You are using an out of date browser. It may not display this or other websites correctly.

You should upgrade or use an alternative browser.

You should upgrade or use an alternative browser.

- Thread starter mpnv1990
- Start date

The USTA doesn't publish or want players to know for fear they'd use the information to manage ratings, but it is generally understood to be at least several tenths into the 4.0 range, e.g. they have to be playing at a solid 4.0 level, perhaps even closer to the 4.5 level than 3.5 level, to get strikes. So there is ample room for a player to be well on their way to being bumped up at year-end but not get strikes.

One could argue the strike threshold are too high (I have), but the USTA tends to give the benefit of the doubt to new players not wanting to "punish" them unfairly and give them a bad experience.

What you’re saying makes sense. I saw one USTA district say that it would have to be 3.80 to generate a strike.The USTA doesn't publish or want players to know for fear they'd use the information to manage ratings, but it is generally understood to be at least several tenths into the 4.0 range, e.g. they have to be playing at a solid 4.0 level, perhaps even closer to the 4.5 level than 3.5 level, to get strikes. So there is ample room for a player to be well on their way to being bumped up at year-end but not get strikes.

One could argue the strike threshold are too high (I have), but the USTA tends to give the benefit of the doubt to new players not wanting to "punish" them unfairly and give them a bad experience.

If somebody plays doubles against two upper end 3.5s (let’s say 3.35 dynamic) and their partner is also around 3.35, what are the odds of generating a strike if the final outcome is 6-2, 6-2?

You are assuming the dynamic ratings you listed are accurate? Assuming they are, I don't think that would generate a strike. Would have to be a 4.0 player that is higher rated and that self rate beat them 1,1.What you’re saying makes sense. I saw one USTA district say that it would have to be 3.80 to generate a strike.

If somebody plays doubles against two upper end 3.5s (let’s say 3.35 dynamic) and their partner is also around 3.35, what are the odds of generating a strike if the final outcome is 6-2, 6-2?

TR would give that a 3.99 match ratingYou are assuming the dynamic ratings you listed are accurate? Assuming they are, I don't think that would generate a strike. Would have to be a 4.0 player that is higher rated and that self rate beat them 1,1.

Yes, I’m making an assumption. The result is real though.You are assuming the dynamic ratings you listed are accurate? Assuming they are, I don't think that would generate a strike. Would have to be a 4.0 player that is higher rated and that self rate beat them 1,1.

Doubt that gets a strike.Yes, I’m making an assumption. The result is real though.

This is where I don’t understand usta thinking. If a high 4.0 beats a low 4.0 0,0 then how does any level 3.5 beat a 4.0 of any level 2,2??Doubt that gets a strike.

Im referring to OP’s second example or scenario. Not the first.This is where I don’t understand usta thinking. If a high 4.0 beats a low 4.0 0,0 then how does any level 3.5 beat a 4.0 of any level 2,2??

What would @schmke calculate the match rating to be?What you’re saying makes sense. I saw one USTA district say that it would have to be 3.80 to generate a strike.

If somebody plays doubles against two upper end 3.5s (let’s say 3.35 dynamic) and their partner is also around 3.35, what are the odds of generating a strike if the final outcome is 6-2, 6-2?

According to this, the strike threshold is 3.80 for 3.5.

IMO, one of the guys I played yesterday generated a strike a few weeks ago.

He and his partner played two guys with a 5.00~ UTR. Tennis Record has one of the opponents at 3.50 and the other at 3.60. The guy and his partner won 6-2, 6-0.

USTA made me keep my rating after a 4-yr layoff from USTA matches. I had achilles surgery and recovery as well as an additional 2-yr layoff from any tennis and they still denied my appeal from 4.5 to 4.0

My question is not about level, but whether you will be a C or S rated player.USTA made me keep my rating after a 4-yr layoff from USTA matches. I had achilles surgery and recovery as well as an additional 2-yr layoff from any tennis and they still denied my appeal from 4.5 to 4.0

That can be tough… do flex leagues / tournaments count for your location? Playing through that may adjust your rating?USTA made me keep my rating after a 4-yr layoff from USTA matches. I had achilles surgery and recovery as well as an additional 2-yr layoff from any tennis and they still denied my appeal from 4.5 to 4.0

A competitive 4.5 team may still have 1 or 2 4.0s so you could probably still find a team even if youre not at your previous strength.

This is a misunderstanding of statistics (on behalf of both the USTA person who wrote that and the readers) that comes up occasionally on here. The "expected" result between a high 4.0 and a low 4.0 isThis is where I don’t understand usta thinking. If a high 4.0 beats a low 4.0 0,0 then how does any level 3.5 beat a 4.0 of any level 2,2??

It was poorly written and is commonly misinterpreted because "expected" has a statistical meaning. Expected value is the sum of all possible values times their probability of happening. Given that "games won" is only defined as a non-negative number, then a statistical expected value of 0 can only be achieved with a single probability distribution with 100% probability on 0. Probabilistically, as soon as you allow any probability at all for a result other than 0, the expected value is >0. An actual "expected" outcome of 6-0 6-0 would imply that there is zero chance that the opponent will ever win a game, which is obvious nonsense. We see non-bagel matches between people playing up all the time.

We need to stop the thinking of 6-0 6-0 as the"expected" result between the top and bottom of a range. It is better to think of as "not unexpected", or in statistical terms, there is high enough probability that it should not be viewed as proof that a player is out of level without any other evidence.

After 3+ years you would be a S rated player.My question is not about level, but whether you will be a C or S rated player.

I went through all this at the beginning of this year after not playing USTA for 5 years.

First I had to self rate again - had to answer various questions about my playing history, which turned out all to be pointless because right at the end of the process the last question was "what was your most recent USTA rating if any", and then it said I had to self-rate at that last rating at a minimum. So it could have cut to the chase and led with that up front.

Then I submitted an appeal to the self-rating, since (1) there are no leagues at that level in my area, (2) I haven't played a "serious" match in 5 years, and (3) I'm now over 55. Took a couple of weeks to get the response that my appeal was denied. Oh well.

I cannot +1 this enough. +100!This is a misunderstanding of statistics (on behalf of both the USTA person who wrote that and the readers) that comes up occasionally on here. The "expected" result between a high 4.0 and a low 4.0 isNOT6-0 6-0. That blurb that says it is was written (and worded very poorly) in the context of a FAQ to explain that a 6-0 6-0 result is not an outlier in that case and cannot be (by itself) the basis for a grievance, but the true expected result (or expected value) is more like 3 games for the lower player with occasional double bagels not out of the question, but a rare close match or even an upset in the far tail of the distribution not out of the question, either. Remember, the results are outcomes of random probability variables, not deterministic based on input ratings, so a range of outcomes is always possible.

It was poorly written and is commonly misinterpreted because "expected" has a statistical meaning. Expected value is the sum of all possible values times their probability of happening. Given that "games won" is only defined as a non-negative number, then a statistical expected value of 0 can only be achieved with a single probability distribution with 100% probability on 0. Probabilistically, as soon as you allow any probability at all for a result other than 0, the expected value is >0. An actual "expected" outcome of 6-0 6-0 would imply that there is zero chance that the opponent will ever win a game, which is obvious nonsense. We see non-bagel matches between people playing up all the time.

We need to stop the thinking of 6-0 6-0 as the"expected" result between the top and bottom of a range. It is better to think of as "not unexpected", or in statistical terms, there is high enough probability that it should not be viewed as proof that a player is out of level without any other evidence.

if you're self rated, you can try appealing down again after losing 3 matchesUSTA made me keep my rating after a 4-yr layoff from USTA matches. I had achilles surgery and recovery as well as an additional 2-yr layoff from any tennis and they still denied my appeal from 4.5 to 4.0

The language USTA uses (still there on the USTA ratings FAQ page) is this:This is a misunderstanding of statistics (on behalf of both the USTA person who wrote that and the readers) that comes up occasionally on here. The "expected" result between a high 4.0 and a low 4.0 isNOT6-0 6-0. That blurb that says it is was written (and worded very poorly) in the context of a FAQ to explain that a 6-0 6-0 result is not an outlier in that case and cannot be (by itself) the basis for a grievance, but the true expected result (or expected value) is more like 3 games for the lower player with occasional double bagels not out of the question, but a rare close match or even an upset in the far tail of the distribution not out of the question, either. Remember, the results are outcomes of random probability variables, not deterministic based on input ratings, so a range of outcomes is always possible.

It was poorly written and is commonly misinterpreted because "expected" has a statistical meaning. Expected value is the sum of all possible values times their probability of happening. Given that "games won" is only defined as a non-negative number, then a statistical expected value of 0 can only be achieved with a single probability distribution with 100% probability on 0. Probabilistically, as soon as you allow any probability at all for a result other than 0, the expected value is >0. An actual "expected" outcome of 6-0 6-0 would imply that there is zero chance that the opponent will ever win a game, which is obvious nonsense. We see non-bagel matches between people playing up all the time.

We need to stop the thinking of 6-0 6-0 as the"expected" result between the top and bottom of a range. It is better to think of as "not unexpected", or in statistical terms, there is high enough probability that it should not be viewed as proof that a player is out of level without any other evidence.

"A typical match result for a player with a 3.01 rating versus a 3.49 player, both of whom are 3.5s, would be 6-0, 6-0 in favor of the higher rated player."

The word "typical" does not have a clear statistical meaning, and you're right in this case it cannot possibly refer to the expected value i.e. arithmetic mean. However, it's possible that it could be referring to median and/or the mode, which are other common statistical definitions of "typical."

If 6-0, 6-0 is the median result that means USTA expects a 50% chance or higher for that score to be the result. If that's true, then I would think that the mean number of games expected to be won by the underdog is well less than one. While it's possible for the median to be zero and the the mean to be >1, that would have to be a very skewed distribution with high variance.

If 6-0, 6-0 is the mode (most common single result) then its chance could be lower than 50%. For example there could be something like a 40% chance of the underdog winning 0 games, 30% chance of exactly 1 game, 20% chance of exactly 2 games, and 10% chance of >2 games. So the underdog is more likely than not to avoid getting double bageled, but it's still the most likely single result. In that case, I'd still think that the mean is less than one (but maybe close to one).

It's also possible that the mean, median, and mode, are all 1 game or more won by the underdog, and "typical" just means "not unlikely" like you say. But given their language I would guess that they do assume that the double bagel is the median and/or mode result, in which case the mean is probably less than one game for the underdog.

I suppose I could take a look at the "typical" score between players 0.5 or more apart ...

In 2023 in 18+ singles, I show 8208 singles matches where the players were rated 0.45-0.55 apart.

I suppose I could take a look at the "typical" score between players 0.5 or more apart ...

Looking at distinct scores, the most common are:

6-1,6-1 - 806

6-1,6-0 - 626

6-0,6-1 - 594

6-1,6-2 - 480

6-2,6-1 - 432

...

6-0,6-0 - 212

There are 12 scores more common than 6-0,6-0 in this scenario, and even the most common 6-1,6-1 is only 10% of the total.

I'll stand by my comment that 6-0,6-0 is not "typical" when players 0.5 apart play each other.

Interesting data. If I'm understanding correctly, another way to look at it is:In 2023 in 18+ singles, I show 8208 singles matches where the players were rated 0.45-0.55 apart.

Looking at distinct scores, the most common are:

6-1,6-1 - 806

6-1,6-0 - 626

6-0,6-1 - 594

6-1,6-2 - 480

6-2,6-1 - 432

...

6-0,6-0 - 212

There are 12 scores more common than 6-0,6-0 in this scenario, and even the most common 6-1,6-1 is only 10% of the total.

I'll stand by my comment that 6-0,6-0 is not "typical" when players 0.5 apart play each other.

0 games won by loser: 2.6% (212/8208)

1 game won by loser: 14.9% (1220/8208)

2 or more games won by loser: 82.6% (6776/8208)

If you have the data for 6-0,6-2 and 6-2,6-0 we could combine with 6-1,6-1 and further separate 2 games won from 3+ games won.

But already it's clear that even 0/1 game won is somewhat rare (< 18% of the time).

My own anecdotal experience is that I'm more likely to win 0&0 against a player somewhat close in ability to me versus one way weaker. Just because of my level of focus. I tend to somewhat play to the level of my opponent once I think a match is well in hand, which I'm guessing most people do.

So if you add 6-2, 6-0 or 6-0, 6-2 scores also, you can expect a proper drubbing in about half the matches. USTA probably adds the double bagel comment to make teams think twice before they make ‘sandbagging’ accusations against the winner.In 2023 in 18+ singles, I show 8208 singles matches where the players were rated 0.45-0.55 apart.

Looking at distinct scores, the most common are:

6-1,6-1 - 806

6-1,6-0 - 626

6-0,6-1 - 594

6-1,6-2 - 480

6-2,6-1 - 432

...

6-0,6-0 - 212

There are 12 scores more common than 6-0,6-0 in this scenario, and even the most common 6-1,6-1 is only 10% of the total.

I'll stand by my comment that 6-0,6-0 is not "typical" when players 0.5 apart play each other.

There is a small chance it could be the mode with a spike at zero then still meaningful probability on 1 through 5 or 6 and diminishing but nonzero after that, although my guess would be that the mode is probably more like 2 or 3.The language USTA uses (still there on the USTA ratings FAQ page) is this:

"A typical match result for a player with a 3.01 rating versus a 3.49 player, both of whom are 3.5s, would be 6-0, 6-0 in favor of the higher rated player."

The word "typical" does not have a clear statistical meaning, and you're right in this case it cannot possibly refer to the expected value i.e. arithmetic mean. However, it's possible that it could be referring to median and/or the mode, which are other common statistical definitions of "typical."

If 6-0, 6-0 is the median result that means USTA expects a 50% chance or higher for that score to be the result. If that's true, then I would think that the mean number of games expected to be won by the underdog is well less than one. While it's possible for the median to be zero and the the mean to be >1, that would have to be a very skewed distribution with high variance.

If 6-0, 6-0 is the mode (most common single result) then its chance could be lower than 50%. For example there could be something like a 40% chance of the underdog winning 0 games, 30% chance of exactly 1 game, 20% chance of exactly 2 games, and 10% chance of >2 games. So the underdog is more likely than not to avoid getting double bageled, but it's still the most likely single result. In that case, I'd still think that the mean is less than one (but maybe close to one).

It's also possible that the mean, median, and mode, are all 1 game or more won by the underdog, and "typical" just means "not unlikely" like you say. But given their language I would guess that they do assume that the double bagel is the median and/or mode result, in which case the mean is probably less than one game for the underdog.

I'm almost 100% certain zero is not the median. Given that the median is the 50th percentile and zero is the first point on the distribution, that would imply that over half of those matches are double bagels, which is not supported even slightly by my observations, but I don't have the hard numbers on that. It's possible @schmke could shed some light on matches with >0.45 rating difference and see what the distribution of outcomes really is.

Edit: looks like I'm late to the party with this request. The pitfalls of working in between typing sentences. LOL.

Yeah, the probability based on Schmke's analysis seems to be about 2.5% or 1-in-40. While that sounds rare, 2.5% in a probability distribution is significant enough that the USTA is right that it is not enough to be proof in and of itself that a player is out of level. If you're talking about a 1-in-1000 event or 1-in-10000 or something, then there's a case that you won't see it much, but 1-in-40? That's nothing.So if you add 6-2, 6-0 or 6-0, 6-2 scores also, you can expect a proper drubbing in about half the matches. USTA probably adds the double bagel comment to make teams think twice before they make ‘sandbagging’ accusations against the winner.

They probably make the double bagel comment to make sure that the 50% of opponents who get beaten winning only 2 or 3 games in a match when they play a player with close to 0.5 dynamic rating difference don’t think they should gripe about this.Yeah, the probability based on Schmke's analysis seems to be about 2.5% or 1-in-40. While that sounds rare, 2.5% in a probability distribution is significant enough that the USTA is right that it is not enough to be proof in and of itself that a player is out of level. If you're talking about a 1-in-1000 event or 1-in-10000 or something, then there's a case that you won't see it much, but 1-in-40? That's nothing.

if they should win 3 games total, their dynamic would rise to 3.04. Winning 4 games, the typical 6-2 6-2 drubbing, would raise their dynamic to 3.07.

There is no reason whatsoever to believe that what whoever wrote that blurb called "typical" and what the USTA uses behind the scenes in their rating tables is internally consistent in any way.

Perhaps you're right, but is there any reason to believe they're not consistent? I'm not aware of any evidence one way or the other. We're both just guessing, right?There is no reason whatsoever to believe that what whoever wrote that blurb called "typical" and what the USTA uses behind the scenes in their rating tables is internally consistent in any way.

Because it is only 3 games and sometimes people are better at singles or doubles. I suppose if you are better at singles your captain should account for that in the lineups and have you play some doubles first and vice versa. The two guys I can think of that got promoted or DQed are at the bottoms of their levels. The guy who got promoted midyear to 3.5 has a 3.99 UTR and the guy that got DQed up to 4.0 had a 5.45 singles and a 4.94 doubles UTR. Both of those UTRs are not at all unusual for the level they were at before promotion/DQ.

That is surprisingly low. It is also interesting that the 6-1,6-1 6-1,6-0 and 6-0,6-1 is higher then everything but then it jumps to the losing player winning three games instead of 6-0,6-2 and 6-2,6-0.In 2023 in 18+ singles, I show 8208 singles matches where the players were rated 0.45-0.55 apart.

Looking at distinct scores, the most common are:

6-1,6-1 - 806

6-1,6-0 - 626

6-0,6-1 - 594

6-1,6-2 - 480

6-2,6-1 - 432

...

6-0,6-0 - 212

There are 12 scores more common than 6-0,6-0 in this scenario, and even the most common 6-1,6-1 is only 10% of the total.

I'll stand by my comment that 6-0,6-0 is not "typical" when players 0.5 apart play each other.

Two quibbling/minor questions and one main issue:

1) Were there more total matches in the .45-.50 range then the .50 to the .55 range? I still wouldn't expect this to be so close even in the .50-.45 set, but I wonder if it was lopsided.

2) Also what percentage of the matches were women's matches. I suspect the women's levels are more narrow overall.

3) Mainly I just question using singles. I know plenty of singles players that hit loopy consistent shots and returns but get killed in doubles. Since USTA is like 90% doubles I think much of this can be explained by some of these singles players tearing up their rating playing doubles. If you take 3.5 men's doubles and the average rating difference is .5 (so 1.00 total difference between the sides) then I would bet the lower rated team is losing more total games then the numbers you posted in singles.

- Replies
- 5

- Views
- 2K

- Replies
- 58

- Views
- 6K

- Replies
- 0

- Views
- 919

- Replies
- 14

- Views
- 2K