Share your WTN - Crowd sourcing the NTRP to WTN mapping

[BallBag]

I was looking at this chart way too long thinking why is there so many female 6.0s.
Looks like the peaks are separated by 3-4 WTN levels. So that's the separation for an expected double bagel beat down. It's 2.5 for UTR so I guess it kinda scales with the full range.
@schmke are you able to mine UTR ratings?

 

[travlerajm]

I was looking at this chart way too long thinking why is there so many female 6.0s.
Looks like the peaks are separated by 3-4 WTN levels. So that's the separation for an expected double bagel beat down. It's 2.5 for UTR so I guess it kinda scales with the full range.
@schmke are you able to mine UTR ratings?

I was looking at the chart, and noticing that there are virtually zero females 5.0 and higher.

I think this explains why 5.0 women can do well in 9.0 mixed (they are rare unicorns with skillsets in the top 1 percentile of all females).

 

[schmke]

I just posted the analysis for men's doubles. Here is the chart:

A little more "normal" looking to the right than the women, but there is an odd blip of WTN 3's that stands out, including a decent number of 4.5s that are WTN 3. The range of WTNs for an NTRP level is quite large though even if you ignore the long tails. A decent number of 4.0s are WTN 12 and 13, but also 29 and 30!

Also, noted in my blog write-up is that the average male 3.0 is a 28.4 while the average female 3.0 is a 28.2. This goes against the conventional wisdom that the same level NTRP male and female would have the male roughly 0.5 NTRP better on a neutral scale. That 0.5 NTRP is one level and you'd expect a roughly 3-4 WTN difference, but here the female is rated better. At 3.5, 4.0, and 4.5 the male is higher but just a few hundredths or at the most, 1.5 for the 4.5 level.

 

[leech]

NTRP: 4.0C (male)
Singles WTN: 14.9
Doubles WTN: 15.5

Bonus stats:
TR: 3.64
TR MXD: 3.96
Schmidt report: 3.77
UTR Singles: 5.67
UTR Doubles: 6.05

 

[BallBag]

NTRP: 4.0C
Singles WTN: 14.9
Doubles WTN: 15.5

Bonus stats:
TR: 3.64
TR MXD: 3.96
UTR Singles: 5.67
UTR Doubles: 6.05

TR is doing you dirty.

 

[schmke]

I was looking at this chart way too long thinking why is there so many female 6.0s.

My goof perhaps confused things. I added 6.0 and failed to add a color so it was using purple for 6.0s and 2.5s :O They were at opposite ends of the chart so hopefully clear enough, but I've changed the 2.5 color to be distinct.

 

[travlerajm]

With my TR bonus data included:

NTRP: 4.5C
WTNs: 13.0 (all results at least 4 years ago)
WTNd: 13.1
TR men’s: 4.36 (based on doubles only)
TR mixed: 4.50 (I play primarily mixed 8.0).

 

[BallBag]

My goof perhaps confused things. I added 6.0 and failed to add a color so it was using purple for 6.0s and 2.5s :O They were at opposite ends of the chart so hopefully clear enough, but I've changed the 2.5 color to be distinct.

I'm color blind so I thought they actually look different to everyone else.

Looking at the women's chart again, it kinda looks like there's a small blip at 10. Might be an artificial ceiling that was added to the systems when they were showing a bunch of league players rated higher than current tour players. Nathan Ponwith is currently 678 ATP and WTN 3.6. Gabriella Price is women's 662 and is WTN 10. That is good company to be in even for a 5.5 league player.

In other news, @dsp9753 who beat me 4 and 1 the other day and is WTN 11.2, should be able to have a good match with Gabriella. Meanwhile I will have a good battle with WTA 1514 Katya Townsend who's rated 13.5.

 

[ChaelAZ]

From @schmke 's blog post, found this very interesting.

Back on the genders though, comparing the average WTN for males and females, we get:

• 3.0 - women 28.2 vs men 28.4
• 3.5 - women 24.8 vs men 24.6
• 4.0 - women 21.9 vs men 21.4
• 4.5 - women 18.7 vs men 17.2

 

[travlerajm]

I just posted the analysis for men's doubles. Here is the chart:

A little more "normal" looking to the right than the women, but there is an odd blip of WTN 3's that stands out, including a decent number of 4.5s that are WTN 3. The range of WTNs for an NTRP level is quite large though even if you ignore the long tails. A decent number of 4.0s are WTN 12 and 13, but also 29 and 30!

Also, noted in my blog write-up is that the average male 3.0 is a 28.4 while the average female 3.0 is a 28.2. This goes against the conventional wisdom that the same level NTRP male and female would have the male roughly 0.5 NTRP better on a neutral scale. That 0.5 NTRP is one level and you'd expect a roughly 3-4 WTN difference, but here the female is rated better. At 3.5, 4.0, and 4.5 the male is higher but just a few hundredths or at the most, 1.5 for the 4.5 level.

Do the results suggest that the 4.5C players with WTNs of 3 are a glitch in the system? Or are those 3 ratings legit?

 

[schmke]

Do the results suggest that the 4.5C players with WTNs of 3 are a glitch in the system? Or are those 3 ratings legit?

I don't think we know enough to say either way.

We know there are outliers with NTRP that, for whatever reason, are clearly better/worse than their assigned level. It is possible these players are the ones in the chart with WTN 3's or are the ones resulting in the high standard deviations and broad ranges within an NTRP level, and their WTN is more accurate. It just seems like there are an awful lot of these exceptions, and if WTN is right, it is throwing a little shade on the veracity of NTRP.

Now, like I said in my blog posts, WTN does calculate singles vs doubles ratings while NTRP doesn't. This alone is likely a significant reason for some of what we are seeing. Whether that is the key reason and it explains it all or not, I can't say.

 

[BallBag]

I don't think we know enough to say either way.

We know there are outliers with NTRP that, for whatever reason, are clearly better/worse than their assigned level. It is possible these players are the ones in the chart with WTN 3's or are the ones resulting in the high standard deviations and broad ranges within an NTRP level, and their WTN is more accurate. It just seems like there are an awful lot of these exceptions, and if WTN is right, it is throwing a little shade on the veracity of NTRP.

Now, like I said in my blog posts, WTN does calculate singles vs doubles ratings while NTRP doesn't. This alone is likely a significant reason for some of what we are seeing. Whether that is the key reason and it explains it all or not, I can't say.

I'm more inclined to think that WTN is way off base. I was looking at WTNs for pro women in the 1000 to 1500 ranges at I was seeing anything between 13 to 5. I would think that those ratings should be most reliable with the bluest checkmark.

 

[Moon Shooter]

NTRP 3.0s

TR: 2.92 adult and 3.22 mixed

Singles WTR 21.1 no blue check(edited)
Doubles WTR 26.5 blue checked (edited)
UTR for today singles 2.67
UTR for today doubles 4.16 (It was hovering around 5.00 for about a 6 months but then recently started to drop)

 

[DavidEDH]

I just call the WTN the WTF number. We now have NTRP, UTR and WTF - it seems quite silly. I have grown fond of UTR over the last few years and the correlation between the three ratings seem inconsistent. Just my thoughts.

 

[Moon Shooter]

Do we know what matches are being included in the WTN number? If they are only including USTA matches then this system will have many of the problems of the current USTA system.

Men and women almost never play each other in singles in USTA. So it should not be surprising that the women have the same WTN as men in singles. I am not sure if incorporating mixed would help even things out in doubles. Do they include mixed doubles? I think they really need to include MM v FF matches if they want to claim this is truly gender neutral.

I hope they add some transparency to this and show what matches are even included and give a performance rating for each match. Right now it seems like a number plucked out of thin air.

Schmidt Computer Ratings

computerratings.blogspot.com

 

[BallBag]

Do we know what matches are being included in the WTN number? If they are only including USTA matches then this system will have many of the problems of the current USTA system.

Men and women almost never play each other in singles in USTA. So it should not be surprising that the women have the same WTN as men in singles. I am not sure if incorporating mixed would help even things out in doubles. Do they include mixed doubles? I think they really need to include MM v FF matches if they want to claim this is truly gender neutral.

I hope they add some transparency to this and show what matches are even included and give a performance rating for each match. Right now it seems like a number plucked out of thin air.

Schmidt Computer Ratings

computerratings.blogspot.com

WTN should enable the USTA to run mixed gender events, leagues and tournaments. Actually doing so is up to your section.

 

[schmke]

Do we know what matches are being included in the WTN number? If they are only including USTA matches then this system will have many of the problems of the current USTA system.

Men and women almost never play each other in singles in USTA. So it should not be surprising that the women have the same WTN as men in singles. I am not sure if incorporating mixed would help even things out in doubles. Do they include mixed doubles? I think they really need to include MM v FF matches if they want to claim this is truly gender neutral.

I hope they add some transparency to this and show what matches are even included and give a performance rating for each match. Right now it seems like a number plucked out of thin air.

Schmidt Computer Ratings

computerratings.blogspot.com

I would assume that all USTA sanctioned events would be included, but I have no confirmation of that. And specifically for league players there are a variety of leagues that are/aren't included for NTRP calculations so does WTN include all or follow what NTRP does? I'm efforting to find out.

 

[TennisOTM]

I just posted the analysis for men's doubles. Here is the chart:

A little more "normal" looking to the right than the women, but there is an odd blip of WTN 3's that stands out, including a decent number of 4.5s that are WTN 3. The range of WTNs for an NTRP level is quite large though even if you ignore the long tails. A decent number of 4.0s are WTN 12 and 13, but also 29 and 30!

Also, noted in my blog write-up is that the average male 3.0 is a 28.4 while the average female 3.0 is a 28.2. This goes against the conventional wisdom that the same level NTRP male and female would have the male roughly 0.5 NTRP better on a neutral scale. That 0.5 NTRP is one level and you'd expect a roughly 3-4 WTN difference, but here the female is rated better. At 3.5, 4.0, and 4.5 the male is higher but just a few hundredths or at the most, 1.5 for the 4.5 level.

Very cool charts! I'm surprised, though, at how much you are reserving judgment against WTN doubles based on these data. I can't see how these within-level ranges could possibly be anywhere close to correct. They are basically saying that a healthy portion of 4.5C players, for example, are good enough at doubles that they could partner with a barely-played beginner and defeat many teams of two 4.5C players.

E.g. a WTN 10 could partner with a WTN 36 and be favored against any teams of two WTN 23+ players (of which there are many among 4.5C guys). And that's not even going very close to the far tails of the 4.5C distribution.

I don't think it's too early to conclude that many of the WTN doubles numbers are nonsense. I'm willing to believe that NTRP gets some players "wrong" because of averaging singles and doubles or other reasons, but could those flaws really be this frequent and extreme? I can't believe it, unless there's something I'm missing?

 

[schmke]

Very cool charts! I'm surprised, though, at how much you are reserving judgment against WTN doubles based on these data. I can't see how these within-level ranges could possibly be anywhere close to correct. They are basically saying that a healthy portion of 4.5C players, for example, are good enough at doubles that they could partner with a barely-played beginner and defeat many teams of two 4.5C players.

E.g. a WTN 10 could partner with a WTN 36 and be favored against any teams of two WTN 23+ players (of which there are many among 4.5C guys). And that's not even going very close to the far tails of the 4.5C distribution.

I don't think it's too early to conclude that many of the WTN doubles numbers are nonsense. I'm willing to believe that NTRP gets some players "wrong" because of averaging singles and doubles or other reasons, but could those flaws really be this frequent and extreme? I can't believe it, unless there's something I'm missing?

I've been critical of things the USTA does, and I'll continue to be where warranted. But I am by no means a USTA hater with a vendetta to make them look bad or anything, and so with an initial release of WTN not even being a week old, I'll cut them a little slack for the time being. I've noted observations where things seem unexpected, I'm just not prepared to say WTN is wrong across the board, even if anecdotally there seem to be some very clear gaffes based on personal experience. Who is to say WTN isn't right and NTRP is wrong? I don't think that is the case mind you, just not wanting to prematurely condemn WTN.

But you raise an interesting point, in singles at least, players have a Game zONe which is their rating +/- ~2 that is supposed to provide a competitive match. If WTN is right, they have players, who by being at the same NTRP level are supposed to be "compatible" being 15-20 WTN levels apart. So is WTN saying these players aren't compatible and NTRP's claim they are is false?

 

[BallBag]

I wouldn't fault the USTA for WTN being off. WTN is an ITF program with the USTA providing all their data to the ITF. The ITF then decides which matches are included and how they affect your rating.

I don't see how WTN can correctly rate two players 20 levels apart that are in the same NTRP. Even if you take an edge case of a singles only player vs. doubles only player, a difference of 20 WTN means there are 4 levels of 0 and 0 beat downs between them.

 

[socallefty]

From @schmke 's blog post, found this very interesting.

• 3.0 - women 28.2 vs men 28.4
• 3.5 - women 24.8 vs men 24.6
• 4.0 - women 21.9 vs men 21.4
• 4.5 - women 18.7 vs men 17.2

The USTA says that men’s NTRP ratings are not the same as for women with the men being better by at least 0.5 rating compared to a woman with the same rating. Most players who play mixed have probably experienced this that a 4.5 man is way better than a 4.5 woman. WTN seems to be completely wrong as a result if men and women with similar NTRP ratings have the same average WTN.

It is hard to have an universal rating where men and women don’t play each other much in real life or adults and juniors don’t play each other much. In that case, you have to calibrate ratings of adults/juniors, men/women etc. based on more theoretical models rather than an actual database of real matches played by players in different categories. UTR has the same problem, but they seem to deal with it better than this first experience with WTN.

Also, it is tough to extend universal ratings across different international regions when the players from different countries never play each other - except pros and a few age-group world championship tournaments.

 

[Moon Shooter]

Just a few somewhat random observations:

I am finding some players on my team (2 out 6 checked) that have played quite a few USTA mixed as well as adult doubles matches with no WTN rating at all. So it maybe the database is still filling up.


I found this:
"If you've played with the USTA within the last 5+ years, either singles or doubles, you have an ITF World Tennis Number. Click here to find your profile. "

ITF World Tennis Number

The new ITF World Tennis Number is free to use and is being adopted by over 135 countries, including the U.S. Explore all that the ITF WTN has to offer and get some of your most frequently asked questions answered.

www.usta.com

"The ITF World Tennis Number algorithm uses match result data as provided by the United States Tennis Association (USTA) to calculate your ITF World Tennis Number. Results provided by the USTA from as far back as 2016 can be used in calculating your rating."

So it seems they will go back about 6 years but only use USTA results. So we will never actually get a true gender neutral rating from this unless USTA starts to allow men and women to regularly play each other in singles and allows more variations on doubles. At least all FFvMF etc.

Also how can they claim this is a truly worldwide rating when they only use national numbers?? Are other countries' teams going to be invited to surprise Arizona?


"How does WTN rating differ from a USTA Ranking?

Match wins make up your USTA ranking. Ranking points are awarded by either points per round or points per win, depending on the tournament level. Match scores make up your WTN rating. Match scores are analyzed at set level. The algorithm takes into account each individual set as its own result."

https://customercare.usta.com/hc/en-us/articles/6006524686228

They say they look at sets won as opposed to games won but I believe I recall a youtube video where they do consider the score of the set - that is games won.

Honestly the last thing tennis needs is another mystery number based insufficient data. ITF will at least look at Mixed results but it is unclear if this will help the accuracy so much (if at all) since the men's and women's rating systems will remain fairly separate.

The USTA profile page also says:
"Player results by rank list are available beginning the week of September 20th."
But I am not sure if that has anything to do with ITF or if that is just with their wonky tournament rankings thing.

 

[Moon Shooter]

Who is to say WTN isn't right and NTRP is wrong? I don't think that is the case mind you, just not wanting to prematurely condemn WTN.

Which number will be more accurate will by and large depend on which number considers the most relevant data. I think many people on this forum live in areas where people play many USTA matches. I think FYB said he plays about 80 rated USTA matches per year. If you figure the average number of rated per USTA player, is about 8 per year. Then the median player likely only plays about 4 matches per year. So the median player is going to have a very inaccurate NTRP.

Will adding mixed ratings help the accuracy? It probably will for the median player.

Its annoying that these faqs never answer real questions. Like how do you expect the number to be predictive across gender and national lines when you only include matches with the same nationality and gender on opposite sides of the court?

 

[travlerajm]

Which number will be more accurate will by and large depend on which number considers the most relevant data. I think many people on this forum live in areas where people play many USTA matches. I think FYB said he plays about 80 rated USTA matches per year. If you figure the average number of rated per USTA player, is about 8 per year. Then the median player likely only plays about 4 matches per year. So the median player is going to have a very inaccurate NTRP.

Will adding mixed ratings help the accuracy? It probably will for the median player.

Its annoying that these faqs never answer real questions. Like how do you expect the number to be predictive across gender and national lines when you only include matches with the same nationality and gender on opposite sides of the court?

I think a good algorithm should be able to spit out a very accurate rating even if the player only plays 4 matches per year. That’s 20 matches over a 5-year period.

And each match has a lot more data than just W-L. It also includes the % of games won for each match, and the rating for each opponent. So that’s a pretty robust dataset to work with.

 

[JLyon]

WTNS: 10.4 (way off, I should be closer to 15-20 range, considering the #1 seed at Southern 16's is only a 13 and he would beat me 0,0)
WTND: 15.5
NTRP 4.5C
UTRD: 6.85
UTRS: 7.34

 

[FuzzyYellowBalls]

I think a good algorithm should be able to spit out a very accurate rating even if the player only plays 4 matches per year. That’s 20 matches over a 5-year period.

And each match has a lot more data than just W-L. It also includes the % of games won for each match, and the rating for each opponent. So that’s a pretty robust dataset to work with.

Ahaha, you need to create your own Travlrajm rating system then, let's just imagine there is a huge group of people that only play tennis matches 4 times each year, would they even care about a rating system? They wouldn't be in a league, play any tournaments or anything official so would they even be that concerned about ratings?

 

[Moon Shooter]

I think a good algorithm should be able to spit out a very accurate rating even if the player only plays 4 matches per year. That’s 20 matches over a 5-year period.

Sort of but not quite. It is 4 matches per year and not everyone will play for 5 years. Many will only play one year. And if that is the median player there will be half of the players under that number per year.

But anyway I was just throwing out some numbers I'm not sure if they are accurate. So lets not get stuck on 4 matches, as I am not claiming it is accurate. My point was just to illustrate how very different the amount of data each area might have and therefore how different the accuracy of these NTRP ratings might be depending on where you live.

And each match has a lot more data than just W-L. It also includes the % of games won for each match, and the rating for each opponent. So that’s a pretty robust dataset to work with.

Yes you have the games won and lost to work with. But

1) But we know scores can end up being lopsided when the games are all going to deuce. And we also know the same two people can have lopsided back to back sets. and
2) when you add a partner for doubles especially when that partner may have a completely different rating system things are not so easy.

I think what happens is you have some areas with people that have well established ratings and the current NTRP is fine. But then several/?most? other areas where the ratings are a mess. The messy areas have people all over the board but since they don't have a huge number of players right at the top of each level they tend not to have teams that are going to nationals or even sectionals so these wrinkles never get ironed out.

 

[BallBag]

I think a good algorithm should be able to spit out a very accurate rating even if the player only plays 4 matches per year. That’s 20 matches over a 5-year period.

And each match has a lot more data than just W-L. It also includes the % of games won for each match, and the rating for each opponent. So that’s a pretty robust dataset to work with.

WTN only considers sets won vs. lost. I'm not sure why.

Which number will be more accurate will by and large depend on which number considers the most relevant data. I think many people on this forum live in areas where people play many USTA matches. I think FYB said he plays about 80 rated USTA matches per year. If you figure the average number of rated per USTA player, is about 8 per year. Then the median player likely only plays about 4 matches per year. So the median player is going to have a very inaccurate NTRP.

Will adding mixed ratings help the accuracy? It probably will for the median player.

Its annoying that these faqs never answer real questions. Like how do you expect the number to be predictive across gender and national lines when you only include matches with the same nationality and gender on opposite sides of the court?

Both UTR and WTN add a weighting component based on the opponents rating reliability. So if a player does better than expected vs FYB, his rating would increase more than if he had the same result against a player with the same rating but lower match count. This is the main improvement of the WTN algorithm (Glicko-2) vs ELO.

 

[J_R_B]

4.0 C (M)
WTNs 15.0 (16.8-13.2 Game Zone, grey check)
WTNd 18.2 (blue check)

In my current state of being old, fat, and slow with bad knees, I am definitely NOT better at singles. LOL.

 

[travlerajm]

WTN only considers sets won vs. lost. I'm not sure why.


Both UTR and WTN add a weighting component based on the opponents rating reliability. So if a player does better than expected vs FYB, his rating would increase more than if he had the same result against a player with the same rating but lower match count. This is the main improvement of the WTN algorithm (Glicko-2) vs ELO.

Considering only sets and not games is leaving a lot of useful info on the table.

On the other hand, usta supposedly uses games won %, meaning a 10-point super gets weighted the same as other games. This might bias results. I’d think perhaps weighting the super 2.5x of the other games would be roughly right, since the super is 2.5x as many points on average.

 

[Le Renard]

Same with me— tennislink shows my league matches but the USTA site that lists my WTN doesn’t display my matches.

Same here for NY, is it a major data load issue where they would have forgotten to load all results to calculate the WTN? I have been investigating for 3 times in a week now with them, they say that all is in but can't provide the list of games logged to calculate the WTN... very odd. I am attacking now with the angle of double since I have one game recorded for the USTA league in May 2022 but my double WTN is empty...

 

[BallBag]

Same here for NY, is it a major data load issue where they would have forgotten to load all results to calculate the WTN? I have been investigating for 3 times in a week now with them, they say that all is in but can't provide the list of games logged to calculate the WTN... very odd. I am attacking now with the angle of double since I have one game recorded for the USTA league in May 2022 but my double WTN is empty...

The USTA site only shows tournament results, not league. WTN does not list all the matches that it looks at for your rating at this time. How many doubles league matches do you have?

 

[Moveforwardalways]

Has anyone looked at how WTN maps on to UTR?

 

[Le Renard]

The USTA site only shows tournament results, not league. WTN does not list all the matches that it looks at for your rating at this time. How many doubles league matches do you have?

I contacted them and they say that all the games were counted, even USTA league games, not only tournaments. And it would not make any sense not to count the league games, would be a major flaw. I played one double.

 

[BallBag]

I contacted them and they say that all the games were counted, even USTA league games, not only tournaments. And it would not make any sense not to count the league games, would be a major flaw. I played one double.

The USTA site does not calculate WTN. The USTA sends its data to the ITF and the ITF calculates WTN ratings and then sends it back to the USTA to show on its website. The USTA website that has your rating only has results from tournaments. Tennislink has your results from league.

I don't know how many results are needed to generate a rating but if you only have one doubles game and no rating then its more than one.

 

[schmke]

The USTA site does not calculate WTN. The USTA sends its data to the ITF and the ITF calculates WTN ratings and then sends it back to the USTA to show on its website. The USTA website that has your rating only has results from tournaments. Tennislink has your results from league.

I don't know how many results are needed to generate a rating but if you only have one doubles game and no rating then its more than one.

I don't know what the minimum is for WTN to calculate a rating, but it is important to remember that WTN calculates separate singles and doubles ratings. So while NTRP might have say 5 matches to calculate a rating from, in that scenario WTN has perhaps 2 singles and 3 doubles matches which may preclude one of the ratings being calculated or a lower confidence on both.

 

[Idaho MEP]

4.0T (now 4.0S because I joined a league)
25.8 singles (grey)
23.5 doubles (blue)

I don't know what the minimum is for WTN to calculate a rating, but it is important to remember that WTN calculates separate singles and doubles ratings. So while NTRP might have say 5 matches to calculate a rating from, in that scenario WTN has perhaps 2 singles and 3 doubles matches which may preclude one of the ratings being calculated or a lower confidence on both.

They calculated a singles rating for me based on just 2 matches (but grey check mark).

In those two singles matches, I beat a guy who's now 19.9, and got destroyed by a guy who's now 14.0. (I just got my 3rd singles match in, and beat a guy with a 15.2, but they haven't factored that in yet.)

 

[BallBag]

4.0T (now 4.0S because I joined a league)
25.8 singles (grey)
23.5 doubles (blue)


They calculated a singles rating for me based on just 2 matches (but grey check mark).

In those two singles matches, I beat a guy who's now 19.9, and got destroyed by a guy who's now 14.0. (I just got my 3rd singles match in, and beat a guy with a 15.2, but they haven't factored that in yet.)

WTN updates every Wednesday.

 

[schmke]

Wrote about women's singles. Here is the chart:

 

[schmke]

And also wrote on the men singles:

 

[BallBag]

Looking at the verified charts, the overall means for men(18) and women (21) are separated by 3 levels. This is consistent with men's NTRP and women's NTRP being a level off.

The standard distribution looks too wide. I know WTN just went live but it has years worth of data so I don't see the accuracy getting any better with time.

 

[time_fly]

I was doing some scouting for upcoming mixed doubles districts, and I have found a 4.0 woman with a blue-check doubles WTN of 5.2. Lol. She has had multiple undefeated visits to Nationals but has never been bumped up. It honestly doesn't look like she throws games or matches either; most of her wins are one-sided. It's stuff like that which makes NTRP seem broken.

 

[schmke]

Looking at the verified charts, the overall means for men(18) and women (21) are separated by 3 levels. This is consistent with men's NTRP and women's NTRP being a level off.

The standard distribution looks too wide. I know WTN just went live but it has years worth of data so I don't see the accuracy getting any better with time.

The singles distribution is a bit better than doubles, the standard deviations dropped and just looking at the charts they aren't as wide for a given level, and there isn't as much overlap of NTRP levels at a given WTN, but I agree it seems wider than you'd expect considering they say they are using data from 2016 and should have adequate matches to base the ratings on.

 

[Moon Shooter]

WTN only considers sets won vs. lost. I'm not sure why.

I read it that way too. But in the video in this post they clarify the score within a set matters.

ITF ups the ante in the "global" ratings game

I just wrote on my blog about the ITF announcing the World Tennis Number (WTN) which seems to be their entry in the ratings game to go head to head with UTR. To be fair, the ITF created the International Tennis Number well over 10 years ago so pre-dated UTR, but had clearly not invested in...

tt.tennis-warehouse.com

I also wonder if people will be able to post non-usta match score results into itf. If they could that would solve many of the problems regarding ratings between genders and ratings involving people outside the US as well as just ratings suffering due to insufficient data. The problem is that video is unclear what they will allow and whether the scores will count toward your rating.

WTN could be a better version of UTR since they will not ignore matches more than 12 months old. However if they only include USTA events and don't allow people to post their private matches the amount of data will still be a problem for many areas.

 

[BallBag]

I was doing some scouting for upcoming mixed doubles districts, and I have found a 4.0 woman with a blue-check doubles WTN of 5.2. Lol. She has had multiple undefeated visits to Nationals but has never been bumped up. It honestly doesn't look like she throws games or matches either; most of her wins are one-sided. It's stuff like that which makes NTRP seem broken.

A verified WTN of 5.2 is top 100 WTA. NTRP has problems but her WTN is way off.

I read it that way too. But in the video in this post they clarify the score within a set matters.

ITF ups the ante in the "global" ratings game

I just wrote on my blog about the ITF announcing the World Tennis Number (WTN) which seems to be their entry in the ratings game to go head to head with UTR. To be fair, the ITF created the International Tennis Number well over 10 years ago so pre-dated UTR, but had clearly not invested in...

tt.tennis-warehouse.com

I also wonder if people will be able to post non-usta match score results into itf. If they could that would solve many of the problems regarding ratings between genders and ratings involving people outside the US as well as just ratings suffering due to insufficient data. The problem is that video is unclear what they will allow and whether the scores will count toward your rating.

WTN could be a better version of UTR since they will not ignore matches more than 12 months old. However if they only include USTA events and don't allow people to post their private matches the amount of data will still be a problem for many areas.

I think Heather is wrong on this one. The ITF's FAQ page says its at the set level.
Ratings outside the US are collected by that country's sanctioning body and then sent to the ITF. So if you play a league match in the UK, the LTA will send the result to the IFT to be added to the results sent by the USTA. I don't know how the ITF know that US Moon Shooter is the same as UK Moon Shooter. They might add a "Merge Profiles" feature like UTR has.

 

[schmke]

A verified WTN of 5.2 is top 100 WTA. NTRP has problems but her WTN is way off.

I think Heather is wrong on this one. The ITF's FAQ page says its at the set level.
Ratings outside the US are collected by that country's sanctioning body and then sent to the ITF. So if you play a league match in the UK, the LTA will send the result to the IFT to be added to the results sent by the USTA. I don't know how the ITF know that US Moon Shooter is the same as UK Moon Shooter. They might add a "Merge Profiles" feature like UTR has.

It is a master data management problem (the space my day job has me operating in) and yeah, I'm not sure how the ITF is going to handle it. Clearly they need to, but do they have something to detect and merge duplicates? Or will they eventually get around to having more of a site a player can interact with and update their profile including identifying duplicate profiles? Time will tell.

 

[BallBag]

It is a master data management problem (the space my day job has me operating in) and yeah, I'm not sure how the ITF is going to handle it. Clearly they need to, but do they have something to detect and merge duplicates? Or will they eventually get around to having more of a site a player can interact with and update their profile including identifying duplicate profiles? Time will tell.

They were saying something about the ITF having a common profile. My USTA profile is showing an ITF ID and that's the same as my ITF IPIN. My LTA profile doesn't show my ITF ID and there is no way to put that in. It also has my "self rated" WTN so its not connected to my ITF profile. ITF has my email so I'm guessing that's how it got linked to my USTA profile. The LTA does not have that technology.

 

[schmke]

They were saying something about the ITF having a common profile. My USTA profile is showing an ITF ID and that's the same as my ITF IPIN. My LTA profile doesn't show my ITF ID and there is no way to put that in. It also has my "self rated" WTN so its not connected to my ITF profile. ITF has my email so I'm guessing that's how it got linked to my USTA profile. The LTA does not have that technology.

And thus there are two of you right now ...

 

[Moon Shooter]

A verified WTN of 5.2 is top 100 WTA. NTRP has problems but her WTN is way off.

I think Heather is wrong on this one. The ITF's FAQ page says its at the set level.
Ratings outside the US are collected by that country's sanctioning body and then sent to the ITF. So if you play a league match in the UK, the LTA will send the result to the IFT to be added to the results sent by the USTA. I don't know how the ITF know that US Moon Shooter is the same as UK Moon Shooter. They might add a "Merge Profiles" feature like UTR has.

I give Heather more credit than that. It would be truly bizarre if she did not understand even such a basic aspect of the algorithm.

Edit: also it would seem very dumb to ignore the score of each set.

Are you referring to this:

"Match results are analysed at set level, meaning our algorithm takes into account each individual set as its own result. Simply, if your match ends 2 sets to 1 in your favour, then the system will update your Number with two set ‘wins’ and one set ‘loss’. Even if you don’t win the overall match any sets you have won will count towards your Number."

I agree I would read this the same way you did without clarification from Heather.

But I think the quoted language is clumsily distinguishing between a system that looks at a "match" as a complete unit and their looking at a set as a complete unity. If you play one set and then get injured in the second set and have to withdraw perhaps some systems may not count the match at all in the ratings. But WTN would presumably count the result of the first set as if it were a "match" itself.


BTW this video goes through how you can set up WTN tournaments through the USTA site and it really looks amazing and easy to use! It includes "coed" tournaments in addition to "mixed" so we should be able to get the data we need to greatly improve the ratings.

Edit2: I think USTA is still adding data to the WTN ratings so I anticipate the WTN ratings will be out of whack for a while.

 

[schmke]

"Match results are analysed at set level, meaning our algorithm takes into account each individual set as its own result. Simply, if your match ends 2 sets to 1 in your favour, then the system will update your Number with two set ‘wins’ and one set ‘loss’. Even if you don’t win the overall match any sets you have won will count towards your Number."

I agree I would read this the same way you did without clarification from Heather.

But I think the quoted language is clumsily distinguishing between a system that looks at a "match" as a complete unit and their looking at a set as a complete unity. If you play one set and then get injured in the second set and have to withdraw perhaps some systems may not count the match at all in the ratings. But WTN would presumably count the result of the first set as if it were a "match" itself.

I agree the language is a bit ambiguous and doesn't necessarily rule out using the score.

Edit2: I think USTA is still adding data to the WTN ratings so I anticipate the WTN ratings will be out of whack for a while.

It is the USTA, so you may be right, but they announced this 3 years ago and certainly have the matches in their system, so why on earth would they still be adding data? They already delayed the launch repeatedly, why launch prematurely and just open themselves up to criticism due to WTN being out of whack?