The unintentional irony of you accusing me of a lack of understanding is hilarious. You're aggravating because you say the same thing over and over and over even though everyone on this board who actually does understand the rating system pushes back against it because it's downright stupid and won't work. Yet, you just ignore everyone and repeat the same stupid crap. It was stupid the first time you said it, the second time you said the same thing, and it's still stupid the 357th time. It's really aggravating to see you vomiting the same nonsense all over the board over and over and just ignoring what everyone who's smarter than you is saying.
The dreaded EOY Rating thread
- Thread starter Brilliant
- Start date
Moon Shooter
Hall of Fame
Another post from you with no substance just hate and asserting I’m wrong. And of course it gets a like from my pal Cretin.
This is the Adult League and tournament talk forum. So the vast majority of this tiny community is going to participate in usta and therefore likely not agree that it is deeply flawed. It’s like arguing there are deep flaws with the Catholic Church with a bunch of Catholic priests or arguing there are deep flaws with trump or Biden with people that voted for them. (I am Catholic btw so I’m not just picking on them) I am under no illusion that my views will not be popular with the few people in this forum that represent the small minority of people that think usta is worth it. I understand how group dynamics work and how people just want to be liked.
But I have plenty of friends and don’t need to make friends with a bunch of people pearl clutching over forum likes. I am not here to make friends but to make points. I have been interested in rating systems for while and there are people that recognize I know what I am talking about even if they ultimately disagree with me and even if you and creighton and a few others feel the need to loudly and ignorantly denounce me as some sort of witch. Why not try stick to making points? Because I really have repeatedly made it clear that I don’t care how often you try to attack me personally and just assert I’m wrong. Your opinion is not worth anything to, me unless you have a good reason for it.
Moon Shooter
Hall of Fame
Yeah I know it sucks and I’m sorry it happened in this thread. But there are a couple of posters (assuming they are different people) that seem to do nothing except follow me around and attack me every time I say anything. I gave the long response. The better, and shorter, response is FO.
Moon Shooter
Hall of Fame
.
silverwyvern4
Semi-Pro
TR is free?
Well then if they don't update soon, I'm going to start using an ad blocker whenever I visit next year.
ChaelAZ
G.O.A.T.
I was an admin and served as a moderator on a few different (very high profile) forums for politics and religion discussion. I can tell you, when I can't find decent threads with vastly biased and unrelenting opinions, I bang my head on a brick wall and can have the same kind of satisfaction. I find some of the same in tennis talk (Nadal is GOAT, sandbaggin doesn't happen as much as people think, and the Blade v7 is EXACTLY the same as the v8 no matter what you think you feel), best truck debates (because we KNOW GMC wins every time), or scotch (Dalwhinnie 25year, nuff said). So I'll just agree that you should continue on and have a great point, but @J_R_B is also right.
And in closing, thanks Obama.
FuzzyYellowBalls
Legend
What crime did you commit to receive this punishment?
ChaelAZ
G.O.A.T.
lol. I was younger and idealistic, along with a long history of study in religion and philosophy. And I had time and thought it made me someone important in a way.
Moon Shooter
Hall of Fame
Yes I have also studied philosophy religion and politics. But tend not to participate in forums discussions anymore on those topics. I have a blog and will discuss issues on other blogs though. Of course as a trial lawyer I argue with people quite a bit. It is from the latter that helps me understand when people are presenting reasons that will work with unbiased judges, as opposed to just posts that are simply working the tribalism of forums or blogs - that typically have their own biases. There is a difference.
We can agree to disagree on JRB.
We can agree to disagree on JRB.
What's the point of debating with someone who has said the same obviously stupid thing 357 times? You're not going to change and what you've droned about endlessly isn't going to suddenly be less stupid, either.
Purestriker
Legend
Yes, the T record is free. UTR is also free, you just cannot see the the decimal point after the level unless you pay for the pro version (Ex. 4 vs. 4.13).
silverwyvern4
Semi-Pro
did TR get a new logo on their website?
E.T.
Rookie
It looks like it. Must be updating the site format/algorithms before scraping for new matches.
Seems like they did a much better job this year than last year.
cks
Hall of Fame
Agreed. New logo.
I checked my team that just finished the fall season to see if they adjusted any of the ratings based on the end-of-year NTRP ratings released on 12/1/2022.
1. One guy, who had a TR rating above 3.50 and did not get a NTRP bump from 3.5 to 4.0 has been reset to TR rating of 3.25
2. Two guys, who had a TR ratings above 3.00 and did not get a NTRP bump from 3.0 to 3.5 have been reset to TR rating of 3.00.
They did this with all the ratings. Any player who got bumped up (or down) had their rating adjusted. If you got bumped from NTRP 4.0 to NTRP 4.5 and your year end TR rating was below 4.01, you got reset to 4.01. In a previous comment of mine an NTRP 4.0s with a year end TR 3.80 got bumped up to 4.5 and his new adjusted TR rating is now 4.01
Also interesting that a few of my 4.0 teammates who were previously showing TR ratings in the high 3.8's or low 3.9's are now showing TR ratings in 3.75-3.80 range (including myself) In my case I can see It as I tried to play through some injuries and lost a few winnable matches. If TR is trying to replicate NTRP rating, then they have a long way to go
time_fly
Hall of Fame
TR usually goes dark around this time of year, I assume to recalibrate based on new ratings and make improvements. I’m more surprised that UTR also stopped updating and I wonder if USTA is hindering them in order to help promote the new ITF rankings.
We had a guy in our league who had self rated at 4.0 get bumped down to 3.5. Not a big deal you might say but his rating immediately went from 4.38 (highest on a team of 20 where maybe 8 got bumped up) down to 3.51. Not much left to do but shake your head.
Purestriker
Legend
Good news…TR is not the USTA algorithm. So it doesn’t matter.
Are you sure about those figures? A 4.38 rating is 4.5 territory. Also, 3.51 is still 4.0 level.
Correction 3.50 rating after being bumped down.
That is true, it technically is not the same algorithm the usta uses. How could they possibly be that different? .88 is significant miscalculation or roughly a 20% rating error. Not really sure how that’s possible. We’re not talking getting bumped from 3.65 to 3.50. Nobody would question that. If you look at this particular players matches, his lowest rated match is 4.00.
The hardest part is probably figuring out how the USTA does the initial rating. I can't even imagine how Schmke figured it out.
So the TR system probably gave the credit in the matches to the S rate in question, while the actual formula gave credit to the partner.
OnTheLine
Hall of Fame
And even messier with an S rate plays multiple matches with or against another S rate, or with/against a player with few match results ... that is one big wild card to try and compute.
Purestriker
Legend
I have found they are different. Folks that they had well above the close range of getting bumped and those individuals did not get bumped. Fortunately, that result favored my team as we were thinking that player was gone.
The interesting question here is which algorithm was more accurate? You have TR calling him a high level 4.5 and NTRP bumping him down to 3.5. Have you played with or against him? Is he closer to 3.5 or 4.5?
I have not played him but watched him play a few games at playoffs. He played singles and retired after splitting sets (since his team won their match) with a guy that was a 4.5 bump. In my experience, 3.5 and 4.5 singles should be light years apart. Obviously that’s one match but TR shows a pattern.
cks
Hall of Fame
You are right it does not matter. However, I do fine it interesting on how the sites (Tennis Record, Tennis League Stats) try to reverse engineer the NTRP rating algorithm.
The retirement is a red flag. He likely knows that retirements don’t count toward your rating and very well may have manipulated scores in his other matches as well.
Moon Shooter
Hall of Fame
I agree with JRB the interesting question is generally which rating is more accurate. But TR seems more concerned with trying to mirror USTA than actually trying to deliver accuracy. This is shown when they change everyone's ratings to match what USTA says at year end. E.g., If TR incorrectly thought there would be a bump TR just changes your ratings so you are within the limit for USTA's level.
Heather from USTA said the rating system is 75% accurate. It is unclear what she meant by that. Did she just mean that USTA accurately picks which side won 75% of the time? Flipping a coin will get you to 50% and since there are so many clear mismatches I would think at least 85% would be the minimum. UTR's 12 month rule is a problem but for people that play enough matches that the 12 month rule is not an issue, I would think they are hitting at least 85% in their predictions.
Heather from USTA said the rating system is 75% accurate. It is unclear what she meant by that. Did she just mean that USTA accurately picks which side won 75% of the time? Flipping a coin will get you to 50% and since there are so many clear mismatches I would think at least 85% would be the minimum. UTR's 12 month rule is a problem but for people that play enough matches that the 12 month rule is not an issue, I would think they are hitting at least 85% in their predictions.
It's my understanding that partial scores from a retirement match DO count for ratings if at least one set was completed. If this guy took a set off a year-end 4.5 guy in singles, it's extremely fishy that he would get rated 3.5C. Maybe this is a case where USTA is the one that erred and not TR. How many rating-eligible singles matches did he play, and how many doubles?
OnTheLine
Hall of Fame
That is not accurate .... a retirement counts if at least 6 games were played. If he split sets, at least 12 games were played: that match should have counted.
From the FAQ on NTRP from USTA:
Valid Match: A team match in which a majority of the individual matches, played by the competing teams, have completed at least six games.
I’m sure that match will count towards next years ratings since fall playoffs happen after ratings come out on dec 1. The team is full of guys that were brought in well below their true rating for one purpose. How have they stayed off ratings guys radar?
Not to drop names but we’ve had a private discussion on this and his numbers weren’t far off from TR.
You're going to have to spill the beans when they get to nationals.
75% accuracy for picking the winner seems quite good to me. The very best predictors for professional tennis matches barely get above 70%, see here:
https://doi.org/10.1515/jqas-2015-0059
Getting to 85% accuracy would make you rich in the betting markets.
Now you might argue that predicting USTA league matches should be easier because of higher frequency of clear mismatches compared to pro tournaments. Possibly, but on the other hand the above article also showed that even when you limit to matches of top-30 players (who benefit from many mismatches over a year), the accuracy of the best predictors still only gets to about 75%.
I'd also argue that USTA has a harder job with predicting league matches compared to pro matches in some ways: first, the pro match predictors have a lot more detailed and accurate data about the players and their results to work with. Second, the above pro match prediction results are based only on singles, whereas I'm assuming USTA's 75% result is based mostly on predicting doubles matches, which has got to be a harder job with more pre-match player estimates involved.
I think you’re right, I confused this with the rule that retirements don’t generate strikes toward dynamic disqualification:
I would use UTRs rating system to argue that you have players from a much more similar playing ability competing in pro tournaments. So you’re talking comparing a 15.5 to a 15.3 or a 14.9.
Just as an example, 4.0 tennis is usually a much wider range. I believe it’s 4.5 to 6.8 utr. That would seem to me to make match predictions easier in theory.
Moon Shooter
Hall of Fame
Interesting article.
Only one of the systems was an Elo based system FiveThirtyEight and it scored the highest. I am not sure what the other systems were doing nor do I really care. There are so many silly tennis rating systems that it is hard to take most of them seriously.
You are correct, I would have thought the lower rated players would be easier to predict with an elo system. I would think that trying to predict a grand slam is like trying to predict nationals. But your typical run of the league match you will often have one player playing out of level or at very far parts of each level which should really help you boost the rating higher than the 50% flipping a coin will get you.
It was surprising to me that the 75% were for the most highly ranked players and the lower rated players had less accuracy.
"Using only 1 year of prior performance data, regression models based on player ranking and an Elo approach developed by FiveThirtyEight were the most accurate approaches. The FiveThirtyEight model predictions had an accuracy of 75% for matches of the most highly-ranked players, which was competitive with the bookmakers. The inclusion of career-to-date improved the FiveThirtyEight model predictions for lower-ranked players (from 59% to 64%) but did not change the performance for higher-ranked players."
But I think part of the problem could be the lower rated players matches are not part of fivethirtyeight's rating system. If someone is new to the pro circuit fivethirtyeight may not have ratings for their non-professional matches or the matches of the people they play. So I would have to better understand how that works. But this makes me think the problem is they may not have had many matches for lower rated players since only Grand Slam and ATP tour matches above the challenger level were included:
"Elo ratings were computed for match results of ATP players beginning at the start of the 2013 season (1 year before the earliest matches in the validation data described in the next section) up to the date of the validation match being predicted. All Grand Slam matches and ATP Tour matches above the Challenger level were included. The BCM included the pre-match odds of 7 bookmakers reported on www.tennis-data.co.uk."
I think if they would have included all the matches played for the year before the predictions would have been better than what we saw at the highest levels.
But even so when you look at the tiny margins that can mean big differences in results at the pro level it is hard to believe that it wouldn't be easier to predict amateur tennis. I mean unless a pro player heavily relies on a tennis serve it is not like you would be able to even come close to watching video and saying this person is in a different league than some other top pro. But I know different 3.0 men's players - all c rated and you can definitely see the difference in ability between several of them.
Then you have the issue of people playing out of their level. (or doubles teams leaving .5 on the table either officially or it is just their dynamic ratings are at the low end of the level) I mean those should also be some pretty easy wins. The rest should be no worse than 50%.
Yes I suppose you can say some pros don't belong there. But when you look at all the tennis players around the world trying to get into an ATP tour tennis tournament there is a certain population pressure. I am not sure it makes all that much difference if you have to go with a player ranked 350 in the world instead of 345. But when you have a team with an established rating that .45 difference in dynamic rating I would think it will be close to 90% - at least for the men. (the higher ranked guy(s) should win 9 out of 10 matches.)
Schmke said for his dynamic ratings when the difference is .35-.45 the favorite wins 90% of the time. I think that may have been for singles matches though.
https://tt.tennis-warehouse.com/ind...utr-is-plummeting.694839/page-2#post-15265732
The top 100 UTR singles men right now range from 14.93 to 16.26. Top 200 range is 14.47 to 16.26. Not sure where to draw the cutoff, but a range of about 1.5 UTR units across pro tournament participants (above Challenger level) seems reasonable.
That 1.5 is, in theory, also the range of a men's NTRP level (4.0 range being about 5.25 to 6.75), as discussed in this thread:
NTRP to UTR conversion formula on TLS
I recently noticed that the TLS estimated ratings site includes an estimate of your UTR based on their estimate of your NTRP (both to the hundredth). I looked through the other threads that discuss NTRP / UTR conversions and didn't find this TLS conversion formula mentioned. I figured out the...
tt.tennis-warehouse.com
Yes, league matches probably do have more outlier matchups with >1.5 UTR difference compared to pro matches, because of out-of-level players being included. But I don't think those more extreme mismatches are all that common overall, especially in doubles.
I could be wrong - it'd be an interesting thing to test.
Moon Shooter
Hall of Fame
A 2.3 utr difference takes us from 15.3 to 13.0 and that takes us from Lorenzo Sonego at number 59 in the world according to UTR with a 15.29 rating to the number 1347th player in the world with a 13.00 rating. It is pretty rare that he will play anyone more than .5 UTR different than his own rating.
For sure, in fact I’d be willing to bet #1347 really has zero chance of qualifying and ever playing #59. Otoh, a club player utr 4.5 that plays with his usta 4.0 club team is very likely to see a 6.8 and it’s not going to be pretty.
I would think predicting matches of players within a smaller rating group (atp) would definitely be harder than the USTA League matchups.
Not to mention you have guys tanking matches and that really throws off the ratings in that area. That is one area that probably skews percentages in usta
Can confirm - in fact, I've seen UTR's as low as 3.x and as high as 8.x in 4.0 leagues. Granted, the 8.x is just passing through at 1st singles on the way to 4.5 18+, and the 3.x is likely playing 3rd doubles in 40+/55+ and probably will be bumped down in the near future - but it's not uncommon.
Personally, as long as people aren't playing mixed, I like UTR because that one year cliff tends to capture who's doing better/worse than in past years. If people play a bunch of mixed, that seems to bring their UTR down quite a bit from where I'd expect it to be just from the eye test. Also, all of these are useless if people are actively managing their scores and wins/losses (which is also fairly common, unfortunately).
Regardless, none of these are linear. It's kind of like that Dr. Who quote - is 1 UTR level/1 level NTRP a lot or a little? It depends on the context
Personally, as long as people aren't playing mixed, I like UTR because that one year cliff tends to capture who's doing better/worse than in past years. If people play a bunch of mixed, that seems to bring their UTR down quite a bit from where I'd expect it to be just from the eye test. Also, all of these are useless if people are actively managing their scores and wins/losses (which is also fairly common, unfortunately).
Regardless, none of these are linear. It's kind of like that Dr. Who quote - is 1 UTR level/1 level NTRP a lot or a little? It depends on the context
silverwyvern4
Semi-Pro
If UTR and Tennis Record remain forever broken, all this discussion is Moot. When are they going to fix it??
Moveforwardalways
Hall of Fame
Mixed doubles is the Achilles heel of UTR. Unless both genders are playing against one another in singles and unbalanced co-ed competition over multiple years, UTR will always have a skewed data set. 90% of each person’s rating is against same gender competition and therefore the rating is not validated in mixed competition.
schmke
Legend
Someone tell me who it is!
But I'll be honest, I do miss on year-end levels sometimes and think it is the USTA that really missed as the USTA's year-end level just doesn't pass the smell test, and the player promptly gets to the level I had them at the following year which would seem to support my rating being more representative/accurate.
I wouldn't give TR a pass across the board, there are a lot of misses that are real misses, but it is possible some glitch or manual shenanigans resulted in the USTA's rating not really being representative of this player's ability.
Similar threads
