UTR Statistical Analysis done? Like 538?

[Badmrfrosty]

Has there been any statistical analysis done on what the odds should be for a given UTR number facing another given UTR number? I thought I read UTR was based on ELO which should make it easy to come up with odds. Obviously at the elite level you could calculate this since there are posted odds.

I'm curious if anyone knows of a rule of thumb formula. Like if a UTR X plays UTR Y what are the odds UTR X wins.

You probably have to know what 538 is to even understand what I'm talking about. And I'm not a 538 fan.

 

[schmke]

You are right, it should be doable and I'm sure at the elite level where odds are established, those settings the odds are doing so whether with UTR or algorithms of their own.

I have done this sort of analysis for my Estimated Dynamic NTRP Ratings and in general it falls out like this for the percent of the time the higher rated player/pair wins:

< 0.1 - 56-58%
0.1-0.2 - 68-70%
0.2-0.3 - 78-82%
0.3-0.4 - 85-90%
0.4-0.5 - 90-95%

 

[jmnk]

Has there been any statistical analysis done on what the odds should be for a given UTR number facing another given UTR number? I thought I read UTR was based on ELO which should make it easy to come up with odds. Obviously at the elite level you could calculate this since there are posted odds.

I'm curious if anyone knows of a rule of thumb formula. Like if a UTR X plays UTR Y what are the odds UTR X wins.

You probably have to know what 538 is to even understand what I'm talking about. And I'm not a 538 fan.

It would be easy if we knew exactly the ELO formula that UTR uses. ELO formula allows for a number of parameters that are essentially empirically picked to suit a given scenario: they have parameters for chess, slightly different ones for soccer, and so on.

 

[Badmrfrosty]

It would be easy if we knew exactly the ELO formula that UTR uses. ELO formula allows for a number of parameters that are essentially empirically picked to suit a given scenario: they have parameters for chess, slightly different ones for soccer, and so on.

I'm not even sure you need that. If you could download enough results you could calculate the odds and confidence interval for given absolute value ranking and spread between players.

I'm assuming the odds of a 14 beating a 15 is different than 6 beating a 7, but maybe not. Would be interesting to look at the data or if anyone has done the work.

 

[5sets]

Hey @badmrfrosry please enlighten those of us who didn’t get through the 8th grade what 538 is so we can enjoy the convo, thanks


Sent from my iPhone using Tapatalk

 

[S&V-not_dead_yet]

Hey @badmrfrosry please enlighten those of us who didn’t get through the 8th grade what 538 is so we can enjoy the convo, thanks

https://en.wikipedia.org/wiki/Nate_Silver

Silver established his own blog, FiveThirtyEight.com. Often colloquially referred to as just 538, the website takes its name from the number of electors in the United States electoral college.[34]

 

[Badmrfrosty]

Holy crap look what I found. Someone has written the exact paper I was looking for. Well if any other math nerds/gamblers out there in the future are reading this-This should help them out.

A Bayesian regression approach to handicapping tennis players based on a rating system

"Therefore, our goal is to develop a mapping between the Universal Tennis Rating (UTR) system and the MDP-based handicaps, so that two amateur players can determine an appropriate handicap for their match based only on their UTRs."

 

[kevrol]

You are right, it should be doable and I'm sure at the elite level where odds are established, those settings the odds are doing so whether with UTR or algorithms of their own.

I have done this sort of analysis for my Estimated Dynamic NTRP Ratings and in general it falls out like this for the percent of the time the higher rated player/pair wins:

< 0.1 - 56-58%
0.1-0.2 - 68-70%
0.2-0.3 - 78-82%
0.3-0.4 - 85-90%
0.4-0.5 - 90-95%

Dang. That's pretty disheartening.

 

[S&V-not_dead_yet]

Dang. That's pretty disheartening.

You can choose to be disheartened or you can take the #s with a big grain of salt.

 

[OnTheLine]

You can choose to be disheartened or you can take the #s with a big grain of salt.

OR, take a look at your wins in matches you "should" have lost and be extra proud.

 

[OnTheLine]

Holy crap look what I found. Someone has written the exact paper I was looking for. Well if any other math nerds/gamblers out there in the future are reading this-This should help them out.

A Bayesian regression approach to handicapping tennis players based on a rating system

"Therefore, our goal is to develop a mapping between the Universal Tennis Rating (UTR) system and the MDP-based handicaps, so that two amateur players can determine an appropriate handicap for their match based only on their UTRs."

And I can't believe I read the whole thing. Yes, the biggest challenge to their initial system is using dynamic point use in their handicapping. But I think that is almost the only way to work the handicap in tennis. Very interesting read.

 

[S&V-not_dead_yet]

Okay, m


OR, take a look at your wins in matches you "should" have lost and be extra proud.

I can't find any: OK, now I'm disheartened.

 

[OnTheLine]

I can't find any: OK, now I'm disheartened.

LOL ... I can find losses (or wins with way too tight a scoreline) on matches I should have won handily ... now I am suicidal!

 

[schmke]

Dang. That's pretty disheartening.

Hey, none of what I showed said it is a 100% chance the higher rated player/pair wins! There is always a chance!

 

[schmke]

LOL ... I can find losses (or wins with way too tight a scoreline) on matches I should have won handily ... now I am suicidal!

The cool thing I do with my reports is show a player's record when favored to win or the underdog. A nifty way to see if that gaudy record is really just was expected or if you are pulling upsets to get it. Or sometimes that losing record really isn't so bad if you win those you are supposed to and are still pulling some upsets.

 

[Badmrfrosty]

@OnTheLine
If you liked that this is a simpler read just showing how a small difference in your probability to win a point greatly affects chance to win match. See chart at bottom of post.

Probability to win point 0.54

Equals Prob to Win Game 0.599048
Set 0.763430
Match 0.910138

So just a 54% chance to win a point equals 91% chance you will win match.


http://datagenetics.com/blog/august12018/index.html

 

[OnTheLine]

The cool thing I do with my reports is show a player's record when favored to win or the underdog. A nifty way to see if that gaudy record is really just was expected or if you are pulling upsets to get it. Or sometimes that losing record really isn't so bad if you win those you are supposed to and are still pulling some upsets.

I have finished the 40/18+ seasons with a near .500 record and according to TR my scorelines show nothing but competitive (read close) matches:

Average first set:
Me 4.39 Opponent 4.64

Average 2nd:
Me: 4.55
opponent: 4.46

Average 3rd (17 3rd sets this year!)
Me .62
Opponent: .38

Yeah, I still have your email address tucked away .... It may be time to do a refresh, oddly, I didn't do my own report last year did I?

Data is fun ... unless you obsess about it or use it for evil.

 

[Badmrfrosty]

It would be cool if UTR published their logic so you could know your expected score. My understanding is that they calculate your Expected game wins and then if you exceed that you move up if less than that you move down. (Obviously a simplification of algo)

Just like golf I know what my expected score is and can tell how I played based off that and adjusting for course slope and conditions.

 

[S&V-not_dead_yet]

Data is fun ... unless you obsess about it or use it for evil.

So you're saying evil people don't have fun with data?

 

[OnTheLine]

So you're saying evil people don't have fun with data?

well done... early morning imprecision at its finest.

Yes. Evil people are having fun (I hope) with data!

Should have said Data is good unless you obsess about it or use it for evil.

 

[ChaelAZ]

Hey, none of what I showed said it is a 100% chance the higher rated player/pair wins! There is always a chance!

 

[MathGeek]

Anyone with the same UTR as me (5) has about a 90% chance of beating me.

My UTR is kinda inflated by a couple factors:
1. Some youths I performed well against have gotten lots better since I played them.
2. I had an incredible hot streak against better players rusty from their winter layoffs.

Now that it's summer and no one is rusty, I'm gonna be toast.