Opinion: UTR algo should care about winning

[Diablo_Lover]

My understanding of the UTR algorithm is that it only cares about games won and games lost. More specifically it just cares about the % of games won. I believe it also care about winning matches.

Example: Bob and Carl have an identical UTR rating and play a match, with Bob winning 6-4, 2-6, 6-4. Both players won and lost 14 games, so the UTR algo will treat them as having "tied' for algo purposes and will not adjust either player's rating in either direction. However, I believe that the fact that Bob won gives some evidence that he is nonetheless the better player and I think his rating should bump.

I'm not saying that the winning player is always better. E.g. if Bob won 7-6 (5), 0-6, 7-6 (5) then maybe we should think Carl is better. But winning should count for something in my view, perhaps a 1 game "bonus" per set played?

Thoughts?

 

[travlerajm]

My understanding of the UTR algorithm is that it only cares about games won and games lost. More specifically it just cares about the % of games won. I believe it also care about winning matches.

Example: Bob and Carl have an identical UTR rating and play a match, with Bob winning 6-4, 2-6, 6-4. Both players won and lost 14 games, so the UTR algo will treat them as having "tied' for algo purposes and will not adjust either player's rating in either direction. However, I believe that the fact that Bob won gives some evidence that he is nonetheless the better player and I think his rating should bump.

I'm not saying that the winning player is always better. E.g. if Bob won 7-6 (5), 0-6, 7-6 (5) then maybe we should think Carl is better. But winning should count for something in my view, perhaps a 1 game "bonus" per set played?

Thoughts?

You are suggesting for UTR to work more like WTN.

 

[johnmccabe]

Tennis doesn't care about my opinions and my feelings

 

[ProStaffFanatic]

My understanding of the UTR algorithm is that it only cares about games won and games lost. More specifically it just cares about the % of games won. I believe it also care about winning matches.

Example: Bob and Carl have an identical UTR rating and play a match, with Bob winning 6-4, 2-6, 6-4. Both players won and lost 14 games, so the UTR algo will treat them as having "tied' for algo purposes and will not adjust either player's rating in either direction. However, I believe that the fact that Bob won gives some evidence that he is nonetheless the better player and I think his rating should bump.

I'm not saying that the winning player is always better. E.g. if Bob won 7-6 (5), 0-6, 7-6 (5) then maybe we should think Carl is better. But winning should count for something in my view, perhaps a 1 game "bonus" per set played?

Thoughts?

Wow, didn't know that it worked like that.

Over here in Belgium, the ranking system cares not at all about games / sets won. It only cares about victories.
Winning a tournament where every match was won 7-6 6-7 7-6 will yield the exact same points as when every match was won 6-0 6-0.

 

[ChaelAZ]

Math and stats don't work that way, and UTR specifically doesn't because it is an indicator of your competitive level, not a win/loss ratio. And for the exact reason you you end with, that a winning player isn't always better, so you should never reward someone with that.

Unless you beat Alcaraz or Sinner. Then I will give you a bonus point.

 

[ptrbwmf2026]

Math and stats don't work that way, and UTR specifically doesn't because it is an indicator of your competitive level, not a win/loss ratio. And for the exact reason you you end with, that a winning player isn't always better, so you should never reward someone with that.

Unless you beat Alcaraz or Sinner. Then I will give you a bonus point.

I used UTR to put together a team that made the playoffs in 6.0 mixed.

 

[Moon Shooter]

If your method creates a more accurate rating then I’m in favor of it. If not then I would rather go with the more accurate rating system.

 

[TennisOTM]

I do think that using both the match score (who won) AND the total game score could potentially work better than using only the total game score. It might actually help deal with those who punt games on purpose when they know they can still win the match. But only rigorous testing would confirm if it actually works better in practice. I suspect that if it did improve things, it would be just a minor improvement.

 

[GrassCourtFan]

This feels like very much a "rearrange deck chairs on the titanic" kind of issue.

One of the problems with UTR is that the actual algorithm is opaque, we don't really know what it even is! And to top off the fact that it's opaque, it also (based on its instability) seems to be overly complex. Adding an extra bonus for a win is adding more complexity to something I'm guessing is already overengineered. (The USTA alg, though it's also opaque, seems simple enough that I'm pretty confident I know the gist of how it works even if I don't know the actual math.)

UTR also has the issue that it's unstable - likely because they try to use historical recalculations, a player's rating can drift unpredictably even when they have no change to their matches. Extra bonuses for wins don't address that.

 

[nyta2]

You are suggesting for UTR to work more like WTN.

is that how wtn works? (weighting a "win" slightly higher than the raw games analysis?)

 

[schmke]

is that how wtn works? (weighting a "win" slightly higher than the raw games analysis?)

WTN seems to not look at games at all, instead the finest their granularity goes is sets. So in most matches, it is 2 sets to 0 or 2 sets to 1.