trackmyrating.com

The stronger/weaker split compares your rating to the average rating of your opponents, and by default it runs over every match including mixed. In 8.0 mixed you are partnering a 3.5 and facing a pair averaging around 4.0, so that average sits below you by arithmetic rather than because the competition was weaker. There is a checkbox on your page, "Only matches that count toward my USTA rating", and ticking it should give you a truer read.

Mixed does not generate an adult NTRP rating unless mixed is all you play, USTA Reg. 1.04F(1)d, and my engine gives it zero weight to match. So if nearly all your matches are 8.0 mixed, almost none of them are moving your number. What you are seeing comes from the few that do count plus your last published rating. I keep it in there and generate expectations in the match just so you know how you do. Even if you enter a friendly match by hand it'll tell you what your expected result should have been.

The one thing that kind of bites at 4.5. The higher you go the fewer people there are at your level, so more of your matches are against players below you, and the further above them you are the more you are expected to win by just to hold station. Beating a weaker field 6-3 6-3 might not be a neutral result, it could slightly be below par. That is the mechanic rather than a quirk of mine, and it means your rating should track the competition you actually face.

I had that calibration wrong until recently and 4.5 was the level it hurt most, sitting about half a level low. @Jack the Hack turned me onto it. That is fixed and 4.5 now sits where it should be. So if it still looks wrong to you once you have the counting-only view on, send it through the feedback button and I will chase your specific case.

And the individual match ratings you asked about are in, so you can see what each performance was worth on its own. That was your suggestion so thanks!
For mixed 8.0, the level of the opponent (for purpose of rating calc) should not be the average of two opponents.

The opponent level should be equal to the sum of the two opponent ratings minus the partner rating.

So, for example, if a 4.5c male (4.3 dynamic) player partners with a 3.5c female (3.1 dynamic) partner against two 4.0 players (3.8 and 3.7 dynamic), the opponent level for the 4.5c would be:

3.8 + 3.7 - 3.1 =4.4.
A slightly stronger opponent.

Is this what you did? It’s how UTR does it now, after I caught a bug in the UTR algorithm a few years ago that didn’t handle partner adjustment correctly, and I reported it. The bug was obvious for players that only played mixed.

I caught the bug because UTR said my average opponent level in 2022 was 1 UTR point above my rating, yet I went 10-0 in usta 8.0 that season winning over 60% of games. An impossible result.
 
Last edited:
For mixed 8.0, the level of the opponent (for purpose of rating calc) should not be the average of two opponents.

The opponent level should be equal to the sum of the two opponent ratings minus the partner rating.

So, for example, if a 4.5c male (4.3 dynamic) player partners with a 3.5c female (3.1 dynamic) partner against two 4.0 players (3.8 and 3.7 dynamic), the opponent level for the 4.5c would be:

3.8 + 3.7 - 3.1 =4.4.
A slightly stronger opponent.

Is this what you did? It’s how UTR does it now, after I caught a bug in the UTR algorithm a few years ago that didn’t handle partner adjustment correctly, and I reported it. The bug was obvious for players that only played mixed.

I caught the bug because UTR said my average opponent level in 2022 was 1 UTR point above my rating, yet I went 10-0 in usta 8.0 that season winning over 60% of games. An impossible result.
That is exactly right, and it is what the rating engine does. But you found a real bug in what the site shows you.

The engine compares the two sides rather than averaging your opponents, so your partner's rating is already inside the number you get measured against. Run your rule on both players and they land in the same place: the 4.5 is measured against 4.4 and comes up a tenth short, and applying your formula to his partner she is measured against 3.8 + 3.7 − 4.3 = 3.2 and comes up a tenth short too. That equality is the point. My side comparison has the same property, written per player instead of per team, so mine is exactly half of yours with the same sign and the same value for both partners. I checked it against your example this morning and both of them move by the identical amount.

Where you caught m bug is the panel that splits your record into matches against stronger, same level and weaker opponents. That one takes the plain mean of the two opponents with no partner term, which is (I think) the UTR bug you are describing. Someone carrying a weak partner gets told they beat weaker players. I measured it after reading your post. Across a sample of 104,555 doubles matches, 19.4 percent were sitting in the wrong bucket, and where your partner is more than half a level away from you (your situation) it is 40.3 percent. So on mixed 8.0 it is wrong about two times in five. Its on the bug crushing list.

Thank you for this. You have next level bug hunting chops.
 
That is exactly right, and it is what the rating engine does. But you found a real bug in what the site shows you.

The engine compares the two sides rather than averaging your opponents, so your partner's rating is already inside the number you get measured against. Run your rule on both players and they land in the same place: the 4.5 is measured against 4.4 and comes up a tenth short, and applying your formula to his partner she is measured against 3.8 + 3.7 − 4.3 = 3.2 and comes up a tenth short too. That equality is the point. My side comparison has the same property, written per player instead of per team, so mine is exactly half of yours with the same sign and the same value for both partners. I checked it against your example this morning and both of them move by the identical amount.

Where you caught m bug is the panel that splits your record into matches against stronger, same level and weaker opponents. That one takes the plain mean of the two opponents with no partner term, which is (I think) the UTR bug you are describing. Someone carrying a weak partner gets told they beat weaker players. I measured it after reading your post. Across a sample of 104,555 doubles matches, 19.4 percent were sitting in the wrong bucket, and where your partner is more than half a level away from you (your situation) it is 40.3 percent. So on mixed 8.0 it is wrong about two times in five. Its on the bug crushing list.

Thank you for this. You have next level bug hunting chops.
You are scoring a much higher customer service rating than UTR!

After I reported my “mathematically impossible” record to UTR (having lower rating than my average opponent despite winning all my matches) due this same bug, I got this response by email (paraphrasing):

“We have forwarded your inquiry to the data team”

A few days later:

“The data team has reviewed your inquiry and determined that your rating is correct.”

Hmm…

Then 3 months later:

My doubles UTR suddenly changed from “100% reliability” 6.8 UTR to “100% reliability” 8.5 UTR.

Bug fixed, but someone in the UTR data team didn’t want to own the mistake!

That episode marked the end of my UTR power subscription.
 
Love the site... the interface, graphics and match ratings. Maybe I missed it the first time I looked, but the "who was rated what?" function is really cool and helpful.
 
Mixed does not generate an adult NTRP rating unless mixed is all you play, USTA Reg. 1.04F(1)d, and my engine gives it zero weight to match. So if nearly all your matches are 8.0 mixed, almost none of them are moving your number.
Would it be possible for your site to estimate players' adult rating and mixed rating separately? I know that the mixed rating is ignored by USTA unless it's the only one, but I think a lot of players would find value in knowing what their mixed rating would be if those were their only matches. It would also help mixed captains scouting how well players do specifically in mixed.
 
Would it be possible for your site to estimate players' adult rating and mixed rating separately? I know that the mixed rating is ignored by USTA unless it's the only one, but I think a lot of players would find value in knowing what their mixed rating would be if those were their only matches. It would also help mixed captains scouting how well players do specifically in mixed.
So now does everyone have an ntrp "mixed dynamic" rating, that only uses mixed games and ratings from mixed games, and a separate dynamic rating that uses same gender and only same gender matches? I would be interested to see how those compare.
 
Following up on a few things.

@travlerajm hey, I'll take the early win...being responsive is tough, but people calling out your bugs is a gift. That opponent level fix is live now. Chasing it turned up something bigger that I did not know about when I replied to you so double thanks.

@NoahColorado and @Captain Crunch , thank you both. @Captain Crunch , you did not miss it, the who-was-rated-what view is fairly new, so it is good to know that is the part that landed.

@TennisOTM , a separate mixed rating is not there and it certainly could be. At the moment mixed matches get a match rating and then sit outside the adult number, which is USTA's rule rather than my choice. A mixed only dynamic is a different question and a fair one. I have about 1.7 million mixed match rows and roughly 47,000 players with ten or more mixed matches, so there is enough to give a regular mixed player a real number. It is on the list. Great suggestion.

@schmke good to know you already report both on @Moon Shooter 's question.

@Garmadon pay the man, he's worth it! (Assuming you're talking about @schmke). :D
 
Back
Top