NTRP Win-Loss Adjustment Formula

[ronray43]

With the new ratings coming out, I thought I'd share what I forwarded to the national USTA office a few years back. The below is a win-loss adjustment formula to be applied to end of year ratings using the current NTRP algorithm and is intended to allow win-loss record to be taken into account when publishing final ratings. Under the algorithm below, the maximum adjustment is .1, but this factor can be set to anything. BTW, the national office rejected the proposal because it would "put too much emphasis on winning and losing", even though I explained the .1 constant (maximum impace of win-loss record) can be lowered as desired. Regardless, here's one way to integrate win-loss into NTRP:

NTRP Win/Loss Adjustment Formula

Assumptions:

- Used to adjust end of year NTRP ratings as computed by existing formula

- Max NTRP adjustment for any player not to exceed .10 (.10 used as an example, can be any number)

- Appeal rules for adjusted end of year rated remain as-is

- No adjustment applies to victories over players with lower overall NTRP or losses to players with higher overall NTRP (4.0 defeats 3.5, or 4.5 loses to 5.0)


Variables and values assigned to variables:

X = end of year NTRP rating to hundredth point under existing NTRP formula

Y = adjusted end of year NTRP rating

A = number of wins over players with same overall NTRP rating (.01 adjustment per win; can be any number)

B = number of losses to players with same overall NTRP rating (.01 adjustment per loss; can be any number)

C = number of wins over players with overall NTRP rating above winner’s overall NTRP rating, e.g. a win by a 3.5 over a 4.0 (.02 adjustment per win)

D = number of losses to player with overall NTRP rating below loser’s overall NTRP rating, e.g. a loss to a 3.5 by a 4.0 (.02 adjustment per loss)

Formula:

X + (A x .01) – (B x .01) + (C x .02) – (D x .02) = Y

For X > Y, if X – Y > .1, then Y = X - .1

For Y > X, if Y – X > .1, then Y = X + .1

Examples:

Player 1 rated 4.0 during the year has end of year rating of 3.56 using existing algorithm. Unadjusted new year overall NTRP is 4.0 and outside appeal eligibility. Player has win-loss record of 1 win and 4 losses to 4.0s, and a record of 1 win and 2 losses to 3.5s.

Adjusted rating for player 1 becomes: 3.51; overall 4.0; appeal eligible

3.56 + (1 x .01) – (4 x .01) + (1 x .02) – (2 x .02) = 3.51

Since X > Y and 3.56 – 3.51 < .1, final adjusted end of year NTRP rating remains 3.51






Player 2 rated 4.0 during the year and has end of year rating of 3.96 using existing algorithm. Unadjusted new year overall NTRP is 4.0 and within appeal eligibility. Player has a win-loss record of 20 wins and 2 losses, all to players of same overall NTRP rating.

Adjusted rating for player 2 becomes: 4.06; overall 4.5; ineligible for appeal

3.96 + (20 x .01) – (2 x .01) = 4.14; however, since Y > X and Y – X > .10, adjusted rating becomes 4.06 (X + .10). Player becomes overall NTRP 4.5 and outside appeal criteria.


Player 3 rated 4.0 during the year and has end of year rating of 3.75 using existing algorithm. Unadjusted new year overall NTRP is 4.0 and outside appeal guidelines. Player has win loss record of 6 wins and 4 losses, all to players of the same overall NTRP.

Adjusted rating for player 3 becomes: 3.77; overall 4.0; ineligible for appeal

3.75 + (6 x .01) – (4 x .01) = 3.77

Since Y > X and Y – X < .1; adjusted NTRP remains 3.77

 

[tennismonkey]

does anybody have the cliff notes for this?

 

[gmatheis]

so if you are a 4.0 play a bunch of people who are bad 4.0s during your league season you would have a much higher chance of being bumped just because your opponents happened to be bad.

Sorry but USTA is right to reject this idea, it sucks!

 

[ronray43]

Which is why there is a .1 constant applied to how much your win-loss ratio can raise your NTRP in this particular formula. Additionally, this constant could be reduced to .05 or anything else just in case an individual goes 20-0 and just happened to play all lower end players.

 

[schmke]

so if you are a 4.0 play a bunch of people who are bad 4.0s during your league season you would have a much higher chance of being bumped just because your opponents happened to be bad.

Sorry but USTA is right to reject this idea, it sucks!

On the surface, this proposal may seem logical and straightforward, but yes, the flaw is that every win/loss against a same NTRP opponent counts the same and similarly every win/loss against a player with an NTRP higher/lower counts the same. One might argue it all comes out in the wash, but many players don't play enough matches in a year for things to average out and you'd have situations where someone plays court 3 doubles all year and wins, but against weak competition.

It also has issues with doubles in that a weaker player may play with a very good partner and as a result have a very good record because of that, not because of their own play, particularly if the opposition isn't that strong.

To be fair, the proposal does provide for limiting the adjustment, presumably to avoid really whacky things happening in these situations, but if you limit the adjustment too much, what is the point of doing the adjustment? If the win/loss adjustment is sound and equitable, it shouldn't need an artificial limit placed on it. To me, this adjustment approach doesn't seem sound.

If you want to factor wins/losses in, it should be done as part of the individual match rating calculation, not as some end of year override like this proposal. If I were doing the NTRP all over myself, I'd probably include wins/losses in match calculations, but alas the USTA hasn't asked me to do that, well yet anyway

 

[g4driver]

Or they could just count every match instead. Naw, that makes to much sense.

8 matches counted toward a DNTRP (over a two month period) out of 55 USTA league matches played.

And Level 1 and 2 Tournaments never have any B rated players do they ?