UZR explained

[woodchip2153]

http://sports.yahoo.com/mlb/blog/big_league_stew/post/Everything-you-always-wanted-to-know-about-UZR?urn=mlb,212311

 

I know some people have lots of questions about UZR. This is just a primer that helps explain how it's calculated and why there are large variances. There are links in the article if you want to go into more detail.

 

This is a good series by Yahoo to help explain some of the new stats.

[Prizefight]

Excellent article, thanks for posting this! :thumbsup:

[hardcorecubfan]

Simplifies it and may making the decision easier of who is going to be the defensive replacement at second for us in late innings

[Guest]

An unreliable metric is unreliable. It doesn't matter if one uses one weeks worth of data or ten years worth of data. You simply have more data that are unreliable.

 

That's the problem with UZR not sample size

[woodchip2153]

An unreliable metric is unreliable. It doesn't matter if one uses one weeks worth of data or ten years worth of data. You simply have more data that are unreliable.

 

That's the problem with UZR not sample size

 

what makes it unreliable?

[Roast]

An unreliable metric is unreliable. It doesn't matter if one uses one weeks worth of data or ten years worth of data. You simply have more data that are unreliable.

 

That's the problem with UZR not sample size

 

what makes it unreliable?

http://www.hardballtimes.com/main/blog_article/how-reliable-is-uzr/

 

There's a quick little blurb about the reliability, or lack there of, of UZR.

[Rob]

http://www.fangraphs.com/blogs/index.php/uzr-2008-to-2009

[Guest]

An unreliable metric is unreliable. It doesn't matter if one uses one weeks worth of data or ten years worth of data. You simply have more data that are unreliable.

 

That's the problem with UZR not sample size

 

what makes it unreliable?

http://www.hardballtimes.com/main/blog_article/how-reliable-is-uzr/

 

There's a quick little blurb about the reliability, or lack there of, of UZR.

That's not a bad explanation, but again, reliability problems are not solved by more observations (samples taken), they are solved by re-calibrating the measurement system. It's like stepping on a scale and having the scale weigh you differently each time. Sometimes the weight will be correct, but simply averaging all the times it is incorrect and correct doesn't give you your correct weight. The only way to do that will be to fix the scale. IMO, there's no real way to calibrate UZR. At best it might be a rough measure of something, but just watching and using good judgment is probably much better.

 

Those correlation coefficients are terrible. Anything below a .3 means nothing.

[Rob]

An unreliable metric is unreliable. It doesn't matter if one uses one weeks worth of data or ten years worth of data. You simply have more data that are unreliable.

 

That's the problem with UZR not sample size

 

what makes it unreliable?

http://www.hardballtimes.com/main/blog_article/how-reliable-is-uzr/

 

There's a quick little blurb about the reliability, or lack there of, of UZR.

That's not a bad explanation, but again, reliability problems are not solved by more observations (samples taken), they are solved by re-calibrating the measurement system. It's like stepping on a scale and having the scale weigh you differently each time. Sometimes the weight will be correct, but simply averaging all the times it is incorrect and correct doesn't give you your correct weight. The only way to do that will be to fix the scale. IMO, there's no real way to calibrate UZR. At best it might be a rough measure of something, but just watching and using good judgment is probably much better.

 

Those correlation coefficients are terrible. Anything below a .3 means nothing.

 

So you don't believe in offensive stats either?

[Guest]

That's not a bad explanation, but again, reliability problems are not solved by more observations (samples taken), they are solved by re-calibrating the measurement system. It's like stepping on a scale and having the scale weigh you differently each time. Sometimes the weight will be correct, but simply averaging all the times it is incorrect and correct doesn't give you your correct weight. The only way to do that will be to fix the scale. IMO, there's no real way to calibrate UZR. At best it might be a rough measure of something, but just watching and using good judgment is probably much better.

 

Those correlation coefficients are terrible. Anything below a .3 means nothing.

hmm..

 

As you know, there's accuracy and precision in any measurement system. The system can be accurate, but not precise. If this is the case, then additional measurements gives you increased confidence about the true mean value. The issue, of course, is determining whether or not the measurement system is accurate when you lack a good way to calibrate the system.

 

And yeah, I'm not impressed with the R-squared values for any of the metrics. I seem to remember the offensive metrics having a higher auto-correlation value than that in the studies that I've seen, but I'll admit it's been a long while since I've looked.

[Guest]

Any time there is large variability in measurements from game to game, week to week, or month to month on a performance that is relatively stable is a pretty good indication there is very little accuracy.

[Little Slide Rooter]

Life as a baseball fan was easier when I was a kid and the only stats I looked for on a baseball card were avg. and home runs for a hitter and Wins and ERA for a pitcher.

[woodchip2153]

Any time there is large variability in measurements from game to game, week to week, or month to month on a performance that is relatively stable is a pretty good indication there is very little accuracy.

 

I think that's where you're wrong. I think defensive performance does vary from play to play and from game to game just like offensive performance varies from PA to PA and from game to game. I don't think there is any way a person can prove one way or the other until we get hit f/x and fielder f/x but there is no way you can convince me otherwise. I think your opinion is shared by almost all those that don't like UZR is because they don't believe that defenders go in slumps.

[Guest]

Any time there is large variability in measurements from game to game, week to week, or month to month on a performance that is relatively stable is a pretty good indication there is very little accuracy.

 

I think that's where you're wrong. I think defensive performance does vary from play to play and from game to game just like offensive performance varies from PA to PA and from game to game.

 

We'll have to agree to disagree.

[Tyrant]

Life as a baseball fan was easier when I was a kid and the only stats I looked for on a baseball card were avg. and home runs for a hitter and Wins and ERA for a pitcher.

 

Ignorance is bliss, after all.

[Transmogrified Tiger]

Any time there is large variability in measurements from game to game, week to week, or month to month on a performance that is relatively stable is a pretty good indication there is very little accuracy.

 

I think that's where you're wrong. I think defensive performance does vary from play to play and from game to game just like offensive performance varies from PA to PA and from game to game.

 

We'll have to agree to disagree.

 

That's your argument, that defense is a static ability?

[TruffleShuffle]

there was one year where jason bay went from very good outfielder to mediocre/bad... coincidentally he had a hurt knee.

 

everyone believes in offensive slumps; i don't understand why people can't believe that players, either through injury or poor mechanics, can have a bad year defensively as well, or a good year if they improve their reads on the ball and their throwing mechanics.

[Guest]

Any time there is large variability in measurements from game to game, week to week, or month to month on a performance that is relatively stable is a pretty good indication there is very little accuracy.

 

I think that's where you're wrong. I think defensive performance does vary from play to play and from game to game just like offensive performance varies from PA to PA and from game to game.

 

We'll have to agree to disagree.

 

That's your argument, that defense is a static ability?

No that's not my argument. Static and relatively stable are much different arguments. Every player has a bad game or two in the field, but it is hardly the same as going 12 for 60 at the plate, or even 0 for 3 for that matter.

 

The within human variability of speed, hand-eye-coordination needed to catch a baseball, and arm strength is an order of magnitude or more smaller than hitting a baseball. The point being, the large fluctuations from game to game or year to year mean the system isn't very good. In addition, the differences between what is an average and a great defender is not great. It simply isn't. So when UZR says Fukudome was 15 "runs" below the average CF it doesn't pass the smell test.

[Transmogrified Tiger]

Hitting may require some more refined reflexes that cause fluctuations, but that's all there is, you swing. Defensively there's a much wider range of what you have to do physically, and frankly there's much more time during each defensive play to do something wrong.

[Guest]

I want to say that I'm all for measurement in baseball. I think it adds to the game and in important ways. The problems with quantifying defense with any accuracy are probably unsolvable. First, you have positioning before the pitch, then you have a pitcher missing his spot, then the SD of ability being so narrow, and on and on it goes.

 

Long story short, UZR isn't any better then subjective analysis.

[TruffleShuffle]

I want to say that I'm all for measurement in baseball. I think it adds to the game and in important ways. The problems with quantifying defense with any accuracy are probably unsolvable. First, you have positioning before the pitch, then you have a pitcher missing his spot, then the SD of ability being so narrow, and on and on it goes.

 

Long story short, UZR isn't any better then subjective analysis.

 

that's probably true if you can sit and watch a guy play every game, but most people probably don't have time for that, and uzr is much better than taking the word of one of your scouts who sat and watched a guy play shortstop for 4 games or someone who watched a handful of games on film.