Considerable skepticism on the accuracy of the World Tennis Number (WTN) is centered around the fact that the new rating system only considers results at the set level. In other words, there is no difference in the result calculations between a 6-0, 6-0 double bagel and a 7-6, 7-6 squeaker.
The Adult recreational-competitive players have been conditioned by experience with the NTRP rating system to believe that the number of games won and lost in a match actually matters. Similar questions were asked during the Intercollegiate Tennis Association(ITA) coaches webinar that was conducted shortly after it was announced that the organization had exclusively adopted the new rating system. It is a safe assumption that at least some in that community also harbor the idea that ratings algorithms must necessarily factor in games for maximum accuracy.
I have recently come around to the idea that the set level is sufficient. The primary reason underpinning my logic can be summarized in two words: John Isner.
Isner plays an inordinate number of sets that end in a tiebreaker. Playing matches that appear to be very close on paper is simply a side-effect of his serve-dominant style of play. For the most part Isner will hold serve and rarely generate a break. That means that his matches inevitably come down to who performs the best in the tie-break game. Beyond who won the set, the score actually doesn’t matter.
In the final post of “My Days as an NTRP Sandbagging Bastard” I disclosed how sometimes I would not use my most effective tactics during USTA League matches when it was clear I was going to win anyway. At the time, I did not understand that the NTRP algorithm was based primarily at the game level of granularity. In retrospect, I realize that treating those matches as developmental opportunities was in effect sandbagging.
In USTA League culture, it is fairly common for enterprising captains to advise their team members to drop a few games. There is a player in my close orbit who received an NTRP “strike” in a playoff loss under dubious circumstances. That match was a lopsided beating until just after the opposing team won enough lines to advance. After that point, the score tightened considerably. I think there is an argument for stopping all matches in progress at the playoff and Championship levels once advancement is locked.
By operating only at the set level, WTN is immune to that perturbation. It ultimately comes down to who wins and who loses. Nothing else matters.
I have embraced the elegant simplicity. There might be benefit in migrating the NTRP algorithm in the same direction.
- ITA x ITF World Tennis Number Coach Webinar, YouTube Unlisted Video, Recorded January 17, 2023.
- The Science Behind ITF World Tennis Number, ITF World Tennis Number News Post, September 5, 2022.
What matters in sports is whether you win the game or match. In North America’s “Big Four” sports—football, basketball, baseball, and hockey—wins and losses (hockey counts ties) determine the daily standings you read in newspapers and online. So it is not uncommon to see these standings differ either a little or a lot with point differentials of teams. For example, in the NBA, Team A with a 42-40 record finishes ahead of Team B with a 40-42 record in the regular season standings. But it’s irrelevant that Team B’s average point differential is a plus 1.6, while Team A’s point differential is a minus 2.1.
In “judged sports” — such as boxing, gymnastics, and figure skating — each round or event is scored, and the score is totaled. In golf, in medal play or stroke play, the total strokes are used to determine the winner and where other players finish.
Any sports fan with even a modicum of sports knowledge wants their team to win, regardless of the score. And they know a loss is a loss, whatever the score.
As a diehard Boston Red Sox fan, I’m thrilled when the Red Sox win a four-game series against the New York Yankees. If the Red Sox win the first three games by scores of 3-2, 4-3, and 5-4 and then lose the last game 10-0, that overall result is wonderful. The fact that the Yankees scored seven more total runs in the four-game series is irrelevant.
The same reasoning holds true in tennis. If Teresa beats Linda 6-4, 4-6, 7-6 and 6-7, 7-5, 7-5, and Linda beats Teresa 6-0, 6-1, Teresa is clearly the better player in their head-to-head rivalry because of her 2-1 match edge. It’s irrelevant that Linda has won more total games and total points.
One of the reasons the UTR rankings and the NTRP rating system are seriously flawed — as I explain in an essay in my coming book, “Tennis Confidential III” — is that the scores of matches are factored in the rankings. A case in point is my own experience in USTA senior tennis tournaments during the past 15 years. Because I often defeated opponents very decisively when I won while many players about my level often did not, my UTR ranking was often — wrongly — either a little or a lot higher than their rankings. Furthermore, I occasionally was ranked above players whose record was clearly superior to my record. That was also wrong and unfair.
Rankings matter greatly in their own right — both to tennis players and fans. But they also matter greatly because they are used for seedings. Wrong and unfair rankings, therefore, result in wrong and unfair seedings. The consequences of both mistakes create something of a vicious cycle that — wrongly — benefits some players and hurts other players. That is bad for tennis.
First, going by the number of headlines and participation estimates since the pandemic, it’s clear pickleball has usurped hockey as the fourth biggest sport in North America. If we believe the headlines, we’ll soon see indoor hockey rinks converted to pickleball facilities and the original six NHL teams splinter off to form their own professional pickleball league to get a bigger TV contract. Don’t hate the players…
Second, you should have written, “*Almost* any sports fan… wants their team to win, … “. Ask Brian Flores how his erstwhile team owner viewed winning individual games in a lost season when “earning” the #1 overall draft pick was possible. Houston Texans fans may have felt conflicting emotions when their team came from behind to win the final game of last season and “lose” the top pick. File this under the heading of perverse incentives.
Third, as a lifelong Red Sox fan I plan to spend the upcoming season rewatching my DVD copy of “Faith Rewarded, the Historic Season of the 2004 Boston Red Sox” and looking ahead to a more productive off-season by management.
Fourth (and finally, the real reason for this post!), given the variations of tennis scoring (three full sets or two full sets and a super tie-break or two fast-four sets and a regular tie-break or pro sets, etc.) it’s clear why the WTN had to ignore set scores and simplify its calculations to sets won or lost to match the Glicko-2 methodology.
As for using set results vs the overall outcome (W or L), as I understand the WTN from Teresa’s detailed blogs, in your example above, if Linda and Teresa played each other exclusively, the algorithm would have predicted Teresa would win in three sets prior to the match. As a result of Linda winning the third match, Linda’s WTN would increase and Teresa’s would decrease significantly as a straight set win by Linda was the largest deviation from the expected outcome. A smaller change would have registered had Linda won in three sets. No change would have registered had Teresa won in three sets and there would be a small WTN change in the opposite direction for both players had Teresa won in straight sets.
In a larger population of matches vs other players, this also means a win in three sets vs a player with a much lower WTN may result in the winning player’s WTN dropping and the losing player’s WTN increasing because the algorithm would have predicted a straight set win by the higher rated player. If this is what happens then it still may present an opportunity for tanking to maintain a lower rating should the WTN eventually replace the NTRP ratings. Six of one…
As a corollary, I’ve played in a flex league for 20 years that uses match results to assign points for playoff eligibility (straight set wins are awarded more points than splitting sets) and also uses the differential in games each match to calculate each players’ rating for end of season promotions and demotions to higher or lower levels of play. This happens three season per year and is a dynamic scoring and rating model that makes every game count if you want to make the playoffs and/or advance to a higher level of play or avoid demotion to a lower level of play. In my years of playing tennis this has been the best rating model I’ve experienced but it largely works because most participants want to get promoted rather than sandbag.
I wouldn’t base an argument on an outlier as extreme as John Isner. Makes it easy to say he’s the exception that proves the rule, games matter. To discount games means tanking sets becomes a more viable strategy becuase it won’t hurt the tanker’s WTN. Tanking is bad for tennis and should be discouraged.
There was the argument that tournament directors should be able to determine (and market) how the results generated by their events will be processed by ratings algorithms; either binary win/loss, sets, or games; precisely because of the effects of each method on mental approach to matches. I wanted this to be part of the data standard (a processing code). We have processing codes in TODS! But I do not believe that this idea will ever catch on. One of the best juniors I watched developing, prior to UTR, would always have close matches, regardless of the level of his opponent, because he was always experimenting/learning. He ended up winning Wimbledon juniors. Too many players/parents have an unhealthy relationship with ratings (perverse incentives) and are unable to “play” at tennis this way.