Why are USTA’s dynamic ratings a big secret?

SliceGirl

New User
With the new USTA ratings coming out soon, I’m wondering why there is so much mystery around dynamic ratings. Why doesn’t the organization make dynamic ratings public?
 

schmke

Legend
The USTA is afraid that more transparency into the ratings would result in more players trying to manipulate their ratings to stay at level or get bumped down. And there is probably some truth to that, it is the nature of a level based system where people naturally want to win and that is easier to do if they are at the top of one level instead of the bottom of the next higher level.
 

SaltyDDDog

New User
Can someone share where the official USTA ratings are made available, once the new ones come out in a few weeks? On the USTA site, I've never seen anything other than 3.5C, for example. Thank you!
 

brettatk

Semi-Pro
How do you even manage rating? Lose on purpose?

Sent from my SM-G975U using Tapatalk
Exactly. Most of the time it's losing games on purpose. Where you could have won say 6-1 6-1 you give up enough games to win 6-4 6-3. Some guys even give up a set if they know they'll win the 3rd. Another thing I see is when a team clinches the overall match with 3 points. Someone who manages their rating playing the other two lines might not even try and will lose on purpose. Now it's not the majority of guys but it does happen a lot.

I also think if they did publish the ratings it would be more blatant and happen more. Right now you don't know exactly how low someone is in their rating so it's guess work for those who manage. Maybe it makes it a little tougher on them. I wish there was a solution to this issue but you'll always have those who find ways around it.
 

texacali

Rookie
Does tracking dynamic rating on Tennis Record give any indication of USTA dynamic? Does this help "manage" a rating?
 

Vox Rationis

Semi-Pro
Does tracking dynamic rating on Tennis Record give any indication of USTA dynamic? Does this help "manage" a rating?
Sure. People probably use it for that. But their ratings can be unreliable and people that manage their ratings probably don't need it to know they're on track. It's not like it's hard to cheat the system. Most people are just above that.
 

V-Werks

New User
Can someone share where the official USTA ratings are made available, once the new ones come out in a few weeks? On the USTA site, I've never seen anything other than 3.5C, for example. Thank you!
NTRP year-end ratings are due December 2 or 3rd. So, you won't have to wait to much longer! NTRP year-end ratings are disclosed in half-point increments with a certain designation letter. The letter "C" indicates that the rating was computed by the algorithm based on USTA match play.
 

ChaelAZ

G.O.A.T.
What is interesting to me is, I am no math wizard but seems the rating forumla isn't that complex that folks couldn't figure it out somewhere close anyway. And the TR gives it away for free, +/- a few points. Surprise someone hasn't posted their forumulas yet.
 

Nostradamus

Bionic Poster
With the new USTA ratings coming out soon, I’m wondering why there is so much mystery around dynamic ratings. Why doesn’t the organization make dynamic ratings public?
their computer program is one big piece of garbage and it doesn't work for all players due to variability in amount of matches played. when quantum computing becomes available with true AI built in, It will do a better job. or hire a real competent computer programmers and do the job correctly. Just look at the USTA website, it is full of cutting corners and cheap quick piece of programming all over the place
 

Max G.

Legend
What is interesting to me is, I am no math wizard but seems the rating forumla isn't that complex that folks couldn't figure it out somewhere close anyway. And the TR gives it away for free, +/- a few points. Surprise someone hasn't posted their forumulas yet.
I mean, Shemke did figure it out, so it's not particularly well-hidden.
 

V-Werks

New User
What is interesting to me is, I am no math wizard but seems the rating forumla isn't that complex that folks couldn't figure it out somewhere close anyway. And the TR gives it away for free, +/- a few points. Surprise someone hasn't posted their forumulas yet.
I'd be wary of the Dynamic results on TR especially when "S" players participated in the matches (for either side). Self-rated players create nuances for the ratings of all of the players who participated in the matches with them because the results are not counted in the dynamic ratings. However, the results that were ignored in the dynamic rating calculations (which is the rating the TR is attempting to estimate) might be included in the official year-end rating calculation.

Essentially, if the "S" rated player participates in three valid matches by the end of the rating year their match results, along with those who participated with them, retroactively get included in the year-end rating. Hence, it is easy to see why dynamic rating estimates might differ from year-end ratings from the existence of "S" players. There are also other reasons why results might differ.
 

ChaelAZ

G.O.A.T.
I'd be wary of the Dynamic results on TR especially when "S" players participated in the matches (for either side). Self-rated players create nuances for the ratings of all of the players who participated in the matches with them because the results are not counted in the dynamic ratings. However, the results that were ignored in the dynamic rating calculations (which is the rating the TR is attempting to estimate) might be included in the official year-end rating calculation.

Essentially, if the "S" rated player participates in three valid matches by the end of the rating year their match results, along with those who participated with them, retroactively get included in the year-end rating. Hence, it is easy to see why dynamic rating estimates might differ from year-end ratings from the existence of "S" players. There are also other reasons why results might differ.
Good point. We had a TON of 'S' rated players in 4.0 this year, so wondering where all that will play out in my YE rating. Out last match we played D1 aginst a 3.42 (3.5C) and 3.79(4.0S) and I played with an S partner with no rating yet. We ended up winning 7/5, 6/2 but the match rating was super low and looks like it had no effect on my dynamic rating (though my 0/3 in singles with some poultry results will do more to kill it than anything).

That S is nebulous and suppose that would be hard to account for in a rating formula.

EDIT to add: I just checked on TR and wow...out of the 9 matches this season, 5 show S, so not counted in rating. Crazy!
 
Last edited:

J_R_B

Hall of Fame
What is interesting to me is, I am no math wizard but seems the rating forumla isn't that complex that folks couldn't figure it out somewhere close anyway. And the TR gives it away for free, +/- a few points. Surprise someone hasn't posted their forumulas yet.
TR's rating algorithm is wrong, and their results are often way off because of it. Schmke figured it out a lot better, but I think he used a more complex machine learning or other type of algorithm to match the bumps/non-bumps after he amassed a critical mass of match data.
 
It has always been my unscientific opinion that TR started their ratings from scratch without looking backwards to figure out where each player was whenever they began. This also applies to how they start self-rates. So, for example, all 4.0s started at 3.75. As the years go by and they make adjustments, that massive error will shrink. I think we are still many, many years (or adjustments) away from TR becoming a closer approximation.
 

J_R_B

Hall of Fame
It has always been my unscientific opinion that TR started their ratings from scratch without looking backwards to figure out where each player was whenever they began. This also applies to how they start self-rates. So, for example, all 4.0s started at 3.75. As the years go by and they make adjustments, that massive error will shrink. I think we are still many, many years (or adjustments) away from TR becoming a closer approximation.
They used to do that (start S-rates with a middle-level rating), but in the last year or two, they've changed that to start S-rates with no rating. Their rating formula itself is incorrect in at least one way that I know about.
 

Chalkdust

Semi-Pro
How do you even manage rating? Lose on purpose?

Sent from my SM-G975U using Tapatalk
You don't even have to lose matches. Since rating is based on actual games won/lost not match won/lost. So for example you can have an unbeaten season winning every match 6-4 0-6 1-0 and still actually have your rating go down.
 

jm1980

Talk Tennis Guru
You don't even have to lose matches. Since rating is based on actual games won/lost not match won/lost. So for example you can have an unbeaten season winning every match 6-4 0-6 1-0 and still actually have your rating go down.
The rating is also based on the ratings of your opponents. So you could win every match 6-4 6-4 and your DNTRP could still go down.
 
TR usually updates weekly, pulling in matches from the prior week. This week it looks like it just updated its calculations but didn't pull in any matches post 11/10 - I assume with the objective of mimicking USTA YE ratings coming out in a few days.

I had assumed that 2020 leagues (which started in October here) wouldn't count for 2019 YE ratings but rather for next year's rating. However TR is using any matches prior to 11/10, regardless of "league year." Is that how it's believed that USTA treats matches for the new league year prior to the prior year's match cutoff date?
 

Chalkdust

Semi-Pro
The rating is also based on the ratings of your opponents. So you could win every match 6-4 6-4 and your DNTRP could still go down.
Yes of course.
But in the context of rating manipulation while still winning the match: You don't know your opponents precise ratings, so winning 6-4 6-4 is not a good strategy. Sure your rating could go down if the opponents have low ratings but you just don't know. Your best bet is to lose as many games as possible. The best possible outcome is winning 7-6 0-6 1-0 which pretty much guarantees your rating will decrease unless you are playing someone a full level higher.
 

Vox Rationis

Semi-Pro
Yes of course.
But in the context of rating manipulation while still winning the match: You don't know your opponents precise ratings, so winning 6-4 6-4 is not a good strategy. Sure your rating could go down if the opponents have low ratings but you just don't know. Your best bet is to lose as many games as possible. The best possible outcome is winning 7-6 0-6 1-0 which pretty much guarantees your rating will decrease unless you are playing someone a full level higher.
I disagree. If you have a decent grasp of how the ratings work then it's pretty easy to have an accurate rough estimation of your opponents rating. You don't need to have the precise rating to know how a 6-4 6-4 win could affect you.

Also if you're manipulating the score to be 6-4 6-4 instead of dropping a set then you probably way outclass your opponent and are going for plausible deniability.
 

rogerroger917

Hall of Fame
Seems like a lot of work to try to remember how any games you need to lose against worse players.

Sent from my SM-G975U using Tapatalk
 

V-Werks

New User
TR usually updates weekly, pulling in matches from the prior week. This week it looks like it just updated its calculations but didn't pull in any matches post 11/10 - I assume with the objective of mimicking USTA YE ratings coming out in a few days.

I had assumed that 2020 leagues (which started in October here) wouldn't count for 2019 YE ratings but rather for next year's rating. However TR is using any matches prior to 11/10, regardless of "league year." Is that how it's believed that USTA treats matches for the new league year prior to the prior year's match cutoff date?
Yes.
 

V-Werks

New User
Good point. We had a TON of 'S' rated players in 4.0 this year, so wondering where all that will play out in my YE rating. Out last match we played D1 aginst a 3.42 (3.5C) and 3.79(4.0S) and I played with an S partner with no rating yet. We ended up winning 7/5, 6/2 but the match rating was super low and looks like it had no effect on my dynamic rating (though my 0/3 in singles with some poultry results will do more to kill it than anything).

That S is nebulous and suppose that would be hard to account for in a rating formula.

EDIT to add: I just checked on TR and wow...out of the 9 matches this season, 5 show S, so not counted in rating. Crazy!
I understand that players want to think of ratings based on win/loss results. But, this is a mistake since neither the dynamic nor year-end NTRP ratings consider win/loss results in the methodology. The element that is considered is actual service breaks vs. estimated based on valid match play. To increase your rating you must outperform expectations regardless of win/loss results.
 

Matthew ATX

Semi-Pro
Service breaks? How could that possibly be a factor? There's no way to know how many breaks occurred in a match using just the score.
 
Top