A ratings question: what rating would these players be?

[AndrewD]

There have been a lotof threads over the past 12 months or so that ask people to rate players based on the perceived quality of their strokes (as seen in a video). They've been entertaining but, ultimately, very inconclusive. The missing ingredient is always competition which is, obviously, the truest test of a player's rating.

What I thought might interesting is to see how big a gap, in terms of games won, people believe there is between ratings.

So;

If player X is capable of winning no more than 0-2 games in any set against player Y (scores are indicative of their average level of play, not one playing better or worse than usual) what would be the difference in rating between them (eg: 0.5pts, 1pt, 2pts)

If player X is capable of winning no more than 4 games in any set against player Y, what would be the difference in rating between them

If player X is capable of winning at least one set against player Y, what would be the difference in rating between them.

 

[Hokiez]

A) 1pt
B) 0.5pts
C) equal by NTRP, but 0.25pts in the actual computer value.

That's my opinion at least.

 

[raiden031]

When I play with one of my playing partners, I usually win sets either 6-0 or 6-1, but sometimes I might lose 6-4 and occasionally even lose a set against them. So really I don't think one can expect the same score differential to consistently occur, unless the level of play is really different, in which case it might always be 0 and 0 or 0 and 1.

Hokiez, as an example, do you think a weak 3.0 will typically lose to a weak 3.5 by a score of 6-4, 6-4? I definitely don't see that happening consistently.

USTA thinks 6-0, 6-0 is a minimum of 0.35 apart. 6-2, 6-2 is around .25 or something and splitting sets is pretty much even. I mostly agree with this but I think its kinda difficult to double bagel someone who is 0.35 below. Usually people will blow a game or two every match due to UEs at least. So if you can consistently beat someone 6-0, 6-0, then I think you're way better than someone who can consistently beat someone 6-0, 6-1 or 6-1, 6-1.

 

[tfm1973]

there was another thread floating around and i hope i didn't misread it - but it looks like the USTA feels that within any given rating level - let's say 4.0 -- that there is a rather large range of playing ability even within players who are both 4.0

at first i thought it was ridiculous to think that the lowest 4.0 should consistently lose 0-6, 0-6 to the highest rated 4.0 player.

but the more i think about it - to me it makes sense. in my leagues there is a great range within players who are 4.0 - so as a doubles specialist if i played our #1 singles player and consistently lose love and love - it doesn't necessarily mean that he's greater than 4.0 and it doesn't necessarily mean that i'm worse than 4.0 -- it would just mean i'm a bottom feeding 4.0 with bad court coverage.

and i bet to a spectator we probably look pretty similar with the exception of one player losing most of the key points.

scores don't mean as much as we assign. and unfortunately neither do the ratings. it would be great if someone could come up with a meaningless number that you can calculate as you play a match - kinda like the quarterback rating when you watch the NFL.

 

[raiden031]

scores don't mean as much as we assign. and unfortunately neither do the ratings. it would be great if someone could come up with a meaningless number that you can calculate as you play a match - kinda like the quarterback rating when you watch the NFL.

Amen. Too many people on this board think that the USTA computer-rating is the holy grail of measuring tennis skills. I think everyone knows the system is not perfect and that it could be improved, however I think for the most part the system actually does what is intended of it, to match players with other players of similar skill level. Just because someone is a computer-rated 3.5, doesn't mean they are any better than 3.0s or worse than 4.0s. It just means that based on a few matches they played for 3 months out of the year, the system calculated some rating for them. Alot of factors affecting the rating could be whether they were playing doubles with a strong/weak player, whether they just started playing tennis again after years off, maybe this is their first USTA league and they mentally broke down, etc.

Others think that if someone at their same skill level beats them pretty easily, then they have been wronged and that the system is flawed.Not everyone is going to win a National Championship and many people (especially on this board) need to accept that.

We all need to take the NTRP ratings with a grain of salt. There is good reason why the pros don't use this system. Because scores don't tell as much as a Win or Loss. Fed could beat the same guy 7-6, 7-6 probably 10 times in a row yet USTA will say they are equal in skill.

 

[Kevo]

I don't think you can tell much by point rating unless they're close. I mean a reasonable 4.0 will double bagel a 3.5 most of the time if they're playing seriously. A high 4.0 almost 4.5 will be able to double bagel a low 4.0 much of the time.

I am a fairly high 4.0 and I played 3 sets a couple of weeks ago against a 4.0 who wins about half of his matches. I lost the first set 6-4. I won the next two 6-1, 6-2. It was the first time I've played this guy. I think if I played him again it might be more like 1 and 0.

There is just a huge range of things that can happen in an actual match that affects the score. The numbers are mainly designed to keep the matches competitive. You can have a fairly competitive match and still lose big. It wasn't like my opponent was getting blown off the court in those last two sets.

 

[NoBadMojo]

I don't think you can tell much by point rating unless they're close. I mean a reasonable 4.0 will double bagel a 3.5 most of the time if they're playing seriously. A high 4.0 almost 4.5 will be able to double bagel a low 4.0 much of the time.

I am a fairly high 4.0 and I played 3 sets a couple of weeks ago against a 4.0 who wins about half of his matches. I lost the first set 6-4. I won the next two 6-1, 6-2. It was the first time I've played this guy. I think if I played him again it might be more like 1 and 0.

There is just a huge range of things that can happen in an actual match that affects the score. The numbers are mainly designed to keep the matches competitive. You can have a fairly competitive match and still lose big. It wasn't like my opponent was getting blown off the court in those last two sets.

this is true for sure. there is a big difference when someone puts a bagel and a breadstick on someone where the opponent doesnt get many points compared to when someone wins by the same score when the games go to 30all and such.

personally, i think there are too many levels...they should eliminate the .5 levels...it's pretty easy to tell a 4.0 from a 5.0 for example..things get really fuzzy when trying to separate a 3.5 from a 4.0 for example however..or a 4.0 from a 4.5

 

[kevhen]

If you only get a game a set, the guy is at least a level above, maybe 1.5 levels or .75 NTRP. 6-1, 6-1

If you double bagel someone then it's two full levels. 6-0, 6-0

If you can't win sets but win games then you are about one level below (.5 NTRP). 6-3, 6-2

If you can win sets but not matches then you are at the same level but like .1 to .25 below. 7-6, 3-6, 3-6

 

[cak]

At one point the actual algorithm was posted on the net somewhere. I vaguely remember that a 6-4 6-4 result or closer would not change your rating, as that was considered close enough that you were considered the same level. Splitting sets would also be considered the same level. 6-2 6-2 would be considered maybe .5 difference. Anything below that could be anywhere from .5 difference to 1 or even 2 point difference. (I mean there are double bagels where you get some points, and there are double bagels where you get no points...)

Does anyone know if the algorithm is still on the net somewhere. (My old link doesn't work.)

 

[raiden031]

At one point the actual algorithm was posted on the net somewhere. I vaguely remember that a 6-4 6-4 result or closer would not change your rating, as that was considered close enough that you were considered the same level. Splitting sets would also be considered the same level. 6-2 6-2 would be considered maybe .5 difference. Anything below that could be anywhere from .5 difference to 1 or even 2 point difference. (I mean there are double bagels where you get some points, and there are double bagels where you get no points...)

Does anyone know if the algorithm is still on the net somewhere. (My old link doesn't work.)

This is probably what you're referring to:

http://tt.tennis-warehouse.com/archive/index.php/t-116331.html

here is the differential chart:

7-6, 7-6 = 0.06
7-6, 6-4 = 0.09
6-4, 6-4 = 0.12
6-3, 6-4 = 0.15
6-3, 6-3 = 0.18
7-5, 6-2 = 0.18
6-2, 6-3 = 0.21
6-2, 6-2 = 0.24
6-3, 6-1 = 0.24
6-1, 6-2 = 0.265
6-3, 6-0 = 0.27
6-2, 6-0 = 0.295
6-1, 6-1 = 0.295
6-1, 6-0 = 0.325


I assume 6-0, 6-0 is a special case and probably means their new ratings must be at least .35 apart.

 

[AndrewD]

When I play with one of my playing partners, I usually win sets either 6-0 or 6-1, but sometimes I might lose 6-4 and occasionally even lose a set against them.

If that is the case then the next part doesn't fit.

So really I don't think one can expect the same score differential to consistently occur, unless the level of play is really different, in which case it might always be 0 and 0 or 0 and 1.

You said that you "usually win sets either 6-0 or 6-1". That means you are consistantly (usually/on average) getting the same type of score.

 

[raiden031]

If that is the case then the next part doesn't fit.



You said that you "usually win sets either 6-0 or 6-1". That means you are consistantly (usually/on average) getting the same type of score.

Usually as in maybe 75% of the time. Then the other 25% might range from 6-2 to me losing a set. I really don't think I'm that much better than the person that I should be bageling them as much as I do, but it is completely the way we match up.

 

[cak]

Thanks Raiden, that's not the chart I remember, but that one makes more sense, as I seldom see people move up or down without quite a few bagels involved.

 

[Ace]

I have a spreadsheet I created using the algorithm that was once posted on the internet. I use it during the season just for the fun of it. I can post it somewhere if anyone wants to try it out.
Its not 100% accurate, because you have to estimate the rating of your opponent, but its ok, and if nothing else, it proves that losing, but being competitive, doesn't necessarily bring down your rating.

 

[kevhen]

Easitennis says that 6-2, 6-2 is one level of difference but USTA thinks that 6-0, 6-1 is still well within one level but obviously their formula is messed up. It's very hard to double bagel someone that is a level below you. Easy to double bagel someone two levels down though. I got 5 games off a USTA computer rated 5.0 and I am a weaker 4.5 in singles. 6-2, 6-1 is about normal score for one level difference. You are winning about 60% of points, 80% of games, and 100% of matches then.

USTA formula makes it easy to sand bag and stay within level by just losing a couple of games.

 

[AndrewD]

Usually as in maybe 75% of the time. Then the other 25% might range from 6-2 to me losing a set. I really don't think I'm that much better than the person that I should be bageling them as much as I do, but it is completely the way we match up.

Sounds like you are either genuinely a level above them or you guys have played so much you've fallen into a pattern where, unless you're feeling a bit off your game, you expect to win 6-0, 6-1 and he expects to lose by that sort of score.

Still , I was surprised to see, in the chart you listed, that a 6-0, 6-1 win doesn't equal a full 1 point difference between two players. I was expecting to see something along the lines of, an average win of 6-3/6-2 indicating a true half point between the players, an average win of 6-1/6-0 indicating a true 1 point between the players and an average win of 6-0/6-0 indicating more than 1 pt difference between the players.

 

[LuckyR]

This is probably what you're referring to:

http://tt.tennis-warehouse.com/archive/index.php/t-116331.html

here is the differential chart:

7-6, 7-6 = 0.06
7-6, 6-4 = 0.09
6-4, 6-4 = 0.12
6-3, 6-4 = 0.15
6-3, 6-3 = 0.18
7-5, 6-2 = 0.18
6-2, 6-3 = 0.21
6-2, 6-2 = 0.24
6-3, 6-1 = 0.24
6-1, 6-2 = 0.265
6-3, 6-0 = 0.27
6-2, 6-0 = 0.295
6-1, 6-1 = 0.295
6-1, 6-0 = 0.325


I assume 6-0, 6-0 is a special case and probably means their new ratings must be at least .35 apart.

This would make sense if the ratings (which are groups) were used for individuals and had four significant digits (when they have 2). The reality is that two identical players cannot tie, by the rules of tennis, so common scores like split sets, tiebreaker wins, winning sets by a single break (for example 6-3) are all very common for identical players. Even winning by 2 breaks will happen to identical players (who are having better and worse days). The reason to label someone as a particular rating, say 4.0, has to do with how they played over a season (taking into account better and worse days cancelling themselves out) to get an accurate rating.

 

[JavierLW]

There have been a lotof threads over the past 12 months or so that ask people to rate players based on the perceived quality of their strokes (as seen in a video). They've been entertaining but, ultimately, very inconclusive. The missing ingredient is always competition which is, obviously, the truest test of a player's rating.

What I thought might interesting is to see how big a gap, in terms of games won, people believe there is between ratings.

So;

If player X is capable of winning no more than 0-2 games in any set against player Y (scores are indicative of their average level of play, not one playing better or worse than usual) what would be the difference in rating between them (eg: 0.5pts, 1pt, 2pts)

If player X is capable of winning no more than 4 games in any set against player Y, what would be the difference in rating between them

If player X is capable of winning at least one set against player Y, what would be the difference in rating between them.

a) Around .5
b) .25 or less (it varys)
c) .25 or less (it varys)

I would base this on what happens when two people play each other say 100 times, you may see that:

1) One player wins almost all of them
2) There are all diffrent scores from day to day, sets are won, 6-4's, and lower
3) Blowouts are not the norm, but may happen.

The USTA computer system really would only work well for someone who has a fairly good sample of matches, especially since it's based on games.

I believe that if you notice that someone is getting consistantly crushed at any level, or they are crushing everyone else at any level, then the system works well.

But when you're talking about people who are close within any given level and you add that they might only have a small sample of matches you are going to see tons of inconsistancys occur.

There isnt any real getting around that though, and I think it's still far better than letting humans decide what happens.

 

[Ace]

Another thing that causes USTA ratings to be less accurate is players who refuse to play at their own rated level, and only play "up".

Sure, some people might be bored at their own level, but you should be able to prove you don't belong there by playing at that level and winning. Often, a player only plays "up", when they wouldn't do any better at their own level anyway. Because they are playing higher-rated players, they can get more "credit" for losing at the higher level, then losing at their own level, and it actually brings their rating up.

 

[JavierLW]

This is probably what you're referring to:

http://tt.tennis-warehouse.com/archive/index.php/t-116331.html

here is the differential chart:

7-6, 7-6 = 0.06
7-6, 6-4 = 0.09
6-4, 6-4 = 0.12
6-3, 6-4 = 0.15
6-3, 6-3 = 0.18
7-5, 6-2 = 0.18
6-2, 6-3 = 0.21
6-2, 6-2 = 0.24
6-3, 6-1 = 0.24
6-1, 6-2 = 0.265
6-3, 6-0 = 0.27
6-2, 6-0 = 0.295
6-1, 6-1 = 0.295
6-1, 6-0 = 0.325


I assume 6-0, 6-0 is a special case and probably means their new ratings must be at least .35 apart.

This chart is bunk (bogus). If you follow it to the original document (which comes from someone in the Mid Atlantic Section....), you will see that in the description they address .05 as the APPROXIMATE differential per service break. (http://www.shively.net/howNTRPisCalculated.pdf)


They then go on to mention what most of us know, we dont really know what the differential is because the USTA will not publish it.

But then they went on to make that chart anyway. It's an example based on data that we do not know, it's not exactly how things work, so dont assume anything about the .35 per 6-0, 6-0 beating.

To be honest, Ive always assumed it was .5, and Ive run my own simulations where I was able to predict who will be rated out (based on limited data even) or at least who was close. But I dont know it's .5 because they wont publish that info.

So nobody can make an argument based on how the USTA system is doing it wrong because you dont have enough information to say that. People keep using hearsay or they misread articles and then they go on to base their whole argument on it.

 

[JavierLW]

Another thing that causes USTA ratings to be less accurate is players who refuse to play at their own rated level, and only play "up".

Sure, some people might be bored at their own level, but you should be able to prove you don't belong there by playing at that level and winning. Often, a player only plays "up", when they wouldn't do any better at their own level anyway. Because they are playing higher-rated players, they can get more "credit" for losing at the higher level, then losing at their own level, and it actually brings their rating up.

Actually I think that's debatable though, depending on where you think the differential should be.

If you believe that 6-0, 6-0 = .5 then you could be wrong. (I feel this way but I'll admit it's my opinion)

The reason why you could be wrong, is if you meet someone who is .5 or more below you, and you fail to drub them 6-0, 6-0, that's YOUR FAULT, and perhaps you are not really at the correct level to begin with.

If you do drub them 6-0, 6-0, then I should hope that the system just ignores that rating, since it's meaningless (since you cant beat them any better than that). Otherwise it would pull your rating down and then I would agree with you that it's bad for your rating.

If that was the case, I would agree with you, but it's really dumb if it does it that way. However again, we dont know if they do or dont so we should be careful about how we go about complaining about it.

 

[JavierLW]

Amen. Too many people on this board think that the USTA computer-rating is the holy grail of measuring tennis skills. I think everyone knows the system is not perfect and that it could be improved, however I think for the most part the system actually does what is intended of it, to match players with other players of similar skill level. Just because someone is a computer-rated 3.5, doesn't mean they are any better than 3.0s or worse than 4.0s. It just means that based on a few matches they played for 3 months out of the year, the system calculated some rating for them. Alot of factors affecting the rating could be whether they were playing doubles with a strong/weak player, whether they just started playing tennis again after years off, maybe this is their first USTA league and they mentally broke down, etc.

Others think that if someone at their same skill level beats them pretty easily, then they have been wronged and that the system is flawed.Not everyone is going to win a National Championship and many people (especially on this board) need to accept that.

We all need to take the NTRP ratings with a grain of salt. There is good reason why the pros don't use this system. Because scores don't tell as much as a Win or Loss. Fed could beat the same guy 7-6, 7-6 probably 10 times in a row yet USTA will say they are equal in skill.

Pros do not use a system that RATES their SKILL because they have surpassed beyond skill ratings at their stage in the game.

Otherwise everyone would be complaining because Federer would be the biggest sandbagger of them all.

I agree with you that the system is not perfect, and that we shouldnt take it too seriously. But that ought to go for the people who constantly appeal, or self rate at a certain level just for the sake of winning as well.

Outside of appealees and self raters though I would agree that some people are just going to be better than you are in your rating.

As a player if I play a match and I played my absolute best and still lost because the other player is better than me, I think that's awesome. I think sometimes it's actually harder to accomplish that (because it's easy to let down if you feel you are going to lose), then it is to beat someone more around your skillset.

But as a team captain, Im running a team and our goal is to win team matches and finish as high in the standings as possible. Otherwise we wouldnt play team tennis, we would just join some Open Challenge Ladder and play a lot of individual tennis (which I do a lot of). So if some team is recruiting self rates that are several levels ahead, or constantly appealing, then Im going to complain about it. And if it happens to be against the rules, then I am going to REALLY complain about it. (like if someone self rates who has a known history that would preclude them from playing 3.5)

 

[heninfan99]

I'd say the first scenario can either be a .5 or a 1.0 difference.

If player X is capable of winning no more than 0-2 games in any set against player Y (scores are indicative of their average level of play, not one playing better or worse than usual) what would be the difference in rating between them (eg: 0.5pts, 1pt, 2pts)

IMHO, this second scenario doesn't tell you much. I say that because I've taken sets off my coach when I was in the zone and he's definitely way, way above me. Everyone can play out-of-their-head once or twice a year.

If player X is capable of winning no more than 4 games in any set against player Y, what would be the difference in rating between them

These players sound equal.

If player X is capable of winning at least one set against player Y, what would be the difference in rating between them.