Mens Tennis Forums banner

Alternative player ranking system

102K views 1K replies 33 participants last post by  G.100sic 
#1 ·
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.
 
#45 ·
Comparison of Elo ranking to the official ATP race ranking:
Code:
"Race"                       ATP race 
rank   Player                Rank     Diff
  1    Novak Djokovic           1        0
  2    David Ferrer             3        1
  3    Tomas Berdych            6        3
  4    Juan Martin Del Potro    5        1
  5    Rafael Nadal             2       -3
  6    Andy Murray              4       -2
  7    Jo-Wilfried Tsonga       9        2
  8    Roger Federer            7       -1
  9    Richard Gasquet          8       -1
 10    Stanislas Wawrinka      17        7
 11    Gilles Simon            12        1
 12    Kevin Anderson          15        3
 13    Nicolas Almagro         11       -2
 14    Marin Cilic             13       -1
 15    Milos Raonic            13       -2
 16    Denis Istomin           24        8
 17    Bernard Tomic           27       10
 18    Ernests Gulbis          20        2
 19    Julien Benneteau        15       -4
 20    Sam Querrey             17       -3
 
#46 ·
Unfortunately, I have discovered a bug in my database for some time now (nearly a month), which I believed affected some isolated records. But, working on a correction I found out many more erroneous records, which, due to the heavy interdependence of past results to current calculations in Elo system, renders almost the entire ranking incorrect.

I am working intensively on a solution, so hopefully an update after Miami will be error free.
 
#48 · (Edited)
Unfortunately, I have discovered a bug in my database for some time now (nearly a month), which I believed affected some isolated records. But, working on a correction I found out many more erroneous records, which, due to the heavy interdependence of past results to current calculations in Elo system, renders almost the entire ranking incorrect.
...
BTW, errors in the database are corrected. A relatively small number of erroneous records early on in the table propagated through calculations to almost every player's record towards the end. I may post the corrected results for week 11 update later in the evening.

It was interesting to see how larger differences from the beginning even out towards the end of the recalculation during iteration. It is actually the feature of Elo ranking system. For some players though, where error produced the largest differences, there was a significant change in positions.
 
#47 ·
An interesting fact appeared in yesterday's R16 match Berdych vs. Falla in my Elo points calculation, since I take into account every set individually. Due to the difference in points of nearly 200 between the 2, Falla earned 6 points in his defeat, exactly the same as Berdych in victory, for taking 1 set off of Berdych.
 
#54 ·
Here is week 17 update:
http://www.menstennisforums.com/showpost.php?p=12637457&postcount=32

Rafa Nadal have already made it to second position! There is still quite substantial difference to Novak, Nadal is trailing by over 70 points.
Rosol shot up 27 positions, thanks to his win in Bucharest (and congratulations on his first ATP title), to the 23rd place.
Robredo is improving rapidly after the sluggish start of the season, and is now at the 25th place.
So far Cipolla and Mathieu have had absolutely catastrophic start of the season, and are trailing the list. Verdasco not far from the bottom also.
 
#56 ·
New comparison to the official ATP race ranking:
Code:
“Race”                       ATP race    
rank   Player                rank     Diff
  1    Novak Djokovic           1        0
  2    Rafael Nadal             2        0
  3    David Ferrer             4        1
  4    Andy Murray              3       -1
  5    Jo-Wilfried Tsonga       7        2
  6    Stanislas Wawrinka      12        6
  7    Richard Gasquet          8        1
  8    Tomas Berdych            5       -3
  9    Milos Raonic            13        4
 10    Juan Martin Del Potro    6       -4
 11    Nicolas Almagro          9       -2
 12    Roger Federer           10       -2
 13    Tommy Haas              11       -2
 14    Marin Cilic             16        2
 15    Kevin Anderson          17        2
 16    Fabio Fognini           18        2
 17    Ernests Gulbis          27       10
 18    Bernard Tomic           35       17
 19    Sam Querrey             22        3
 20    Jarkko Nieminen         24        4
 
#57 ·
#62 ·
According to their current Elo ranking points, odds for Nadal vs Djokovic (if their match was to be played tomorrow) would be 1.94 : 2.06, respectively.
 
#63 ·
Going into the second week of RG, after first three rounds wins, Nadal slightly increased his lead in Elo points (403.9 versus Djokovic's 391.3), and consequently his odds (Elo expectation) in the eventual semifinal vs Djokovic to 1.930:2.075.

While it seemed that Djokovic had a tougher draw then Nadal, Nadal actually accumulated more Elo points in his first three matches, which means that he had better ranked opponents (by Elo race rankings). Nadal earned 15.3+17.9+17.5 points, while Djokovic had 15.1+16.9+17.3 points.
 
#64 ·
Having said that, it should also be noted that Nadal allowed his opponents more Elo points then Djokovic in the first 3 rounds.
Nadal: 8 points to Brands; 8.7 to Klizan; -5.9 to Fognini; 10.8 total.
Djokovic: -3.3 to Goffin; -3.1 to Pella; -4.9 to Dimitrov; -11.3 total.
 
#65 ·
Elo expectation for the blockbuster SF after today's QF matches is in favor of Nadal, 1.919:2.088.
But Nadal is heading into the semi with more points given to his opponents then Novak (totaling +1.4 vs -18.9)!
Djokovic earned a total of 84.2 Elo points up to the SF, and Nadal earned 88.1 point.
 
#68 ·
Although my policy is to take into account matches played at exhibition tournaments, I have decided not to include results from the Boodles challenge tournament. The reason for that is because matches are not played in the conventional "best of 3" format (there is a super-tiebreak instead of regular 3rd set), and there is no strict tournament structure that determines the overall winner, so my feeling is that it is not sufficiently competitive for the players.
 
Top