How is this UTR scenario possible?

[TheStrenuousLife]

I recently played a match with a guy who is 1.3 UTR points ahead of me. We played a match last week and I won 6-4 6-4. Once it was fed into UTR, I did not change at all, and my rating stayed the same. I checked my opponents, and he went up .36 from his last rating. I am confused as to why this happened. We have both played nearly 50 matches into UTR, and both of us were our last opponents in the last 2 weeks. We are also within the same age range.

Is there a reason he got such a large bump losing to me when he is 1.3 (now 1.6) points higher, yet I stayed the same?

 

[Cashman]

There is a lot more that goes into rating than your last match

The rating period is a moving window and the ratings of your and his prior opponents are dynamic

if your match causes a worse one to drop off his rating or one of your prior opponents has had their rating corrected downwards then it can easily offset any gains you might have achieved from this win

 

[nyta2]

I recently played a match with a guy who is 1.3 UTR points ahead of me. We played a match last week and I won 6-4 6-4. Once it was fed into UTR, I did not change at all, and my rating stayed the same. I checked my opponents, and he went up .36 from his last rating. I am confused as to why this happened. We have both played nearly 50 matches into UTR, and both of us were our last opponents in the last 2 weeks. We are also within the same age range.

Is there a reason he got such a large bump losing to me when he is 1.3 (now 1.6) points higher, yet I stayed the same?

are both 100% reliable?

 

[schmke]

Another UTR conundrum.

I played a doubles match on 1/26 and when it was first recorded in UTR, it had me as a 6.22. Over the next several days, without a new match being recorded for me or any of my partners or direct opponents in the most recent two matches, my UTR went 6.75, 6.68, 6.47, 6.49, 6.25, 6.95, 7.01. I may have had a result fall off the tail end of the year window, and the partner/opponent may have as well, but that is still a lot of variation from no additional matches and just side-effects of the algorithm.

Then, my match played 2/2 got recorded today, but appears to not be calculated yet (no tennis ball in the corner), but my UTR dropped to 6.61. In that new match, my partner (7.01 today) and I beat a 7.74 and 7.43 6-3,6-0. You'd think I'd get a significant bump from that match, I guess we'll see tomorrow when the match gets calculated.

Yes, I and my partner are 100% reliable as are the opponents. I've played over 30 matches in the past 12 months.

The short story is, one's UTR is going to jump around for no obvious reason. Perhaps factoring in how past opponents have done and/or matches fall out of the year window makes it more "accurate", but that "reliable" rating fluctuating over 0.75 in the span of a few days where I didn't play a match is confounding.

 

[TennisOTM]

Another UTR conundrum.

I played a doubles match on 1/26 and when it was first recorded in UTR, it had me as a 6.22. Over the next several days, without a new match being recorded for me or any of my partners or direct opponents in the most recent two matches, my UTR went 6.75, 6.68, 6.47, 6.49, 6.25, 6.95, 7.01. I may have had a result fall off the tail end of the year window, and the partner/opponent may have as well, but that is still a lot of variation from no additional matches and just side-effects of the algorithm.

Then, my match played 2/2 got recorded today, but appears to not be calculated yet (no tennis ball in the corner), but my UTR dropped to 6.61. In that new match, my partner (7.01 today) and I beat a 7.74 and 7.43 6-3,6-0. You'd think I'd get a significant bump from that match, I guess we'll see tomorrow when the match gets calculated.

Yes, I and my partner are 100% reliable as are the opponents. I've played over 30 matches in the past 12 months.

The short story is, one's UTR is going to jump around for no obvious reason. Perhaps factoring in how past opponents have done and/or matches fall out of the year window makes it more "accurate", but that "reliable" rating fluctuating over 0.75 in the span of a few days where I didn't play a match is confounding.

Since you have your own rating system, I'm curious if you might see similar fluctuations in your ratings if you tried to do what UTR does, which is to run the "year-end" version of your ratings every day, using a moving window of the last year of match results. Have you ever tried something like that?

 

[schmke]

Since you have your own rating system, I'm curious if you might see similar fluctuations in your ratings if you tried to do what UTR does, which is to run the "year-end" version of your ratings every day, using a moving window of the last year of match results. Have you ever tried something like that?

Excellent question.

I have done other rating systems for other sports (football, basketball) where I do iterations that factor in the performance of past opponents and let the rating (hopefully) converge, and you do see some interesting results where a win can be counteracted by the results of prior opponents. But this kind of variation over consecutive days with no new results was not something I observed.

 

[schmke]

Another UTR conundrum.

I played a doubles match on 1/26 and when it was first recorded in UTR, it had me as a 6.22. Over the next several days, without a new match being recorded for me or any of my partners or direct opponents in the most recent two matches, my UTR went 6.75, 6.68, 6.47, 6.49, 6.25, 6.95, 7.01. I may have had a result fall off the tail end of the year window, and the partner/opponent may have as well, but that is still a lot of variation from no additional matches and just side-effects of the algorithm.

Then, my match played 2/2 got recorded today, but appears to not be calculated yet (no tennis ball in the corner), but my UTR dropped to 6.61. In that new match, my partner (7.01 today) and I beat a 7.74 and 7.43 6-3,6-0. You'd think I'd get a significant bump from that match, I guess we'll see tomorrow when the match gets calculated.

Yes, I and my partner are 100% reliable as are the opponents. I've played over 30 matches in the past 12 months.

The short story is, one's UTR is going to jump around for no obvious reason. Perhaps factoring in how past opponents have done and/or matches fall out of the year window makes it more "accurate", but that "reliable" rating fluctuating over 0.75 in the span of a few days where I didn't play a match is confounding.

Not that anyone other @travlerajm cares, but today's twist in the "why did my UTR do that?" saga is ...

As noted above, yesterday, without my latest match calculated, I was at 6.61 (having been at 7.01 the day before with no new match added). The added match had me and my 7.01 partner beating a 7.74/7.43 pair by the score 6-3,6-0, a seemingly good result.

One might expect that a nice upset like that might improve my and my partner's rating? Nope, at least for me. I dropped to 6.45 (-0.16) while my partner went up to 7.14 (+0.13) while one opponent dropped to 7.64 (-0.10) and the other to 6.92 (-0.51). Seemingly the "secondary" effects of other matches far outweigh the direct effect of a nice upset win. And those effects are significantly different for different partners.

The gap between our average rating pre-match calculation was 0.775 and post match it is 0.485 so we indeed did get closer, but as a whole the four of us dropped 0.16. Many rating systems are "closed" in that ratings points are transferred between players, e.g. player A wins over B, A goes up X and B goes down X, the points in the system remain constant. If UTR is such a system, other players in the system somehow were given the 0.16 that went missing between players in my match.

 

[Purestriker]

Not that anyone other @travlerajm cares, but today's twist in the "why did my UTR do that?" saga is ...

As noted above, yesterday, without my latest match calculated, I was at 6.61 (having been at 7.01 the day before with no new match added). The added match had me and my 7.01 partner beating a 7.74/7.43 pair by the score 6-3,6-0, a seemingly good result.

One might expect that a nice upset like that might improve my and my partner's rating? Nope, at least for me. I dropped to 6.45 (-0.16) while my partner went up to 7.14 (+0.13) while one opponent dropped to 7.64 (-0.10) and the other to 6.92 (-0.51). Seemingly the "secondary" effects of other matches far outweigh the direct effect of a nice upset win. And those effects are significantly different for different partners.

The gap between our average rating pre-match calculation was 0.775 and post match it is 0.485 so we indeed did get closer, but as a whole the four of us dropped 0.16. Many rating systems are "closed" in that ratings points are transferred between players, e.g. player A wins over B, A goes up X and B goes down X, the points in the system remain constant. If UTR is such a system, other players in the system somehow were given the 0.16 that went missing between players in my match.

The UTR drops after an upset are so mind boggling.

 

[travlerajm]

Many rating systems are "closed" in that ratings points are transferred between players, e.g. player A wins over B, A goes up X and B goes down X, the points in the system remain constant. If UTR is such a system, other players in the system somehow were given the 0.16 that went missing between players in my match.

I’ve mentioned this before:

Juniors are always going up in rating when they become inactive, due to past opponent improvement from biological reasons. For the UTR rating system to maintain stable, someone else in the system (anyone who is a non-junior) has to eat their gain by continually taking some rating decay for the greater good.

 

[TennisOTM]

Excellent question.

I have done other rating systems for other sports (football, basketball) where I do iterations that factor in the performance of past opponents and let the rating (hopefully) converge, and you do see some interesting results where a win can be counteracted by the results of prior opponents. But this kind of variation over consecutive days with no new results was not something I observed.

I'd expect that ratings for those other sports are more stable because all the teams play pretty much the same number of games at regular intervals. Whereas for adult rec tennis you have wide variation, with a sizable portion of players in the network with only a few results or less.

 

[TennisBro]

I recently played a match with a guy who is 1.3 UTR points ahead of me. We played a match last week and I won 6-4 6-4. Once it was fed into UTR, I did not change at all, and my rating stayed the same. I checked my opponents, and he went up .36 from his last rating. I am confused as to why this happened. We have both played nearly 50 matches into UTR, and both of us were our last opponents in the last 2 weeks. We are also within the same age range.

Is there a reason he got such a large bump losing to me when he is 1.3 (now 1.6) points higher, yet I stayed the same?

I've said it before and I'll say it again that I see the UTR of players' ratings higher in some areas while lower in others. Players that are in higher UTR ranking areas more often get the advantage of playing with opponents with higher rankings which ultimately helps improve rankings of others too.

 

[Moon Shooter]

Since you have your own rating system, I'm curious if you might see similar fluctuations in your ratings if you tried to do what UTR does, which is to run the "year-end" version of your ratings every day, using a moving window of the last year of match results. Have you ever tried something like that?

It is not just his results but *everyone's* results that shift every day as matches fall off of the 12 month cliff and new matches are added.

I can't imagine how this is better then using a sort of sliding scale. Where matches are weighted less as time moves on *and* you have played more recent matches. If I know a 40 year old had a 10.45 rating based on 20+ doubles matches 12 months ago I have *some* information about their tennis ability. To completely erase that information must hurt the predictions. Now if you are talking about children then I think they have a better argument. Children can improve quite a bit in 1 year but even there. If a 14 year old had a 10.45 rating after 20+ matches from 12 months ago I still likely have some useful information about their tennis ability.

Sure injuries can change things. But injuries can change things regardless of a 12 month cliff.

 

[Moon Shooter]

I've said it before and I'll say it again that I see the UTR of players' ratings higher in some areas while lower in others. Players that are in higher UTR ranking areas more often get the advantage of playing with opponents with higher rankings which ultimately helps improve rankings of others too.

This should equal out as players move or play people from different areas. But pockets can form not only based on location but based on USTA rating level. There was a while where many of the men's 3.0 singles players in my area had higher singles UTRs then the 3.5 men. Also there can be huge differences with men and women. For a while it was clear that a male player could gain rating points by playing against females.

World tennis number is currently a disaster when comparing men and women. 3.5 women are rated about the same as 4.0 men. (27-29). If you wanted to improve your rating as a male you would just play against women - at least in my area. USTA doesn't have any truly coed events so these pockets tend to last longer then they should.

 

[Max G.]

Now if you are talking about children then I think they have a better argument.

I've always thought that was the root cause of many of the issues with UTR. My guess is they've primarily designed their system to work for juniors, and every time they had to make a tradeoff in system design, they've prioritized the experience of "fast-improving junior" over any other use case. That led them to make a way overengineered system that's way too responsive to recent results, and now they can't make big changes to it without messing up the rating of current juniors so they're stuck trying to expand this brittle mess to adults instead of just designing something more stable.