[G.100sic]

Hello everybody

I am interested in creating and maintaining an alternative tennis players ranking list, based on Arpad Elo rating system (http://en.wikipedia.org/wiki/Elo_rating_system), which is used in chess players rating, and which is suitable for comparison of players from different epochs.

[Henry Chinaski]

well good luck with that

[Purple Rainbow]

Great idea. I've toyed with the idea too, but it's too much work.

Please post the results when you're done!

[G.100sic]

There were some attempts before, but that was only an one time effort...
http://tennisplanet.wordpress.com/20...afafan-thanks/

At that time, ranking of players was (top 10, at least):
1 Roger,Federer,SUI 2007.7 2890.2
2 John,McEnroe,USA 1985.3 2835.6
3 Ivan,Lendl,USA 1990.1 2814.3
4 Rafael,Nadal,ESP 2009.4 2799.6
5 Bjorn,Borg,SWE 1980.5 2771.4
6 Jimmy,Connors,USA 1980.1 2665.1
7 Novak,Djokovic,SRB 2009.9 2619.4
8 Pete,Sampras,USA 1997.2 2609.6
9 Andy,Murray,GBR 2010.1 2584.2
10 Rod,Laver,AUS 1975.2 2567.8
...

The first number is year.month when the peak points were achieved, the second is the actual peak rating points.
Compared to chess players:
1. Kasparov, Gary……………… g RUS 2851 2851
2. Anand, Viswanathan………….. g IND 2771 2769
3. Kramnik, Vladimir…………… g RUS 2760 2758
4. Shirov, Alexei……………… g ESP 2734 2751
5. Morozevich, Alexander……….. g RUS 2758 2748
6. Leko, Peter………………… g HUN 2701 2725
7. Adams, Michael……………… g ENG 2708 2715
8. Ivanchuk, Vassily…………… g UKR 2702 2709
9. Bareev, Evgeny……………… g RUS 2698 2709
10. Topalov, Veselin……………. g BUL 2690 2702

[G.100sic]

Originally Posted by Purple Rainbow

Great idea. I've toyed with the idea too, but it's too much work.

Please post the results when you're done!

Well, calculating points is not to much work, but maintaining the whole database could prove to be demanding.
In the meantime, I would like to discuss some technical/mathematical issues regarding calculations, as well as some practical matter.
For instance, maintaining "career" ranking list, 52-week list, and even year-to-day (ATP/WTA race to championship) list?

[Litotes]

Originally Posted by G.100sic

The first number is year.month when the peak points were achieved, the second is the actual peak rating points.
Compared to chess players:
1. Kasparov, Gary……………… g RUS 2851 2851
2. Anand, Viswanathan………….. g IND 2771 2769
3. Kramnik, Vladimir…………… g RUS 2760 2758
4. Shirov, Alexei……………… g ESP 2734 2751
5. Morozevich, Alexander……….. g RUS 2758 2748
6. Leko, Peter………………… g HUN 2701 2725
7. Adams, Michael……………… g ENG 2708 2715
8. Ivanchuk, Vassily…………… g UKR 2702 2709
9. Bareev, Evgeny……………… g RUS 2698 2709
10. Topalov, Veselin……………. g BUL 2690 2702

Great idea, really, but what kind of list is the one above? They are not present numbers (Magnus Carlsen tops with 2843) and they are not historical highs either, except for Kasparov. Anand and Kramnik has both been above 2800. Ivanchuk has been over 2780, at least.

This is the present live chess list:
http://www.2700chess.com/
and the last officially publicized one:
http://ratings.fide.com/top.phtml?list=men

[Litotes]

Originally Posted by G.100sic

Well, calculating points is not to much work, but maintaining the whole database could prove to be demanding.
In the meantime, I would like to discuss some technical/mathematical issues regarding calculations, as well as some practical matter.
For instance, maintaining "career" ranking list, 52-week list, and even year-to-day (ATP/WTA race to championship) list?

52-week list/year-to-day list would be something else. Nothing like that exists in chess. A top player can go an entire year without playing and will keep his number. How would you go about a 52-week list? I might be unimaginative, but I can't see how you would do anything besides "career" list this way. Unless you have all players start over - like everyone ranked 1500 at the start of the year. In which case playing lots of lowly tournaments would count unproportionally.

What looks more achievable IMO, is to post yearly movements. If player x started the year at 2800 and is now 2840, it has been a good year. But it will not be immediately apparent if it has been better year than for player y moving from 2600 to 2750.

[G.100sic]

Originally Posted by Litotes

Great idea, really, but what kind of list is the one above? They are not present numbers (Magnus Carlsen tops with 2843) and they are not historical highs either, except for Kasparov. Anand and Kramnik has both been above 2800. Ivanchuk has been over 2780, at least.

This is the present live chess list:
http://www.2700chess.com/
and the last officially publicized one:
http://ratings.fide.com/top.phtml?list=men

That list was just copied from the link I referenced (http://tennisplanet.wordpress.com/20...afafan-thanks/), as an example for comparison. Thanks for your input.

[G.100sic]

Originally Posted by Litotes

52-week list/year-to-day list would be something else. Nothing like that exists in chess. A top player can go an entire year without playing and will keep his number. How would you go about a 52-week list? I might be unimaginative, but I can't see how you would do anything besides "career" list this way. Unless you have all players start over - like everyone ranked 1500 at the start of the year. In which case playing lots of lowly tournaments would count unproportionally.

What looks more achievable IMO, is to post yearly movements. If player x started the year at 2800 and is now 2840, it has been a good year. But it will not be immediately apparent if it has been better year than for player y moving from 2600 to 2750.

52-week list/year-to-day list would reflect not what exists in chess, but what exists in tennis. Official 52-week list determines actual ATP/WTA ranking, which determines seeding for the tournaments. The Elo list would reflect just the same, results from matches a player played in the last 52 weeks, and whatever the numbers show from that, just with any statistics, is fine. The same goes for year-to-day list.

For me, it would be interesting to compare those lists with official ATP/WTA ranking, because official ranking favors higher prized tournaments. If we imagine the situation, for instance, that one player wins 2 tournaments, one from 500 series, and one from 250 series, playing against exactly the same opponents on both tournaments, in exactly the same number of sets, he would be awarded double points for the 500 series tournament for exactly equal accomplishment as on 250 series tournament. Elo rating system irons those differences, taking into account only head-to-head achievements.

[Litotes]

Originally Posted by G.100sic

52-week list/year-to-day list would reflect not what exists in chess, but what exists in tennis. Official 52-week list determines actual ATP/WTA ranking, which determines seeding for the tournaments. The Elo list would reflect just the same, results from matches a player played in the last 52 weeks, and whatever the numbers show from that, just with any statistics, is fine. The same goes for year-to-day list.

For me, it would be interesting to compare those lists with official ATP/WTA ranking, because official ranking favors higher prized tournaments. If we imagine the situation, for instance, that one player wins 2 tournaments, one from 500 series, and one from 250 series, playing against exactly the same opponents on both tournaments, in exactly the same number of sets, he would be awarded double points for the 500 series tournament for exactly equal achievement as on 250 series tournament. Elo rating system irons those differences, taking into account only head-to-head achievements.

But what would the starting points be? Imagine you're at Doha 2013. Year just started. You're calculating the "race" ranking. Federer wins 1st round against....say Tomic. What is their ranking before and after this one match? Something like 1500 --> 1520 for Federer and 1500 --> 1480 for Tomic?

[G.100sic]

Originally Posted by Litotes

But what would the starting points be? Imagine you're at Doha 2013. Year just started. You're calculating the "race" ranking. Federer wins 1st round against....say Tomic. What is their ranking before and after this one match? Something like 1500 --> 1520 for Federer and 1500 --> 1480 for Tomic?

Well, the floor in Elo rating system is actually arbitrary, and most commonly it is set to 100 or 400. So, both Federer and Tomic have the same, say 400 points, to start with. In that respect it is not very different from ATP race list, where everybody starts with 0 points. So, if Federer beats Tomic, he would then have 412 points, and Tomic would fall to 388 points. But, with ATP point system it is not the same if Federer beat Tomic in final or in R32 match, whereas in Elo system it is not important in which phase of the tournament a player wins.

Interestingly, if the floor was set to 2000 instead to 400, the difference between the 2 players after the match would stay the same +12 points, and -12 points - 2012 and 1988, respectively.

[Litotes]

Originally Posted by G.100sic

Well, the floor in Elo rating system is actually arbitrary, and most commonly it is set to 100 or 400. So, both Federer and Tomic have the same, say 400 points, to start with. In that respect it is not very different from ATP race list, where everybody starts with 0 points. So, if Federer beats Tomic, he would then have 412 points, and Tomic would fall to 388 points. But, with ATP point system it is not the same if Federer beat Tomic in final or in R32 match, whereas in Elo system it is not important in which phase of the tournament a player wins.

You're thinkin 12 as quotient, I see. This could conceivably be higher for important tournaments and lower for unimportant ones.

It will take a long time for rankings to show a true picture if a victory over Federer initially is worth no more than a victory over a 16-year old WC in his first ATP match.

[G.100sic]

Originally Posted by Litotes

You're thinkin 12 as quotient, I see. This could conceivably be higher for important tournaments and lower for unimportant ones.

It will take a long time for rankings to show a true picture if a victory over Federer initially is worth no more than a victory over a 16-year old WC in his first ATP match.

Not really, (+/-)12 is not the quotient here, but just the result of the calculation, and it reflects the difference in player's "strength". In the case that both players have the same rating (points), it is 12, but say Federer had 2890 points, and Tomic had 2000, then if Federer beat Tomic, he would gain 24 points, and Tomic would loose the same. But if Tomic beat Federer, then they would both gain 0 points, that is, they would both retain their previous points. And if we imagine that there could be a draw in tennis match, then Federer would still gain 12 points from the draw, and Tomic would loose 12 points. That is simply how calculations in Elo system work. And the point is not to rate tournaments as important or unimportant, the point is that wining is awarded the same way on any tournament (between the same players).

The difference in points reflects probability that one player would win over the other, something like determining the betting odds. If a player with higher ranking wins lower ranking player, firstly, it is actually expected, and secondly, by wining, that player increases his chances of winning again, and that is reflected on increased difference in points. Rating system is in a way cumulative.

What I would like to discuss, is how to grade a result of a match.

In chess, there are 2 outcomes, win/loose and draw, and win is awarded 1, draw 0.5, and loosing 0. In Elo rating calculations, result should map to a range between 0 and 1. In tennis, there can be only win/loose, so simple way would be to award 1 for win, and 0 for loosing. But, I am thinking of a finer grading, that could reflect more subtle differences between player's performances, and also establish the difference between matches on grand slams and matches on other tournaments (winning in 3 sets vs. winning in 2 sets).

[Litotes]

Originally Posted by G.100sic

Not really, (+/-)12 is not the quotient here, but just the result of the calculation, and it reflects the difference in player's "strength". In the case that both players have the same rating (points), it is 12, but say Federer had 2890 points, and Tomic had 2000, then if Federer beat Tomic, he would gain 24 points, and Tomic would loose the same. But if Tomic beat Federer, then they would both gain 0 points, that is, they would both retain their previous points. Thats how calculations in Elo system work. And the point is not to rate tournaments as important or unimportant, the point is that wining is awarded the same way on any tournament (between the same players).

The difference in points reflects probability that one player would win over the other, something like determining the betting odds. Rating system is in a way cumulative.

What I would like to discuss, is how to grade a result of a match.

In chess, there are 2 outcomes, win/loose and draw, and win is awarded 1, draw 0.5, and loosing 0. In Elo rating calculations, result should map to a range between 0 and 1. In tennis, there can be only win/loose, so simple way would be to award 1 for win, and 0 for loosing. But, I am thinking of a finer grading, that could reflect more subtle differences between player's performances, and also establish the difference between matches on grand slams and matches on other tournaments (winning in 3 sets vs. winning in 2 sets).

No it isn't. Take it from me - I have played competitive chess and have an international Elo-ranking myself. If two players ranked 1500 meet and the winner gains 12 points, then we have the following scenarios for players ranked 2890 and 2000. If 2890 wins - plus one point, he goes to 2891 (not precisely one, I rounded off). If 2000 wins, plus 23 points, he goes to 2023. As you see, winning over a higher ranked opponent is worth much more than winning over someone lower, or your own class. But winning over a lower ranked player is never completely worthless.

In theory, there is no problem calculating more results than two or three. But I don't see what winning in three sets should be worse than winning in two. In chess, there is no bonus for mating the opponent in eight moves. You get the same score if you force him into submission only after 100 moves.

[G.100sic]

Originally Posted by Litotes

No it isn't. Take it from me - I have played competitive chess and have an international Elo-ranking myself. If two players ranked 1500 meet and the winner gains 12 points, then we have the following scenarios for players ranked 2890 and 2000. If 2890 wins - plus one point, he goes to 2891 (not precisely one, I rounded off). If 2000 wins, plus 23 points, he goes to 2023. As you see, winning over a higher ranked opponent is worth much more than winning over someone lower, or your own class. But winning over a lower ranked player is never completely worthless.

In theory, there is no problem calculating more results than two or three. But I don't see what winning in three sets should be worse than winning in two. In chess, there is no bonus for mating the opponent in eight moves. You get the same score if you force him into submission only after 100 moves.

Hmmmm... I may have misinterpreted the formulas. Thanks for pointing that out, I will look into it!

As for the grading of the result... In tennis, winning sets determines the result of a match, and that is also recorded in statistics, so it makes sense to include that in calculations. What my proposal is - player who wins 3 sets is awarded 1; player who wins 2 sets is awarded 2/3, or 0.666...; player who wins 1 set is awarded 1/3, or 0.333...; player who wins no sets gets 0. And that is regardless of winning a match. So, a 3:2 win would map to 1:0.666..., a 2:1 win would map to 0.666...:0.333..., 3:1 win would map to 1:0.333..., 3:0 win maps to 1:0, and 2:0 win maps to 0.666...:0. That way, a player would be awarded more points if he/she chunks more sets off of the opponent. And also, 3:0 win is worth more than 2:0, as a consequence winning a match on a Grand Slam or Davis Cup is worth more than a win on ATP/WTA series tournaments.