TLS Rating vs Tennisrecords

Matthew ATX

Semi-Pro
So when does TR start updating again? I was under the impression that matches played after last year's cutoff would count for this year. Both of mine are still listed as NC.
Also, we are two weeks into the new season and no played matches are showing up yet.
 

J011yroger

Talk Tennis Guru
So when does TR start updating again? I was under the impression that matches played after last year's cutoff would count for this year. Both of mine are still listed as NC.
Also, we are two weeks into the new season and no played matches are showing up yet.
*Sad trombone plays*

J
 

time_fly

Hall of Fame
I was wondering the same thing. I think they may have given up, although they haven't pulled the website down yet. Too bad there's no contact info and the domain is registered anonymously by proxy. I'd even offer to help them. Even if the ratings weren't that accurate, their site made scouting really easy.
 

J_R_B

Hall of Fame
Their ratings have gotten more and more inaccurate over the years. For the Middle States 4.0 40+ mens team that went to nationals, they predicted there would be 1 bump. There were 8.
 

OnTheLine

Hall of Fame
As long as they keep having advertisers on their site, they will stay up. If their visitor volume drops they will lose their advertisers. So, my guess is they will start updating soon .... it is helpful for scouting .... I am annoyed by having to use Tennislink right now ... way less friendly for scouting and creating lineups.

I will say, with no new ratings so far this season .... my teammates are way less stressed than when they saw their rating rise/drop on a weekly basis.
 

time_fly

Hall of Fame
I will say, with no new ratings so far this season .... my teammates are way less stressed than when they saw their rating rise/drop on a weekly basis.
Don't worry, TLS is still doing updates so they have something to obsess over. It's not weekly unfortunately.
 

dbm502

New User
There are 2 sites calculating NTRP ratings currently and I am seeing different rating results for the same player. Which site's numbers do you think is more accurate?
A bunch of us looked at these in depth 2 weeks after year end rankings & TLS was a little more accurate. We went by which site predicted the most players actually getting bumped. TLS predicted 42% & TR 25%. I'm sure both will get you in the ballpark, but TLS did a better job predicting players getting bumped up & down than TR.
 

kelkat

Rookie
So it's been a while since the last post on this thread. TR has me at 3.85, TLS at 3.63. Seems like TR updates at least every 10 days, while Ive seen no change in TLS in weeks....
 

5sets

Professional
So it's been a while since the last post on this thread. TR has me at 3.85, TLS at 3.63. Seems like TR updates at least every 10 days, while Ive seen no change in TLS in weeks....
I like tennisrecord. It’s algorithm seems to at least place people in .10 of actual USTA dntrp. Can’t say the same for TLS, and it’s not as user friendly. If anything tennisrecord is helpful in looking up match history for a player.


Sent from my iPhone using Tapatalk
 

5sets

Professional
Did you ever notice on TR though that it gives different match ratings for yourself on your own profile page then on your opponents? It’s weird. I see this in Doubles matches. It will give me a match rating and then if I click on my partner or one one of my opponents my match rating is sometimes
.10 different.

Anybody care to elaborate or explain?


Sent from my iPhone using Tapatalk
 

Vox Rationis

Semi-Pro
Did you ever notice on TR though that it gives different match ratings for yourself on your own profile page then on your opponents? It’s weird. I see this in Doubles matches. It will give me a match rating and then if I click on my partner or one one of my opponents my match rating is sometimes
.10 different.

Anybody care to elaborate or explain?


Sent from my iPhone using Tapatalk
Do you mean the number in the parenthesis by your name? If you're on your partner or opponents' page, I believe that is your estimated dynamic rating going into the match. Not your match derived rating.
 

5sets

Professional
Do you mean the number in the parenthesis by your name? If you're on your partner or opponents' page, I believe that is your estimated dynamic rating going into the match. Not your match derived rating.
Yes! So then it has different estimated dynamic ratings for me depending on whose page I look at. Try it with you.


Sent from my iPhone using Tapatalk
 

5sets

Professional
Do you mean the number in the parenthesis by your name? If you're on your partner or opponents' page, I believe that is your estimated dynamic rating going into the match. Not your match derived rating.
Ohhh, I got you. I was looking at my match rating on the bottom left and then looking at my opponents page and it lists my rating in parentheses which is my dNTRP going into the match.

That’s why they are different.

Thanks


Sent from my iPhone using Tapatalk
 

J_R_B

Hall of Fame
T
I like tennisrecord. It’s algorithm seems to at least place people in .10 of actual USTA dntrp. Can’t say the same for TLS, and it’s not as user friendly. If anything tennisrecord is helpful in looking up match history for a player.


Sent from my iPhone using Tapatalk
TR definitely wins on user friendliness of the site. For that, it's actually great. For their ratings, much less so.
 

5sets

Professional
Honestly, I use UTR the most now. It’s great because it calculates non League matches as well and and shows updates daily


Sent from my iPhone using Tapatalk
 

5sets

Professional
You got to give TR credit where credit’s due. I mean they don’t have anyone’s initial rating to the hundredth so it will take a few matches before they even get close to what usta’s is. When I look at TLS people seem either drastically overrated or underrated and no updates.


Sent from my iPhone using Tapatalk
 

5sets

Professional
How do you compare and conclude it is within 0.1 of USTA DNTRP?
Just by seeing people I know and have played with and/or against, of course nobody knows the secret USTA algorithm, but TR seems much more close than TLS


Sent from my iPhone using Tapatalk
 

Ft.S

Semi-Pro
Just by seeing people I know and have played with and/or against, of course nobody knows the secret USTA algorithm, but TR seems much more close than TLS
Sent from my iPhone using Tapatalk
I also think TR is more accurate than TLS, but as soon as you have more than 12-14 matches on record, TR discrepency against USTA DNTRP can get as large as 0.3 from my experience, based on reports I have received from Kevin. So, IMO both are useless from a ratings perspective. I do like all the other stats and historical data in TR, very useful and helpful.
 

5sets

Professional
Agreed. When all’s said and done I like playing the tennis ball in front of me. I usually screw myself up if I’m thinking I’m 0.21 higher than the guy on the other side of the net so I should be winning easily. I just like looking at TR. The simple fact that it updates after every match makes it fun to look at even if it is off. And it’s free and more user savvy than tennislink.


Sent from my iPhone using Tapatalk
 

time_fly

Hall of Fame
This morning for fun -- I have a weird idea of fun -- I took a look at our men's 3.5 leagues in TR. There are 58 rated players. The ratings have a roughly normal distribution, but the median and average are both 3.15. The 75th percentile is only 3.28, and there are only 5 players rated 3.40 and above. Of those, one is a self-rate with only 2 matches, one is a recent bump-down, one was a self-rate last year and hasn't played any matches this year, and one was a self-rate last year that plays more 4.0 matches than 3.5. So these guys are all somewhat unusual.

So it seems like there is a -0.1 shift in our ratings compared to what I would expect, and almost everyone who has been in the league for a few years without playing up is 3.39 or lower. I'm not sure if this is an intentional effect to model regional variations or if it is something that has just happened as a side-effect of their ratings calculation.
 

5sets

Professional
This morning for fun -- I have a weird idea of fun -- I took a look at our men's 3.5 leagues in TR. There are 58 rated players. The ratings have a roughly normal distribution, but the median and average are both 3.15. The 75th percentile is only 3.28, and there are only 5 players rated 3.40 and above. Of those, one is a self-rate with only 2 matches, one is a recent bump-down, one was a self-rate last year and hasn't played any matches this year, and one was a self-rate last year that plays more 4.0 matches than 3.5. So these guys are all somewhat unusual.

So it seems like there is a -0.1 shift in our ratings compared to what I would expect, and almost everyone who has been in the league for a few years without playing up is 3.39 or lower. I'm not sure if this is an intentional effect to model regional variations or if it is something that has just happened as a side-effect of their ratings calculation.
Yup, exactly, these sites and Schmke ‘s calculations are just that, fun. They can guesstimate and hypothesize this and that but at the end of the day their ratings are not usta’s. And what does it matter? Do you really want to know your opponents dNTRP going into a match? Afterwards, like I said it’s fun when you’re home, have a shower, and crack open an ice cold beer, but knowing before can mess with your head.


Sent from my iPhone using Tapatalk
 

schmke

Hall of Fame
Yup, exactly, these sites and Schmke ‘s calculations are just that, fun. They can guesstimate and hypothesize this and that but at the end of the day their ratings are not usta’s. And what does it matter? Do you really want to know your opponents dNTRP going into a match? Afterwards, like I said it’s fun when you’re home, have a shower, and crack open an ice cold beer, but knowing before can mess with your head.
They are fun with the right mindset like you say. And no, my ratings aren't perfect, but I think they are a more accurate representation than others ...

And despite my having all my ratings at my disposal, I generally don't rush to lookup my opponent when I find out who it is, but will look afterwards. Knowing I'm supposed to win or lose a match shouldn't affect how I play in the match as I'm going to try my best regardless. The one thing knowing beforehand can help with is in doubles, knowing who, on paper at least, is the stronger/weaker opponent, can be helpful to aid in having an initial strategy of who to target more.
 

ShaunS

Semi-Pro
The one thing knowing beforehand can help with is in doubles, knowing who, on paper at least, is the stronger/weaker opponent, can be helpful to aid in having an initial strategy of who to target more.
It occurs to me that @schmke could be pulling a long-term covert operation. Stay with me here!

He's going to spend a couple decades giving out accurate information. Then when his team is at nationals, and one of the opposing teams asks for a 'Team Report', he'll feed them bad data!

Nice try buddy, nice try.

I generally don't rush to lookup my opponent when I find out who it is, but will look afterwards.
This is what I enjoy most from them. It doesn't matter what their rating is in terms of my play, but I like to know what the NTRP system thinks of my opponent relative to my experience. With that in mind, it's why I don't stress too much about the inaccuracies of the unofficial sources. The NTRP ratings themselves are fairly imprecise because it's hard to rate human tennis players with that kind of precision.
 

schmke

Hall of Fame
It occurs to me that @schmke could be pulling a long-term covert operation. Stay with me here!

He's going to spend a couple decades giving out accurate information. Then when his team is at nationals, and one of the opposing teams asks for a 'Team Report', he'll feed them bad data!

Nice try buddy, nice try.
Darn, you are on to me ...
 

schmke

Hall of Fame
And he's using the profits from the reports to pay for his team to go.
Profits? No one told me about profits! This is all a labor of love :)

Now if I can just get on a team that qualifies to go! Alas, I'm not a ringer at 4.5 that captains are seeking out ...
 

pixel

New User
Seems like TR and TLS both gave up. New England TLS ratings haven't been updated since the beginning of August. TR stopped when ratings came out; maybe they are rethinking their whole algorithm once they compared the real bump list to their predicted one? Or maybe they are tired of being bashed here.

To be fair to these guys, it has to be pretty hard to reverse-engineer the USTA algorithm details with only an annual bump list to go by as ground truth. Because the method involves basically an iterative network calculation, any slight inaccuracies are going to propagate and compound quickly.
More wait time means more website visit and more clicks from players anxious to see their updated rankings. This is how they make money and get ads.
 

pixel

New User
Looks like tennisrecords.com and TLS have slightly different algorithms. I played a team with an average rating of about 2.85. The average rating of my partner and I was 3.15. we won 7-6, 6-4, 1-0 . Tennisrecords.com calculated the match rating as 2.67 and reduced my rating by 0.09 points! Looks like the number of games won or lost is quite important.
 

OnTheLine

Hall of Fame
Looks like tennisrecords.com and TLS have slightly different algorithms. I played a team with an average rating of about 2.85. The average rating of my partner and I was 3.15. we won 7-6, 6-4, 1-0 . Tennisrecords.com calculated the match rating as 2.67 and reduced my rating by 0.09 points! Looks like the number of games won or lost is quite important.
Yes the differential of games won is all important. The way you put in the score here could mean that you you either won by 12-10 or you won by 13-11 .... in either case with a ratings difference of 0.30 you should likely have won by a margin of at least 6 if not 8 so yes your rating would plummet based on the result of only a differential of 2.

Put another way:

With a court difference of 0.30 the computer likely spit out an expected scoreline something along the lines of a 6-2; 6-2 win for you. That is a game difference of 8
Your team greatly under-performed completing with a game difference of only 2
Your team would see a ratings decrease and your opponents a ratings increase.

Technically, someone could lose every single match throughout a year but beat the computer outcome estimation every time and their rating at the end of the year would be higher than at the beginning.
 

pixel

New User
Thanks a lot for the reply. Yes, we won 13-11 games.
It looks like playing with a higher-ranked partner also hurts your rankings unless you win really decisively!
 

schmke

Hall of Fame
How do you think your scores reflect on these other sites?

I think your are more accurate but intrested in your thoughts.

I see big swings in the three.
I think my ratings are more accurate when it comes to predicting year-end levels. The feedback I get from folks is they find my ratings more accurate as well. That said, mine are not perfect, there are adjustments the USTA seems to make as a result of Nationals that don't appear to be purely algorithm based.

Why are mine different? I can see some things the other sites do that is different and I believe from my research what I'm doing is closer to what the USTA does.
 

pixel

New User
schmke, I have read your blog . I would rate, you first, then TLS and then tennisrecord. com for accuracy.
I know for a fact that in the league ( or at least everyone I know), people take tennisrecord.com ratings very seriously. Partly because it updates ( or used to update) very frequently.
USTA considers tournaments and I think that makes a difference.
I remember a long time ago I saw a formula online for the USTA algorithm. I can't find it anymore!
I think after reading all this, I found the secret to move up! Lose no more than 4 games in a match!
 

Rusbus

New User
For my example, I am almost a full level swing between sites. I am a 3.83 on tls and tennis record and a 4.22 through yours. Your levels line up more closely for other players I play but who knows. I think me playing a whole season with an S rated player really skewed things for me.
 

OnTheLine

Hall of Fame
I think my ratings are more accurate when it comes to predicting year-end levels. The feedback I get from folks is they find my ratings more accurate as well. That said, mine are not perfect, there are adjustments the USTA seems to make as a result of Nationals that don't appear to be purely algorithm based.

Why are mine different? I can see some things the other sites do that is different and I believe from my research what I'm doing is closer to what the USTA does.
I as you know think your ratings are the cat's meow.

That said in terms of what the other sites are doing:

TLS has me always very low, TR higher and when I got my ratings from you last year you were nearly 0.20 higher than TR for me and 0.42 higher than TLS which was a crazy swing.

If your ratings are again 0.20 higher than TR I may be bumped to 4.0 this year ... which is not in my plans exactly just yet.
 

OnTheLine

Hall of Fame
schmke, I have read your blog . I would rate, you first, then TLS and then tennisrecord. com for accuracy.
I know for a fact that in the league ( or at least everyone I know), people take tennisrecord.com ratings very seriously. Partly because it updates ( or used to update) very frequently.
USTA considers tournaments and I think that makes a difference.
I remember a long time ago I saw a formula online for the USTA algorithm. I can't find it anymore!
I think after reading all this, I found the secret to move up! Lose no more than 4 games in a match!
First, USTA does NOT consider tournaments in all regions/sections ... only in some. And from information from my LLC they count for much less than people think.

I am guessing you are joking about finding the secret ... :) I wish it was that simple. But if you lose 4 games to opponents that the computer thinks you should double-bagel ..... well, the "secret" won't work.

If I were to put out a "secret" it would be: In doubles, play with the lowest rated person you can find (say a 2.5 playing up on your 3.0 team), play on line 1 assuming you play against higher rated players and greatly beat what the computer thinks the score should be.

The REAL secret is: practice hard, become better, dominate your level and you will be bumped up.
 

schmke

Hall of Fame
I as you know think your ratings are the cat's meow.

That said in terms of what the other sites are doing:

TLS has me always very low, TR higher and when I got my ratings from you last year you were nearly 0.20 higher than TR for me and 0.42 higher than TLS which was a crazy swing.

If your ratings are again 0.20 higher than TR I may be bumped to 4.0 this year ... which is not in my plans exactly just yet.
But was I right? ;-)
 

OnTheLine

Hall of Fame
But was I right? ;-)
You were spot on not only with me, but with every person on the team. You predicted correctly the 2 bump ups as well as the 3 bump downs.

If you wanted to try your hand again .... I have two teams where I think a full 50% will be looking at a bump and where none had post-season play (or tournaments) to muddy the waters ...
 

chatt_town

Hall of Fame
There are 2 sites calculating NTRP ratings currently and I am seeing different rating results for the same player. Which site's numbers do you think is more accurate?
My question is if everyone is putting their scores into USTA site, Where else are these other sites getting their data? The next question is if they are getting it from USTA why would USTA allow them to download their data if they are calculating differently? Maybe someone has an answer to this.
 

a-a-ron

New User
My question is if everyone is putting their scores into USTA site, Where else are these other sites getting their data? The next question is if they are getting it from USTA why would USTA allow them to download their data if they are calculating differently? Maybe someone has an answer to this.
Best case for USTA is TLS, TR and Schmeke are scraping the data from the web site. A simple script could easily harvest this data. Once you have it, apply your guess at the calcualations. Worse, USTA sells it...tho I doubt that. In either case, USTA is doing little to prevent it.

I have paid Schmeke twice as I was hoping I was getting bumped two different years. He was wrong both times but the info was interesting I guess. The Sites were as well, but those are free and provide lots of fodder for those who wish to fret over their numbers.
 

chatt_town

Hall of Fame
Best case for USTA is TLS, TR and Schmeke are scraping the data from the web site. A simple script could easily harvest this data. Once you have it, apply your guess at the calcualations. Worse, USTA sells it...tho I doubt that. In either case, USTA is doing little to prevent it.

I have paid Schmeke twice as I was hoping I was getting bumped two different years. He was wrong both times but the info was interesting I guess. The Sites were as well, but those are free and provide lots of fodder for those who wish to fret over their numbers.
Thanks for the reply. I didn't know you could do that legally. :) Harvest data from a site like that that is. I guess if it's there for the public it's free game. This is interesting.
 

a-a-ron

New User
Thanks for the reply. I didn't know you could do that legally. :) Harvest data from a site like that that is. I guess if it's there for the public it's free game. This is interesting.
I'm sure it violates terms but it appears USTA does little to prevent it.
 

a-a-ron

New User
So, I took a few minutes and grabbed the data from my section, 744 have a calculated dynamic rating in both TLS and TR.

This table shows the variation between the two sites. Meaning, TLS calculates 3.9100 and TR calculates 4.0351, the sites vary by 0.1251 and thus fall in the 0.1001-0.1500 range.

Rating Type0.0000-0.05000.0501-0.10000.1001-0.15000.1501-0.20000.2001-0.35000.2501-0.30000.3001-0.35000.3501-0.40000.4001+
A
4​
2​
2​
2​
3​
0​
0​
2​
0​
C
188​
162​
130​
86​
55​
42​
21​
18​
17​
S
8​
0​
1​
0​
0​
0​
1​
0​
0​

TLS has 19 people bumping up and 134 people down. TR has 17 up and 104 down. Of these results both sites only agree on 2 up and 50 down.

Once Nationals are complete, I'll run the same thing again out of curiosity.
 

travlerajm

G.O.A.T.
The more I know about UTR, the more obvious it is that the algorithm needs serious revision to stabilize ratings.

With the current UTR algorithm, if I were to beat a 12-year-old munchkin, and 5 years later that kid grows up to win Wimbledon. My UTR would get dragged up and say I’m as good as Roger.
They really need to address this if they want to be taken more seriously by everyone. They need to go to an ELO algorithm that takes a snapshot of opponent strength at the time the match was played, and then allows that a player’s level might go up or down over time.
 
Top