The ATP system just doesn't rank

Ted Hopkins
Page 62     ( 621 words )
Saturday, 23 Jan 1999
From section: News Australian open tennis
Publication: Australian Financial Review

An engaging contest is occurring in the cyberspace surrounding the Australian Open tennis tournament. An unheralded Aussie internet player, Darren O'Shaughnessy, is taking on the giants of the game and knocking them over.

O'Shaughnessy is the driving force behind the SRI tennis rankings, which can be accessed at www.rankingsoftware.com. SRI stands for Spectral Radius at Infinity and the spectral radius algorithms applied to calculate the ranking of elite tennis players is walloping the Association of Tennis Professionals (ATP) rankings.

In terms of exposure and prestige, the ATP rankings are unmatched. But how accurate are they at judging genuine quality and competitiveness? In theory, if they are a good measure, then they should provide a reasonable guide for predicting match and tournament outcomes. Thanks to some astute computer modelling by O'Shaughnessy and also by Swinburne University sports researchers, the heat is on the pros.

A key issue with the ATP rankings is they are based on a cumulative points system for winning matches and tournaments. The only adjustment that occurs is where bonus points for prestige events such as the Australian Open are awarded. The reasoning for the bonus and the amount awarded is unclear. Prestige and the mere assumption these big events are tougher to win seems the plausible explanation.

A good example of their shortcomings was the high-standard tussle between Mark Philippoussis and Michael Chang during the week. Over 225 minutes of playing time, the Australian won the match by winning just 12 more points than his American opponent. This suggests that both players deserved close to equal merit for the match. Yet under the ATP system, Philippoussis takes all and Chang gets nothing.

Before the match, the ATP ranked the Australian at No 16 in the world and Chang at 27. The SRI ranked Chang at 17 and Philippoussis at 18, which looked much closer to the money. The pre-Open form of Thomas Enqvist was no surprise to O'Shaughnessy. The SRI had him ranked at No 5 before the tournament, while the ATP ranked him at 21. Clear tournament favourite is Andre Agassi, because SRI ranked him No 1 in the world during the past season. The ATP ranking for Agassi was No 6.

And prior to the Australian Open, the SRI had predicted Pat Rafter and Lindsay Davenport would win the US Open, while the ATP went for Pete Sampras and Martina Hingis. The reason SRI seems the more reliable predictor is that its method used in rankings is adjustive, rather than cumulative.

SRI ranks strong wins higher than weak wins. For example, a 6-0 6-0 win is worth one point, while a tough 6-7 6-3 6-4 win will earn 0.67 points and the losing opponent is given 0.33. Other "adjustment" items include strength of scheduling, court surface and the algorithmic weighing of wins or losses against weak or strong players.

SRI is focused on assessing whether the result was better or worse than expected. If a player at the bottom of the rankings or seedings during the Australian Open takes Agassi to five sets and loses, O'Shaughnessy's method awards the player with extra points. And Agas- si's win would not receive the same number of points as he would if he blitzed his opponent. In tennis cyberspace, O'Shaughnessy is a hobbyist who is a pro at mathematics and modelling and is outpointing the big guns. Ted Hopkins is a director of Champion Data Pty Ltd. More details at www.swin.edu.au/sport


This article reproduced with the permission of the author and publisher.
darren@rankingsoftware.com