesser matches, but I don't think that happens much nowadays.

[ashkor87]

here are the links to that fivethirtyeight discussion

https://fivethirtyeight.com/features/se ... -all-time/
https://fivethirtyeight.com/features/dj ... -all-time/

[ashkor87]

just for reference, peak ELO ratings for current players are like this:
Serena 2508
Azarenka 2326
Barty 2220
Osaka 2200
Kerber 2145

I think it looks fine! a reasonable ordering, dont you think?
source: http://tennisabstract.com/reports/wta_elo_ratings.html

[ashkor87]

the men's would be:
Djokovic 2470
Federer 2382
Nadal 2370
Nishikori 2210 (suprise!)
Zverev 2149
Medvedev 2172
Tsiptsipas 2132

source same as above.. tennisabstract.com

[Liamvalid]

Nate Silver site fivethirtyeight has an extensive graphic detailing ELO ratings for tennis..Djokovic has the highest rating ever..Serena is a bit lower than Navratilova, and, as the article points out, the reason could be that Serena did not have opponents with high ELO for most of her career.. Sharapova, for instance, is so far below Serena that beating her gave Serena hardly any increase in Elo rating .if only Henin had played longer...this, by the way, is the real weakness of ELO- the big 3 today all have the advantage of having played many times against each other, all very high ELO, so it keeps their Elo high. Serena does not have that advantage.
At some simplistic level, the GOAT debate is about - who would have beaten whom? The ELO answers that directly. It does not weight the importance of the tournament or the round, but then, I am sure, Nadal, Djokovic and Federer don't want to lose to anyone at any round in any tournament, so they probably fit the ELO well! In the past, there have been players, even great players, who did not 'show up' for lesser matches, but I don't think that happens much nowadays.

Regarding Serena. I just looked at a WTA ranking for 2002 when Serena became dominant. The top 12 was Venus, Capriati, Henin, Clijsters, Mauresmo, Seles, Hantuchova, Dokic, Hingis, Myskina, Davenport, with the likes of Pierce out injured. Surely Navratilova didn’t have to contend such a strong roster of players when she came to prominence? (Maybe I’m wrong as that was before my time). Maybe it’s recency bias but when comparing eras, I can’t see a stronger time than the late 90s-late 2000s so I don’t think this ELO is accurate in that respect. I’m guessing Navratilova gains all her points from her matches with Evert and Graf?

[ashkor87]

not sure, but the article i posted a link to, has some detail...

[ponchi101]

Martina came in when Court and BJK were starting to fade, Chrissie was clearly at the top, and then the sport was dominated by both of them for about ten years, with brief interruptions by Mandlikova, who simply had the same raw deal as every single ATP player of today (having to go against two monsters for a long time). So depth of competition may be one issue against some players, but that has always been a circular argument to me: if you dominate your era, then you make the era seem weak because nobody else won squat because you dominated the era.

[mick1303]

Nate Silver site fivethirtyeight has an extensive graphic detailing ELO ratings for tennis..Djokovic has the highest rating ever..Serena is a bit lower than Navratilova, and, as the article points out, the reason could be that Serena did not have opponents with high ELO for most of her career.. Sharapova, for instance, is so far below Serena that beating her gave Serena hardly any increase in Elo rating .if only Henin had played longer...this, by the way, is the real weakness of ELO- the big 3 today all have the advantage of having played many times against each other, all very high ELO, so it keeps their Elo high. Serena does not have that advantage.
At some simplistic level, the GOAT debate is about - who would have beaten whom? The ELO answers that directly. It does not weight the importance of the tournament or the round, but then, I am sure, Nadal, Djokovic and Federer don't want to lose to anyone at any round in any tournament, so they probably fit the ELO well! In the past, there have been players, even great players, who did not 'show up' for lesser matches, but I don't think that happens much nowadays.

Edberg lost Roland Garros final to Chang. He never reached other finals there before or since. Clay was stylistically his worst surface. Then several years later he beaten Chang also in Roland Garros in some early round. Everyone remembers their first RG encounter, but hardly anyone remembers the second. Don't tell me these two matches shall have the same weight and round/tournament tier does not matter.

[ponchi101]

How about if that subtraction that is done in ELO gets smaller as you reach final rounds? The formula, in theory, looks fair, so it would be a matter of tweaking it, if somebody were to decide to go this route.

[ashkor87]

Nate Silver site fivethirtyeight has an extensive graphic detailing ELO ratings for tennis..Djokovic has the highest rating ever..Serena is a bit lower than Navratilova, and, as the article points out, the reason could be that Serena did not have opponents with high ELO for most of her career.. Sharapova, for instance, is so far below Serena that beating her gave Serena hardly any increase in Elo rating .if only Henin had played longer...this, by the way, is the real weakness of ELO- the big 3 today all have the advantage of having played many times against each other, all very high ELO, so it keeps their Elo high. Serena does not have that advantage.
At some simplistic level, the GOAT debate is about - who would have beaten whom? The ELO answers that directly. It does not weight the importance of the tournament or the round, but then, I am sure, Nadal, Djokovic and Federer don't want to lose to anyone at any round in any tournament, so they probably fit the ELO well! In the past, there have been players, even great players, who did not 'show up' for lesser matches, but I don't think that happens much nowadays.

Edberg lost Roland Garros final to Chang. He never reached other finals there before or since. Clay was stylistically his worst surface. Then several years later he beaten Chang also in Roland Garros in some early round. Everyone remembers their first RG encounter, but hardly anyone remembers the second. Don't tell me these two matches shall have the same weight and round/tournament tier does not matter.

Maybe, but these things average out over a long enough period/dataset.

[mick1303]

Maybe, but these things average out over a long enough period/dataset.

They do average if you consider the whole pool of players. But there are players who even have a special name given for their psychological disposition - "big match players". Elo-based calculation would diminish their accomplishments.

[ponchi101]

Wouldn't ELO just be one more point when talking about this? I would find it hard to say that one single metric can be used for this conversation.
But we agree on that (I guess).

[ashkor87]

Maybe, but these things average out over a long enough period/dataset.

They do average if you consider the whole pool of players. But there are players who even have a special name given for their psychological disposition - "big match players". Elo-based calculation would diminish their accomplishments.

True ...