Dynamic rating goes down after 6-0, 6-0 win?

[rallyfries]

So I am a 3.5C and had an estimated dynamic rating of 3.51 on TR. I recently played a league match against a 3.07 and won 6-0, 6-0 and my dynamic rating went down to 3.49 and my opponent's rating goes up to 3.09. How is that possible? At worst I would expect it to stay the same. I think my UTR also went down as well.

 

[denoted]

They get your point count from local league officials. Too many deuce games.

 

[travlerajm]

So I am a 3.5C and had an estimated dynamic rating of 3.51 on TR. I recently played a league match against a 3.07 and won 6-0, 6-0 and my dynamic rating went down to 3.49 and my opponent's rating goes up to 3.09. How is that possible? At worst I would expect it to stay the same. I think my UTR also went down as well.

Just a guess. TR averages your last X number of results. If playing that match caused a really good result X matches ago to fall out of the rolling average, then your rating could go down.

 

[OnTheLine]

Just a guess. TR averages your last X number of results. If playing that match caused a really good result X matches ago to fall out of the rolling average, then your rating could go down.

What he said.

Same with UTR ... sometimes after a drop or a rise you can see that a particularly good win or miserable loss just dropped off.

 

[rallyfries]

What he said.

Same with UTR ... sometimes after a drop or a rise you can see that a particularly good win or miserable loss just dropped off.

Hmm...according to UTR algorithm it takes the last 30 matches in the last 12 months but I've only played 15 matches this year and am 15-0. My UTR dropped from 6.20 to 6.14 after the double bagel.

 

[FlamingCheeto]

Don't worry, Swiatek just did the same to what's he name in the final of Rome

 

[J_R_B]

So I am a 3.5C and had an estimated dynamic rating of 3.51 on TR. I recently played a league match against a 3.07 and won 6-0, 6-0 and my dynamic rating went down to 3.49 and my opponent's rating goes up to 3.09. How is that possible? At worst I would expect it to stay the same. I think my UTR also went down as well.

1. TR is wrong, so it's likely their calculation is bad.
2. for a 6-0 6-0 match, you still have to assume that indicates a certain difference in level. It looks like TR might assume 0.4, which is nearly a full level. If you're starting with a larger differential, that match might move you closer together. This is a weakness in the calculation and a reason they threw out 6-0 6-0 results for years. The differential for this match score is really a minimum and it's impossible to tell what the real differential is. If you played Djokovic, the score would be 6-0 6-0, but your NTRP level difference for that match is certainly not 0.4.

 

[travlerajm]

1. TR is wrong, so it's likely their calculation is bad.
2. for a 6-0 6-0 match, you still have to assume that indicates a certain difference in level. It looks like TR might assume 0.4, which is nearly a full level. If you're starting with a larger differential, that match might move you closer together. This is a weakness in the calculation and a reason they threw out 6-0 6-0 results for years. The differential for this match score is really a minimum and it's impossible to tell what the real differential is. If you played Djokovic, the score would be 6-0 6-0, but your NTRP level difference for that match is certainly not 0.4.

Seems like a difference of 0.8 or more would make more sense than 0.4 if they are trying to keep the algorithm simple.

 

[OnTheLine]

Hmm...according to UTR algorithm it takes the last 30 matches in the last 12 months but I've only played 15 matches this year and am 15-0. My UTR dropped from 6.20 to 6.14 after the double bagel.

But even if an older very good result didn't drop off entirely, as they age they factor less into the current rating.

So let's say prior to this match, your match #14 was contributing 5% to your rating ... but after you add another match it is now contributing 3% .....

That an UTR does weirdo modifications all the time to their ratings so it may jump tomorrow for no reason whatsoever ... or plummet for no reason whatsoever.

 

[denoted]

When it comes to ratings, everything is connected. Butterfly effect and so forth.

 

[CosmosMpower]

The real question is how is your UTR a 6 something if you're only a 3.49, should be like a 3.75-3.8 at UTR 6.

 

[rallyfries]

The real question is how is your UTR a 6 something if you're only a 3.49, should be like a 3.75-3.8 at UTR 6.

TR didn't count my NTRP Nationals tournament wins in April where I bageled and breadsticked everyone.

 

[time_fly]

TR didn't count my NTRP Nationals tournament wins in April where I bageled and breadsticked everyone.

That’s annoying. I bet they’re not giving you any credit for your Challenger points either. A lot of 3.5 self-rates have issues like that.

 

[rallyfries]

That’s annoying. I bet they’re not giving you any credit for your Challenger points either. A lot of 3.5 self-rates have issues like that.

I don't think any tournament matches count on TR which is unfortunate.

 

[OnTheLine]

When it comes to ratings, everything is connected. Butterfly effect and so forth.

Yup. Especially if someone that you beat 7-6; 7-6, then gets beaten by 0&1 by someone you lost to.

 

[mikelikesgatorad]

I once lost a match in juniors 6-0, 6-0 where I was pretty sure there were 7-8 deuce games.

And on the other side I played a JV player filling in for a HS singles match and won 6-0, 6-0, winning every point that I served

"Throwing out" those matches could make sense

 

[Purestriker]

So I am a 3.5C and had an estimated dynamic rating of 3.51 on TR. I recently played a league match against a 3.07 and won 6-0, 6-0 and my dynamic rating went down to 3.49 and my opponent's rating goes up to 3.09. How is that possible? At worst I would expect it to stay the same. I think my UTR also went down as well.

TR is garbage. Don't use it.

 

[travlerajm]

TR is garbage. Don't use it.

Nah... TR is a great resource. It’s public, transparent, user-friendly, easy-to-search, free, reasonably logical, and with a more stable and broadly useful algorithm than UTR.
If you don’t like the rating it spits out... train hard and play better next time.

 

[Purestriker]

Nah... TR is a great resource. It’s public, transparent, user-friendly, easy-to-search, free, reasonably logical, and with a more stable and broadly useful algorithm than UTR.
If you don’t like the rating it spits out... train hard and play better next time.

I don’t have an issue with the rating it gives me. But I do have an issue how it addresses self rates and I don’t think it’s logical. I prefer UTR.

 

[travlerajm]

I don’t have an issue with the rating it gives me. But I do have an issue how it addresses self rates and I don’t think it’s logical. I prefer UTR.

UTR ratings implode for anyone who is not playing usta matches all the time and hasn’t played in any matches for a few months. This means that UTR ratings are completely meaningless for 90% of the rec player population. I’d get more behind UTR if they fixed this.

 

[time_fly]

UTR ratings implode for anyone who is not playing usta matches all the time and hasn’t played in any matches for a few months. This means that UTR ratings are completely meaningless for 90% of the rec player population. I’d get more behind UTR if they fixed this.

I'm pretty sure UTR doesn't care much about USTA rec league players. I think they include us as a half-hearted way to try to pull us into UTR tournaments to make some extra money. But UTR is really for more formally competitive players -- either playing for school or playing regularly on some tournament circuit.

 

[travlerajm]

I'm pretty sure UTR doesn't care much about USTA rec league players. I think they include us as a half-hearted way to try to pull us into UTR tournaments to make some extra money. But UTR is really for more formally competitive players -- either playing for school or playing regularly on some tournament circuit.

I think the UTR people would make more money if they fixed the weak points in their algorithms so that more rec players actually thought UTR was useful.

 

[Purestriker]

UTR ratings implode for anyone who is not playing usta matches all the time and hasn’t played in any matches for a few months. This means that UTR ratings are completely meaningless for 90% of the rec player population. I’d get more behind UTR if they fixed this.

Ok. That makes sense. Fortunately in our area we play USTA and UTR all year around. So it doesn’t move as much or isn’t as out of date. TLS used to be another solution, but the site is rarely updated.

 

[Moveforwardalways]

TR ratings are low, very low. Lower than USTA.

 

[travlerajm]

One thing I have noticed about TR ratings in the past. The algorithm appears to take into account the USTA computer rating as part of the calibration to keep ratings from drifting too far out of alignment with the USTA computer.

For example, if the TR math calculates your rating to be in the 4.0 range based on match results, but the USTA computer says you are 3.5C, then TR apparently overrides the math and sets your rating at 3.49. It then probably uses this forced rating value at the top of the 3.5 range to calculate dynamic ratings for other players.

 

[Moveforwardalways]

One thing I have noticed about TR ratings in the past. The algorithm appears to take into account the USTA computer rating as part of the calibration to keep ratings from drifting too far out of alignment with the USTA computer.

For example, if the TR math calculates your rating to be in the 4.0 range based on match results, but the USTA computer says you are 3.5C, then TR apparently overrides the math and sets your rating at 3.49. It then probably uses this forced rating value at the top of the 3.5 range to calculate dynamic ratings for other players.

It does exactly that. Paradoxically, this is also what keeps it from closely tracking the USTA dynamic rating for bumps up and down.

 

[J_R_B]

Nah... TR is a great resource. It’s public, transparent, user-friendly, easy-to-search, free, reasonably logical, and with a more stable and broadly useful algorithm than UTR.
If you don’t like the rating it spits out... train hard and play better next time.

TR is definitely a nice resource where you can get some good information. The biggest issue is that insofar as they try to actually claim they replicate the USTA DNTRP calculations, that is wrong. Their ratings may or may not be a more accurate assessment of a player's actual underlying ability, but they are definitely not the same as the USTA's, and if you try to use TR as a proxy for your USTA DNTRP, that is a mistake. They are WAY off in some cases.

 

[denoted]

TR is definitely a nice resource where you can get some good information. The biggest issue is that insofar as they try to actually claim they replicate the USTA DNTRP calculations, that is wrong. Their ratings may or may not be a more accurate assessment of a player's actual underlying ability, but they are definitely not the same as the USTA's, and if you try to use TR as a proxy for your USTA DNTRP, that is a mistake. They are WAY off in some cases.

I know it's probably not a good use of my time to think about this, but I'm curious about why they are way off. The algorithm does not seem to be especially complex. I doubt that it can be. A commenter above suggests that the errors that result when it recalibrates to USTA end-of-year rankings are responsible. I think the small interaction effects might add up to large errors (that part is complex). And TR also only goes back to 2014, whereas the published USTA data goes back about twenty years. (Do they have earlier results digitized as well?)

 

[schmke]

I don't know what TR does, so can't speak to it, but I can say I observe some behaviors that indicate they do things differently than I do with my ratings. Does that make them "wrong" with respect to an accurate rating? I dunno, but observations from many in this forum do seem to indicate it doesn't track the USTA's year-end levels as well as you might hope if they are trying to replicate those.

 

[McLovin]

The TR algorithm for NTRP is, by definition, wrong. Why? Because NTRP is a scale defined and maintained by the USTA. Their algorithm is the only accurate way to calculate NTRP (unless they've hacked into the USTA and stolen their code). Anything else is simply an estimate.

The question is...do you believe TR's algorithm gives a more accurate representation of a player's skills? That was the whole idea behind UTR: A different way to estimate a player's level...NOT a different way to calculate NTRP.

 

[J_R_B]

The TR algorithm for NTRP is, by definition, wrong. Why? Because NTRP is a scale defined and maintained by the USTA. Their algorithm is the only accurate way to calculate NTRP (unless they've hacked into the USTA and stolen their code). Anything else is simply an estimate.

The question is...do you believe TR's algorithm gives a more accurate representation of a player's skills? That was the whole idea behind UTR: A different way to estimate a player's level...NOT a different way to calculate NTRP.

The problem is that TR seems to want to market itself as an offline calculation of NTRP, because,why else would they "reset" all their bump misses at year end? They would be mor genuine in saying, hey, this is our calculation and it's different, like UTR is doing.

 

[Moveforwardalways]

In 2019 there were many (as in, a lot) of players rated X.75 on TR that were bumped by the USTA to the next level.