UTR algorithm

pmoreira

New User
I have read about UTR (Universal Tennis Rating) and am curious. How is the UTR algorithm?

I know the general idea from myutr.com but I want the details. The answer can be a reference elsewhere.
 

OnTheLine

Hall of Fame
I don't have the details ... but I will add my thoughts ...

They have an algorithm
It includes how your opponents have performed after you have played them

Then they decide that the entire chain of ratings is wrong or "off"
Then they put in a "fix" that changes sub UTR 10 player's ratings with swings of up to 1.25 randomly. Up or down. Can be on a near daily basis even if you have not played a match (nor have your opponents/partners) while they tweak their numbers
If you are an adult their system suppresses your rating by a minimum of 1.5
If you are a junior their system inflates your rating by a minimum of 1.5
 

BallBag

Professional
Thanks for your answer, but I am interested in the maths.
The algorithm is proprietary and I don't know of anyone that reverse engineered it. The ITF's WTN algorithm should do something similar and will be based on Glicko 2.
 
Except that the ELO system is actually stable with ratings that don’t implode upon inactivity.
Well there's that... :D

For UTR, as with any data set, the smaller the pool, the more variance. I'm guessing that the algorithm was not optimized [or perhaps even back-tested] for inactivity, hence all sorts of "fat tail" events are happening [my UTR jumped 200 points in the last few months without me playing a single match].

Elo* originally had ratings gently decline during inactivity which sandbaggers used to artificially depress their rating before a big tournament. At some point they changed this because now one's rating does not decline during inactivity.

*it's "Elo" as in "Dr. Arpad Elo", the inventor of the algorithm. It's not ELO, which is Electric Light Orchestra, IMO.
 

ChaelAZ

G.O.A.T.
UTR is wonderful when it has lots of data points, but can be less accurate/useful for rec players who don't play as often.
 

schmke

Hall of Fame
Except that the ELO system is actually stable with ratings that don’t implode upon inactivity.
I don't believe UTR is anything like pure Elo. They may have taken some inspirations from it and/or the core algorithm to determine match ratings likely uses it or something very similar.
 

OnTheLine

Hall of Fame
UTR is wonderful when it has lots of data points, but can be less accurate/useful for rec players who don't play as often.
Define "as often"
I play 6-8 rated matches per month. I likely play more rated matches than many juniors.
Consistent with perhaps 2 weeks off at a time.
My UTR will swing from high 2 to low 5 randomly ... I can assure you my match score results are pretty consistent.
 

jmnk

Hall of Fame
Define "as often"
I play 6-8 rated matches per month. I likely play more rated matches than many juniors.
Consistent with perhaps 2 weeks off at a time.
My UTR will swing from high 2 to low 5 randomly ... I can assure you my match score results are pretty consistent.
hmm, that is weird. Clearly can't verify without a name - but I have not seen a single junior UTR ranking varying soo much while playing so many rated matches (assuming those 'rated' matches are against properly ranked UTR players).
 
Define "as often"
I play 6-8 rated matches per month. I likely play more rated matches than many juniors.
Consistent with perhaps 2 weeks off at a time.
My UTR will swing from high 2 to low 5 randomly ... I can assure you my match score results are pretty consistent.
It would be interesting to track your opponents' UTRs and see if they are moving similarly.

If your rating jumps around but their's are too, that at least is consistent [although what's the source of their ratings moving around: their opponents? Then you get into a series of loops...which might all come back to you!].

But if your opponents' UTRs are relatively stable, it makes no sense at all that yours moves so much.
 

travlerajm

G.O.A.T.
Define "as often"
I play 6-8 rated matches per month. I likely play more rated matches than many juniors.
Consistent with perhaps 2 weeks off at a time.
My UTR will swing from high 2 to low 5 randomly ... I can assure you my match score results are pretty consistent.
My singles UTR drifted from 10 to 5 before it decided to go to UR. That means you and I had the same UTR at some point.

Meanwhile, the other ratings sites said I was better at maintaining my 4.5-5.0 level.
 

OnTheLine

Hall of Fame
hmm, that is weird. Clearly can't verify without a name - but I have not seen a single junior UTR ranking varying soo much while playing so many rated matches (assuming those 'rated' matches are against properly ranked UTR players).
I haven't played against or with player's without a verified rating much if at all. And how can you verify with a name (not that I would be providing it) and see how the rating shifted? Talking about a rating on a Monday at 3.29, the next morning (no matches added or subtracted) a 5.34 then by Wednesday a 4.43 still with no new data points for me or opponents. (a down 2 week period in league matches, no local tournaments either)


It would be interesting to track your opponents' UTRs and see if they are moving similarly.

If your rating jumps around but their's are too, that at least is consistent [although what's the source of their ratings moving around: their opponents? Then you get into a series of loops...which might all come back to you!].

But if your opponents' UTRs are relatively stable, it makes no sense at all that yours moves so much.
Our entire area has had wild swings the past few months. Men and women. For a while most 3.5 women were majority 2s with some 3s. Then majority 3s with some 4s but only after shifting to majority 4s with some 5s. Men I played with in mixed saw the same thing happening with shifts from 4s and 5s to 6s and 7s then back to 5s and 6s

These are shifts that are on a random day to day basis where it seems everyone shifts then re-shifts. Someone is playing with the algorithm to tweak it.

My singles UTR drifted from 10 to 5 before it decided to go to UR. That means you and I had the same UTR at some point.

Meanwhile, the other ratings sites said I was better at maintaining my 4.5-5.0 level.
That is hilarious.
 

J011yroger

Talk Tennis Guru
At one point I think I saw a 4.5 male and their 3.5 female mixed partner both UTR 5s.
I played 9.0 and my 63 year old 4.0 lady partner had a higher UTR than I did...

Meanwhile back in reality, I am handily beating 9&10 UTR juniors.

J
 

CosmosMpower

Hall of Fame
I played 9.0 and my 63 year old 4.0 lady partner had a higher UTR than I did...

Meanwhile back in reality, I am handily beating 9&10 UTR juniors.

J
I don't think juniors UTR ratings are comparable to adults. They only play other juniors mostly to get their rating and it's a whole different ball game against an adult unless you're talking 16/17 year old superchamps
 

ChaelAZ

G.O.A.T.
Define "as often"
I play 6-8 rated matches per month. I likely play more rated matches than many juniors.
Consistent with perhaps 2 weeks off at a time.
My UTR will swing from high 2 to low 5 randomly ... I can assure you my match score results are pretty consistent.

I meant more like juniors and college players that play more often, or even rec folks who do a few dozen matches a season. The point being more data points give more reliable data so all.

I kid you not, I my UTR is only 2.79 at the moment, down from almost 6 last summer. Crazy swings in play and few matches cause crazy swings in UTR, and I only had like one singles match I played like absolute *** in and a few decent dubs wins, but low ranked opponents. Once all the current matches start posting I assume it will swing back up. Thanks COVID. lol.
 

OnTheLine

Hall of Fame
I meant more like juniors and college players that play more often, or even rec folks who do a few dozen matches a season. The point being more data points give more reliable data so all.

I kid you not, I my UTR is only 2.79 at the moment, down from almost 6 last summer. Crazy swings in play and few matches cause crazy swings in UTR, and I only had like one singles match I played like absolute *** in and a few decent dubs wins, but low ranked opponents. Once all the current matches start posting I assume it will swing back up. Thanks COVID. lol.
Yeah totally get that .... so far for YTD 2021 I have only 7 rated matches ... but for last 12 months 46 ... my matches from mid July are starting to fall off as they only keep X number as part of your rating.

So I looked this morning and I was at 4.13. Just for fun, looked again right now ... went to 3.83. Trust me: nothing happened in those 6 hours. No new matches. Absolutely nothing .... except for the UTR algorithm fairy doing something.
 

Purestriker

New User
I don't have the details ... but I will add my thoughts ...

They have an algorithm
It includes how your opponents have performed after you have played them

Then they decide that the entire chain of ratings is wrong or "off"
Then they put in a "fix" that changes sub UTR 10 player's ratings with swings of up to 1.25 randomly. Up or down. Can be on a near daily basis even if you have not played a match (nor have your opponents/partners) while they tweak their numbers
If you are an adult their system suppresses your rating by a minimum of 1.5
If you are a junior their system inflates your rating by a minimum of 1.5
I actually like that it is constantly updating. I do agree that the junior ratings seemed to be inflated, but I have found the adult ratings to be more accurate. The key is to have enough matches for their algorithm to get an accurate rating.
 
I have read about UTR (Universal Tennis Rating) and am curious. How is the UTR algorithm?

I know the general idea from myutr.com but I want the details. The answer can be a reference elsewhere.
I do have a UTR but I also do agree since I'm a junior my ranking inflates like crazy even though here in hawaii our juniors aren't ranked very high so ours don't fluctuate like in europe or the 48 states.
 

OnTheLine

Hall of Fame
Finally caught up and I am back in the mid 5 range. lol. Again, without a lot of data points it really is kind of nebulous for rec players at times.
In our area over the past 3-4 days they did ANOTHER weird adjustment .... most men moved about a full 1.0 up and all the women dropped nearly the same.

I went from a 4.13 last Monday now a 3.58! One match went in and is was a 6-2; 6-3 win against a pair who had both been 4s and now are 3s.

And it is clearly a manual adjustment as you can look across your history and partners/opponents everyone was dropped.

So now once again there are women who are 2.5/3.0/3.5C rated players .... who are all UTR 2. Yeah, whatever.
 

travlerajm

G.O.A.T.
In our area over the past 3-4 days they did ANOTHER weird adjustment .... most men moved about a full 1.0 up and all the women dropped nearly the same.

I went from a 4.13 last Monday now a 3.58! One match went in and is was a 6-2; 6-3 win against a pair who had both been 4s and now are 3s.

And it is clearly a manual adjustment as you can look across your history and partners/opponents everyone was dropped.

So now once again there are women who are 2.5/3.0/3.5C rated players .... who are all UTR 2. Yeah, whatever.
Wake me up when the UTR algorithm wizard figures it out.
 

FIRETennis

Professional
https://support.myutr.com/en/support/solutions/articles/9000151830-understanding-the-algorithm-complete-summary

The biggest issue with UTR is that it has a strange rating decay if you do not play often. This has been exacerbated by the COVID situation in various countries where you have some players playing a lot, and some players haven't had a verified UTR match in a year or more. I do not understand why in light of the current situation, UTR hasn't adjusted their algo to account for this. Aside this issue, if you are playing verified users, often, it's pretty darn good. Not perfect but certainly better than the other options out there (NTRP or "very good, good, intermediate, not so good, beginner" ratings).
 
Top