Curtennis
Hall of Fame
Are there not many 5.0 options in Florida? I would have figured with the exception of maybe California, there’s more tennis in Florida than anywhere else in US.
The dreaded EOY Rating thread
- Thread starter Brilliant
- Start date
Lord Of The Strings
Rookie
That's pretty good. I know my captain is hoping that TR is accurate, but I think we all feel like TR is underrating most of our team by about 0.1. Lots of guys on my spring team are on the TR cusp of bumping up into 4.0 (myself included, but it has me still shy of the bump), but @schmke has me in 4.0 range (fwiw, difference of 0.13 there). None of us would complain about a bump up and I think several of us are hoping for it. We've run the table in our division the past 2 years and had decent showings at states given that we normally don't have half of our best players there (all computer rated players, no self-rates, no appeal downs before anybody jumps down my throat).
All that said, a lot of the guys also play tournaments and our section factors in sanctioned NTRP tournament play into year-end ratings. My guess is that TR's accuracy will vary by section a decent amount, especially sections that factor in sanctioned NTRP tournaments.
silverwyvern4
Semi-Pro
You appealed UP?????
Did you play in leagues up 2 levels?
silverwyvern4
Semi-Pro
Do you think I should buy a schmke rating estimate for my team now even though ratings will come out tomorrow?
silverwyvern4
Semi-Pro
Maybe tennis record ratings are more accurate than USTA ratings.
It's really what you're looking to understand. I certainly wouldn't do it now just to see what his ratings are. I've gotten an individual report in the past which I found very interesting. It's more useful if you've played a lot of matches since there's more data points too.
TR ratings are an attempt to estimate USTA NTRP. In that sense, it cannot be more accurate. I would say with that in mind even if it did more accurately represent someone's level that would be a happy accident.
schmke
Legend
I can't speak for TR, but I know there are times where I look at the players I miss on and wonder how the USTA got it, IMHO, wrong and didn't bump up a player I predicted would be. These situations are often supported by the player dominating at their "too low" level the next year and makes me not feel so bad about predicting a bump that didn't happen.
Do you have any idea how this happens?
If the system is extremely consistent for most players, how could it get a few players so incredibly wrong?
Have you ever tried to join a team at a level you're not rated for? Try telling the team Captain that you can always play up and let me know how it goes. lol.You can always play up a level. You just want the rating so you look cool.
Any proof?
Winners WIN!
TennisManiac
Hall of Fame
That's nonsense
TennisManiac
Hall of Fame
Yes. But I dont feel like getting into it.
TR match ratings are "useful", but they're not right if you are trying to use them to predict USTA bumps. TR generally gets the order of the players within a team right (i.e. they have the stronger players at the top and the weaker players at the bottom), but the actual estimates can be wildly incorrect.
It's possible, and it also depends on your definition of "more accurate". If you're using TR as a proxy for actual USTA ratings for predicting bumps and whatnot, then, no, a proxy is never more accurate than the actual thing it is modeling, by definition. If you are comparing TR and NTRP as two different models of underlying tennis ability, then maybe TR is more "accurate", but that's largely irrelevant for people playing USTA leagues. Also, how do you define "accuracy"? Minimum bias? Minimum squared error from some theoretical incalculable underlying tennis ability? Something else?
Did TR in its year-end projection calculate and include "eligible" self-rate matches? If not, the omission of self-rate tainted matches creates some error.
They keep including them and then excluding them. The person can't make up their mind.
I totally agree with you, but there are many different factors that could contribute to the errors in their year-end bump predictions, other than their match rating formula. One factor we know about for sure is that in many areas they fail to include / exclude the all the matches that count / don't count for USTA ratings in that area. Also we know that they have to guess at the starting value for most players at the beginning of a rating year, which likely feeds forward into errors in the year-end ratings. Then they have to guess at how USTA does the year-end calculation, including weighting different leagues or playoff rounds or tournaments differently, adjustments based on Nationals results, etc. I suspect that those things are the main contributors to the errors, and that their formula for calculating an individual match rating is probably pretty accurate.
travlerajm
Talk Tennis Guru
It’s my understanding that TR doesn’t “guess” at the starting value each season. Rather, TR “calibrates” by setting players that its algo rated outside the ntrp rating band to an “adjusted rating” right at the edge of the player’s official usta computer rating.
For example, a player that finished rated 4.12 in TR that ends up 4.0C will be reset to a 3.99 TR rating to start the season. That is, TR rated the player a 4.5, but it is calibrated so all players start out in same rating band as the usta computer has them.
Moveforwardalways
Hall of Fame
TR is factoring in combo, singles leagues, etc. USTA only factors in adult spring league in most areas.
Lord Of The Strings
Rookie
I think you're best off buying a rating estimate for your team after ratings come out (assuming the team stays mostly intact). This can give you some good insight on doubles pairings/performance for your team as well as individuals. But you can also do this mid-season or just before states/sectionals for more data points.
I'd argue that "calibrate" implies more precision than it is - if the player got a 4.0C then their actual starting rating in the new year could be anywhere from 3.50 to 3.99 - TR is just guessing that it is 3.99, right?
travlerajm
Talk Tennis Guru
It’s more of a system calibration. It ensures that every player in the TR database starts the year with ntrp rating in the same bags as his computer rating. Without this, the algorithm could easily drift. It’s neither right nor wrong. It’s just a way to force-fit it’s results to be closer to the usta computer.
Both of these assertions are false.
In the USTA computer every returning player has a "to the hundredth" starting value, used to begin the new year's calculations. When TR chooses a value to use for that, it's going to be either "right" if it matches USTA's value exactly or "wrong" if it doesn't. They do make adjustments to ensure the starting value estimate is in the correct 0.5-range bin, but their precise value is probably wrong for most players.
I wonder if this partly explains why the TR estimates tend to skew low year after year. The majority of their "misses" are low, i.e. there are many players who receive a higher computer rating than TR had predicted, much more than the opposite. They do then adjust those players up to the minimum of their actual level, but that is not enough of an adjustment, because only some of those players were actually at the minimum in USTA's result - surely many of them were a few hundredths above the minimum and some much higher than that. But because TR starts ALL of that group at the extreme minimum, on average they are starting out too low, which then makes them (and their opponents, and their opponents' opponents, etc.) too low again the next year.
Moveforwardalways
Hall of Fame
They are not assertions, and they are not false.
atatu
Legend
I don't it's accurate to say that USTA only factors in the adult spring leagues, where are you getting that ?
Moveforwardalways
Hall of Fame
In most areas, your NTRP rating is only based on regular adult league results - not combo, singles leagues, tournaments or mixed. Only if you did not play regular adult league can you get a “mixed-only” rating or a “tournament-only” rating. There are only a few areas that are exceptions to this.
nobackhand
New User
Do you know which areas are the exception to this?
silverwyvern4
Semi-Pro
Are the bump and bump down thresholds the same? 3.00 is cutoff to be bumped up from 3.0 to 3.50 and also to be bumped down from 3.5 to 3.0? Or is it harder to get bumped down, like a 3.5C would need a rating under 2.95 to get bumped down?
That's basically correct. But, originally you said "USTA only factors in the adult spring leagues." Your comments about the leagues' TR counts are still incorrect.
In my section, I believe only Adult league matches (whether Winter, Spring, Summer, or postseason) count toward a player's dynamic rating. But TriLevel gender, singles league, and tournament match results still get factored in at year-end (but do not produce strikes for dynamic disqualification).
If I recall correctly, 3.01 is the cutoff for 3.5s. So, 3.01 - 3.50 is the 3.5 range. If you fall to 3.00 you'll get demoted to 3.0. But, if that happened, there is a narrow appeal range (approx 2.97 - 3.0) that you could use to appeal up!
For the Southern Section, TR computes scores from mixed, combo, tri-level, etc., but does not factor them into the spring league ranking that counts for promotion. All of my combo, mixed, etc. matches have a match rating, but the overall rating is static from spring (except for mixed, which seems to have its own category).
It's possible you are right, but I don't think so. Happy to have you prove me wrong if you can. Many areas do include non-Adult League results. For example the entire Southern section, which is huge, is supposed to include NTRP tournament results in their 2022 year-end ratings, according to their regulations document (see Page 13):
cks
Hall of Fame
This is my understanding as well. I checked and found the USTA web page on NTRP which states the same thing.
Moveforwardalways
Hall of Fame
I will defer to @schmke on this. If I am wrong, I’m wrong. But if tournament results count this year for southern, that would be brand new and I would be shocked to have not heard people talking about it. And there will now be a way to throw matches and sandbag like never before in southern.
silverwyvern4
Semi-Pro
Is the appeal up band wider than the appeal down band? Seems like it would be. They like having people Appeal up
What is the highest you've had somebody over the cutoff and gotten wrong?
OnTheLine
Hall of Fame
Many sections include tournament results and results from non-advancing leagues. Intermountain is another example of this. The only leagues that do not count are Mixed and their odd summer league ... but ladies' daytime league, Fall non advancing as well as tri-level count towards ratings. Each league is weighted differently in terms of how they flow into final rating. 18+ and 40+ carry the most weight ... per LLC there the tournament results are very lightly considered.
The weighting concept may be accurate, but it doesn't make sense. For example, the 40 & Over results are now nine months old. The Fall League results are only one month old (perhaps only two weeks old). Yet if what you've said is true, the nine-month-old results would be weighted heavier than the very recent results. The Fall League also has the same players as the 18 & Over League because the Fall League allows 18+-year-olds.
Last edited:
schmke
Legend
I usually chalk it up to A) something in their year-end calculations that gives too much credit for simply playing in playoffs or single unexpected matches there, or B) manual adjustments they sometimes make to certain regions. But I also know of a few cases where players have been bumped up and they ask why and someone checks and they say "ooops" and just correct it, so I guess it is possible this could explain it.
This is what the USTA has said about tournament matches concerning the NTRP:
"Do tournament matches count more or less for my year-end rating than league matches? (Also in Year-End) If a Section has opted to include tournament play in their NTRP ratings, tournament results count the same as league matches."
Hence, this seems to contradict what you've been told by your LLC.
schmke
Legend
Southern is pretty much the only section that only uses advancing leagues for ratings, and yes, even they now include tournaments in year-end calculations. Most other sections also include tournaments, but many also include non-advancing leagues like Tri-Level and singles leagues and other non-advancing doubles leagues that are run locally and only have playoffs locally or perhaps through Sectionals but that is it.
OnTheLine
Hall of Fame
It makes all the sense in the world. If you weighted the Fall non-advancing league the same as an advancing league it would be even more an opportunity to tank and manage ratings than it is already.
In terms of weighting of tournaments ... I will absolutely go with the word of this particular LLC over what the USTA posted somewhere.
Why? because she has been with USTA for more than 20 years ... formerly at national before moving to a semi retirement position locally. That and I really believe that USTA will post inaccuracies solely to protect people from being able to more effectively manage their rating.
silverwyvern4
Semi-Pro
So we're expecting ratings to be released in 12-14 hours from now?
I’m expecting it to be released at 12 AM ET on December 1st. In other words, as soon as the day changes from November 30th to December 1st.
Some Adult Leagues may not advance to Sectionals. Hence, that's an example of a non-advancing league, correct? If an Adult League advances to Sectionals but not Nationals, what's that league called? For instance, we have a league with a local playoff between flights and advances to Sectionals in January. But, play terminates at Sectionals. The LC stated in the registration email that the matches count for NTRP ratings, but no mention was made about weightings.
Also, the source of the info I previously posted about weightings is USTA League Nationals FAQ.
Last edited:
