A Simpler USTA Algorithm

I’ve played tennis for years. The USTA is our governing body, and they have a fancy algorithm that they use to rank players based on outcomes of matches, taking the score into play. For instance, if I win 6-2 6-2, I gain more ground than a 7-6 7-6 win. It makes sense on paper, but I think it’s time for a change.

People game the system. People lose on purpose all the time to “keep their ratings” or get bumped down. It’s easy to do because the system punishes bad losses and rewards big wins, especially over strong opponents.

A simple win-loss system, with a similar algorithm could solve most of the problem.

It would be harder to get bumped down. If you get trounced, it counts just the same as a close loss.

There would be very little reason to throw matches. A win is a win. And a loss is a loss. There is nothing in between. So instead of trying to create information that’s not there, we just use what we know for certain. And if we reduce the benefit of throwing matches, we have a better sport. 

The downside is it would probably take more matches to have meaningful ratings.

For example, two players play each other, both with established ratings of 4.25 and 4.05. The 4.25 player is expected to win. If he loses, he loses points, and the 4.05 player gains points. Half of the difference of the two scores would be the points at stake, and because it is a zero sum game, we can only give those back out. So we subtract 0.1 from 4.25 and have two players rated 4.15. And obviously, there is no rating change between players of the same level. And no change if the higher-rated player wins. 

How would mixed doubles work? Keep it as simple as possible. We combine the ratings of the teams. Let’s say for this example that we have 7.8 [3.4+4.4] and 7.6 [4.0+3.6]. The difference of the scores is 0.2, so there is 0.1 of points available. If the better team wins, there is no change. If the worse team wins, the teams become 7.7 points a piece. To get to that number, we must subtract 0.05 points from each player on the 7.8 team, and add 0.05 points to each player on the 7.6-team. So the teams become: 7.7 [3.35+4.35] and 7.7 [4.05+3.65]. There’s not enough information to tell the reason for the win, or to weight anything according to scores. So a win is a win, and each player benefits [or loses] equally.

If we can clean up the algorithm, we can clean up the sport. But as long as the algorithm values aspects of the sport that the players don’t necessarily care about, you open yourself up to tanking. And when you have a system that rewards tanking in sport, it becomes a competition in sandbagging.