Not to be whiny, but can we just cut Richie Martin loose already?

[weams]

Just now, Frobby said:

No, I meant the defensive ones.   

Well. They have done nothing to improve UZR so I don;t know what to say then. 

[weams]

1 minute ago, Frobby said:

No, I meant the defensive ones.   

Is this another above replacement chicken and egg?

[weams]

3 minutes ago, Frobby said:

No, I meant the defensive ones.   

And I mostly agree with Lichtman 

[awannabegeek]

18 hours ago, Frobby said:

I don’t think they’re lazy or sloppy.    There are smart people working as hard as they can to make this stuff as accurate as they can.    But sometimes, it just doesn’t jibe with what we think we see.    Could be we’re wrong,  could be some flaws in their system.

I’ll say this: it used to be a lot more common for the different advanced metrics to disagree with each other.    These days, they almost always agree in direction, but with variances in order of magnitude.  I don’t find as many cases where I strongly disagree with the numbers as I used to.    But Martin is definitely one that has me scratching my head.  

It's not supposed to jibe with what you see.  That is the entire point of statistical analysis.  What we see is commonly shrouded in a number of different cognitive biases.  Whether you think the numbers are valid or not is one thing but the problem with a lot people who question quantitative analysis is the belief that because it doesn't match your eyes it is invalid, it's the point.   

[Frobby]

12 minutes ago, awannabegeek said:

It's not supposed to jibe with what you see.  That is the entire point of statistical analysis.  What we see is commonly shrouded in a number of different cognitive biases.  Whether you think the numbers are valid or not is one thing but the problem with a lot people who question quantitative analysis is the belief that because it doesn't match your eyes it is invalid, it's the point.   

When a statistic doesn’t jibe with what you see, there are two explanations: what you see is misleading, or there is something wrong with how the stat is compiled.   It’s foolishly stubborn to believe it’s always the same one of those two.    Especially when you’ve got defensive metrics that disagree with each other.   

[Frobby]

18 hours ago, weams said:

Well. They have done nothing to improve UZR so I don;t know what to say then. 

There have been significant changes in how UZR is determined.    Whether those changes are improvements is debatable, I guess.  https://blogs.fangraphs.com/instagraphs/2018-uzr-and-uzr-update/

[awannabegeek]

1 minute ago, Frobby said:

When a statistic doesn’t jibe with what you see, there are two explanations: what you see is misleading, or there is something wrong with how the stat is compiled.   It’s foolishly stubborn to believe it’s always the same one of those two.    Especially when you’ve got defensive metrics that disagree with each other.   

It is even more foolishly stubborn to believe because statistic doesn't match your eye test it must be inaccurate....That seems to be the argument of most anti-stat guys.  

[Sessh]

Quantitative analysis should always be questioned just like anything else. Not doing so is, IMO, naive. If such analysis can't hold up to criticism and questioning, then it's not much of an analysis. Besides, that analysis was compiled by other humans with the same potential for biases and have their own ideas about what should and should not be included in the analysis. Work done by humans is likely just as prone to flaws as the humans that did the work. That said, I'm ok with some of the offensive statistics, but have never been all that impressed with the defensive ones though I did try. They need a lot of work.

[Frobby]

Just now, awannabegeek said:

It is even more foolishly stubborn to believe because statistic doesn't match your eye test it must be inaccurate....That seems to be the argument of most anti-stat guys.  

I agree with you there.     The new OF stats that statcast has developed that measure how good a jump an OF gets etc. are quite useful because they expose guys who make make a lot of catches that look nice, but who don’t really cover the ground that well.

[awannabegeek]

3 minutes ago, Sessh said:

Quantitative analysis should always be questioned just like anything else. Not doing so is, IMO, naive. If such analysis can't hold up to criticism and questioning, then it's not much of an analysis. Besides, that analysis was compiled by other humans with the same potential for biases and have their own ideas about what should and should not be included in the analysis. Work done by humans is likely just as prone to flaws as the humans that did the work. That said, I'm ok with some of the offensive statistics, but have never been all that impressed with the defensive ones though I did try. They need a lot of work.

My post is not to say quantitative analysis is flawless, is that the counter argument to quantitative analysis' accuracy is not "well my eyes see something different so the stats are wrong", that is the point.  Humans of all sorts are prone to cognitive biases but the data scientists which help build these stats know how to control for it and the error rate due to their cognitive bias is I'm going to bet a lot smaller than the average baseball fan.  

[Sessh]

3 minutes ago, Frobby said:

I agree with you there.     The new OF stats that statcast has developed that measure how good a jump an OF gets etc. are quite useful because they expose guys who make make a lot of catches that look nice, but who don’t really cover the ground that well.

They are nice stats to have for sure. I guess to me, I look at it like.. either the guy can makes plays on defense or he can't. If he can't, it's pretty obvious. I don't care if he's diving all over the place or not. All I care about is whether or not he can make the plays. Santander supposedly gets a bad jump on balls, but he's made some nice plays out there and really hasn't looked nearly as bad as I expected him to look. Now, being able to make plays despite getting a bad jump may indicate a guy that has a limited shelf life out there as he ages. That I can buy.

[Frobby]

1 minute ago, awannabegeek said:

My post is not to say quantitative analysis is flawless, is that the counter argument to quantitative analysis' accuracy is not "well my eyes see something different so the stats are wrong", that is the point.  Humans of all sorts are prone to cognitive biases but the data scientists which help build these stats know how to control for it and the error rate due to their cognitive bias is I'm going to bet a lot smaller than the average baseball fan.  

The big issue I have with the advanced defensive metrics is lack of transparency.    I’d love to see UZR broken down into individual plays (that’s how it’s built, after all) and be able to see the plays that got a significantly positive or negative score.   But it’s just a number on a spreadsheet somewhere.

[awannabegeek]

3 minutes ago, Frobby said:

The big issue I have with the advanced defensive metrics is lack of transparency.    I’d love to see UZR broken down into individual plays (that’s how it’s built, after all) and be able to see the plays that got a significantly positive or negative score.   But it’s just a number on a spreadsheet somewhere.

That's fair healthy skepticism.  I can't argue with that.  I think they're the best we (fans) have at the moment, they're not perfect but better than our eyes (for the most part).  Most teams to my knowledge use their own proprietary metrics which are probably (definitely?) more accurate than UZR.

Edited July 21, 2019 by awannabegeek

[weams]

9 minutes ago, awannabegeek said:

My post is not to say quantitative analysis is flawless, is that the counter argument to quantitative analysis' accuracy is not "well my eyes see something different so the stats are wrong", that is the point.  Humans of all sorts are prone to cognitive biases but the data scientists which help build these stats know how to control for it and the error rate due to their cognitive bias is I'm going to bet a lot smaller than the average baseball fan.  

Many defensive statistics require human input, have bias, are wrong. 

[awannabegeek]

6 minutes ago, Sessh said:

They are nice stats to have for sure. I guess to me, I look at it like.. either the guy can makes plays on defense or he can't. If he can't, it's pretty obvious. I don't care if he's diving all over the place or not. All I care about is whether or not he can make the plays. Santander supposedly gets a bad jump on balls, but he's made some nice plays out there and really hasn't looked nearly as bad as I expected him to look. Now, being able to make plays despite getting a bad jump may indicate a guy that has a limited shelf life out there as he ages. That I can buy.

I don't understand this as a rebuttal to defensive metrics? All you care about is if a guy catches a ball? What about the balls that drop in that another more capable outfielder would have caught due to better instincts and\or athleticism?