#### Scoop

##### Rookie
Let's assume this scenario:

Player A - at the very top of the 4.0 ratings
Player B - new bump up in 2020 to 4.0. He works very hard and is now "better" than player A despite the alleged lower NTRP rating.

Player A and B play all of their doubles matches together in 2020 and neither play singles.

Is there any way for the ratings system to have player B "catch up" to player A for year end ratings?

Thanks!

##### Legend
Honestly I don't know. But it is hard to see how that could happen if you're always on the same team. How can they say who makes the bigger contribution?

Also, I don't like having a partner who is trying to show that he makes a bigger contribution. For example I've had doubles partners who upon winning a point where they received serve, announce, "I got mine." Are we a team, or are you playing against me as well as our opponents?

#### therecanbeonlyone

##### Semi-Pro
If they only play doubles with each other, then they will maintain their rating difference (dntrp) all year. The effect of this is that player A will get closer to a bump up because of player B’s improvement.

##### Talk Tennis Guru
If they only play doubles with each other, then they will maintain their rating difference (dntrp) all year. The effect of this is that player A will get closer to a bump up because of player B’s improvement.

I don't think that's correct. At least in UTR, which is based on the same type of math that the Elo chess rating system is and I'm assuming NTRP is similar, different rated players have different expected win rates against the same player.

So let's say A is 3.99 and B is 3.50. They play doubles together against a 3.75 pair. For simplicity's sake, let's say they win every match [the same concept applies regardless of who wins and by how much because it's the relative difference in expected performance between A and B that matters, not the absolute result].

For any given match, B's relative performance is higher than A's because B is lower-rated and therefore is expected to do worse against the opposition than A [the fact that they're playing doubles together and therefore can't have a different outcome is not part of the formula, I don't think].

If A's rating goes up, B's will go up more.
If A's rating goes down, B's will go down less [it's even possible that A's will drop while B's will rise: if they play a team rated 3.75 and barely win 7-6 7-6, A's rating will likely drop because he was supposed to have won easily whereas B's will probably rise because he was supposed to lose]

This is all based on the assumption that NTRP is like UTR in how it accounts for expected performance. UTR has published their algorithm; USTA has not.

#### denoted

##### Semi-Pro
I've always assumed, for lack of seeing how it could be done otherwise, that doubles teams' dynamic ranking is averaged (mean). So 3.50/3.99 ~= 3.75 vs. two 3.75/3.75, predicted draw.

#### BallBag

##### Professional
I don't think that's correct. At least in UTR, which is based on the same type of math that the Elo chess rating system is and I'm assuming NTRP is similar, different rated players have different expected win rates against the same player.

So let's say A is 3.99 and B is 3.50. They play doubles together against a 3.75 pair. For simplicity's sake, let's say they win every match [the same concept applies regardless of who wins and by how much because it's the relative difference in expected performance between A and B that matters, not the absolute result].

For any given match, B's relative performance is higher than A's because B is lower-rated and therefore is expected to do worse against the opposition than A [the fact that they're playing doubles together and therefore can't have a different outcome is not part of the formula, I don't think].

If A's rating goes up, B's will go up more.
If A's rating goes down, B's will go down less [it's even possible that A's will drop while B's will rise: if they play a team rated 3.75 and barely win 7-6 7-6, A's rating will likely drop because he was supposed to have won easily whereas B's will probably rise because he was supposed to lose]

This is all based on the assumption that NTRP is like UTR in how it accounts for expected performance. UTR has published their algorithm; USTA has not.
UTR has published their algorithm? Only thing I read about UTR doubles algorithm is that they average the ratings for the dubs team.

##### Talk Tennis Guru
UTR has published their algorithm? Only thing I read about UTR doubles algorithm is that they average the ratings for the dubs team.

This is an example:

It's not the actual statistical calculations but it does explain the high-level and what factors affect the outcome.

By comparison, NTRP is a black box.

#### OnTheLine

##### Hall of Fame
In the FAQ section on USTA website re NTRP it used to state that the difference between partners would be maintained. That language is no longer there but replaced with this:

Does the dynamic calculation treat doubles partners differently?

No, the dynamic calculation does not treat doubles partners differently. The system knows the players’ information and calculates the dynamics based on the opponent(s) dynamics and the scores. The effect of the outcome of the match is applied equally to both players.

#### Justagamefool

##### New User
If they only play doubles with each other, then they will maintain their rating difference (dntrp) all year. The effect of this is that player A will get closer to a bump up because of player B’s improvement.
This is correct for NTRP ratings.

##### Talk Tennis Guru
Well, I was wrong, at least for sites like TR that attempt to mimic NTRP: I checked several matches I played with lower-rated partners and in every case, the match rating relative to the final rating delta was the same for me and my partner. I still don't see how that makes sense but there you have it.

#### Scoop

##### Rookie
Thanks for all the replies. Definitely doesn't seem to make sense but that's about par for NTRP.

#### schmke

##### Legend
I don't think that's correct. At least in UTR, which is based on the same type of math that the Elo chess rating system is and I'm assuming NTRP is similar, different rated players have different expected win rates against the same player.

So let's say A is 3.99 and B is 3.50. They play doubles together against a 3.75 pair. For simplicity's sake, let's say they win every match [the same concept applies regardless of who wins and by how much because it's the relative difference in expected performance between A and B that matters, not the absolute result].

For any given match, B's relative performance is higher than A's because B is lower-rated and therefore is expected to do worse against the opposition than A [the fact that they're playing doubles together and therefore can't have a different outcome is not part of the formula, I don't think].

If A's rating goes up, B's will go up more.
If A's rating goes down, B's will go down less [it's even possible that A's will drop while B's will rise: if they play a team rated 3.75 and barely win 7-6 7-6, A's rating will likely drop because he was supposed to have won easily whereas B's will probably rise because he was supposed to lose]

This is all based on the assumption that NTRP is like UTR in how it accounts for expected performance. UTR has published their algorithm; USTA has not.
What you outlined is incorrect for NTRP. It is in effect the average of each doubles team's ratings that are used, and partners move up and down together. I can't say for certain what UTR does, have you looked at the algorithm and confirmed what you said?

#### schmke

##### Legend
Thanks for all the replies. Definitely doesn't seem to make sense but that's about par for NTRP.
While the different answers might confuse, what NTRP does is pretty clear. The gap between player's ratings is preserved when they play doubles together. So if player B only plays with player A (and vice versa), the gap between the two will stay the same. Doing anything else is arguably wrong, as given just a result, an algorithm can't say if the result is due more to player A or player B. While in your scenario it appears player B is underrated and having their rating unfairly suppressed, the scenario could equally likely be that player A is underrated and is carrying player B and the gap should be larger. Having the algorithm increase or decrease the gap with no other data (no matches by either player with other partners or in singles) would be hard to justify.

##### Talk Tennis Guru
What you outlined is incorrect for NTRP. It is in effect the average of each doubles team's ratings that are used, and partners move up and down together. I can't say for certain what UTR does, have you looked at the algorithm and confirmed what you said?

In post #10 I recognized I was wrong at least about estimates like TR and I'll take your word for it on anything NTRP.

I could only find a high-level description of UTR, not the algorithm itself and I can't see detailed #s for anyone except myself so I can't do the comparison that I did on TR.

#### Traffic

##### Hall of Fame
Let's assume this scenario:

Player A - at the very top of the 4.0 ratings
Player B - new bump up in 2020 to 4.0. He works very hard and is now "better" than player A despite the alleged lower NTRP rating.

Player A and B play all of their doubles matches together in 2020 and neither play singles.

Is there any way for the ratings system to have player B "catch up" to player A for year end ratings?

Thanks!
No. Unless player B plays a bunch of matches on their own or with other partners with lower rating and continue to win.

Replies
12
Views
2K
Replies
7
Views
963
Replies
28
Views
4K
Replies
14
Views
4K
Replies
1
Views
818