Impact of Defense

[jimmer]

 

This is my point of stats. We often use the stats that benefit our argument and ignore the same stat if it doesn't support our argument.

it helps if we know which stats can also be used. Like, in this case, Off-WAR can't be used like the poster ken did because we are comparing Mauer to Dozier specifically, not how Mauer compared to other 1Bs versus how Dozier compared to other 2Bs. WAR is position specific.

 

And I prefer to look for the truth, whatever that may be.  I don't decide who is better and then cherry-pick stats.  If I'm wrong, I'm wrong.  In our discussion, you preferred not to value RE24 based on a few examples, but instead said that those who SAW the Twins play didn't think Mauer was better offensively. I just used my eyes and old time stats and said I saw something different. I showed how different people's eye tests might show different things for a variety of reasons.  I posted earlier some of those reasons. Perception based on biases, for one.

 

I like Dozier better than Mauer.  I have no reason to look for reasons to say Mauer is better.

[jay]

Curious to hear some thoughts on this...

 

Would it help if defensive measures were designed to present the data in a way that is more similar to more traditional stats?

 

For example, instead of trying to put a potentially convoluted, complicated run value on a player's defense, what if the stat presented it more like a slugging percentage and we called it defensive efficiency percentage?

[TheLeviathan]

Those 130 runs are the difference between the Fangraphs WAR model matching reality and missing badly. If you don't think that the 130 number is correct, and that it should be something significantly smaller, then you should have an argument as to how to adjust hitting and pitching to account for those missing runs.

Correct me if I'm wrong, but if we did a similar analysis of WAR to team wins wouldn't we see a pretty low success rate of correlation? If so, why would I think this argument is strong?

[jay]

Correct me if I'm wrong, but if we did a similar analysis of WAR to team wins wouldn't we see a pretty low success rate of correlation? If so, why would I think this argument is strong?

Here's a macro look at WAR correlated to wins - http://stats.seandolinar.com/predicting-baseball-wins-with-war/

 

The correlation factor (r^2) of .77 is quite strong. That correlation is exceeded by a team's pythag record, but that can be expected to at least some extent given that pythag is capturing the sequencing that leads to runs while WAR intentionally remains agnostic. I've seen other studies call the correlation closer to .8 and we can find higher or lower in a given season.

 

In any fashion, a .77 correlation to real world wins is significantly stronger than any standard stats. For example, batting average correlates to wins at .27.

 

As a side note, the correlation between WAR and real wins has been found to be stronger with the current state defensive components included. This is a big plus in the rationale for why the "creators" include it.

[jimmer]

 

Here's a macro look at WAR correlated to wins - http://stats.seandolinar.com/predicting-baseball-wins-with-war/

The correlation factor (r^2) of .77 is quite strong. That correlation is exceeded by a team's pythag record, but that can be expected to at least some extent given that pythag is capturing the sequencing that leads to runs while WAR intentionally remains agnostic. I've seen other studies call the correlation closer to .8 and we can find higher or lower in a given season.

In any fashion, a .77 correlation to real world wins is significantly stronger than any standard stats. For example, batting average correlates to wins at .27.

As a side note, the correlation between WAR and real wins has been found to be stronger with the current state defensive components included. This is a big plus in the rationale for why the "creators" include it.

Instead of pythag, try checking the correlation to BaseRuns?

[jay]

To more directly answer the point...

Yes, reducing the 130 runs to some smaller number due to the correlation rate of less than 1 could be one potential explanation.

[jay]

Instead of pythag, try checking the correlation to BaseRuns?

You know it? I'm mobile and that's way too hard right now.

[jimmer]

 

You know it? I'm mobile and that's way too hard right now.

Are you asking if I know where to find BaseRuns or are you asking if I know the correlation because to the first question, Yes.  To the 2nd, no :-)

[jay]

Are you asking if I know where to find BaseRuns or are you asking if I know the correlation because to the first question, Yes. To the 2nd, no :-)

Lol, I wish the second was yes.

 

I think I get what you're saying and now I'm curious. If we replace the observed run differential in the pythag theorem with the BaseRuns (expected runs) differential, how does it change the correlation?

 

I'd think it would be lower than the standard pythag for the same reason WAR is lower. Standard pythag captures some level of sequencing and luck that will get you closer to that team's observed win total.

[jimmer]

 

Lol, I wish the second was yes.

I think I get what you're saying and now I'm curious. If we replace the observed run differential in the pythag theorem with the BaseRuns (expected runs) differential, how does it change the correlation?

I'd think it would be lower than the standard pythag for the same reason WAR is lower. Standard pythag captures some level of sequencing and luck that will get you closer to that team's observed win total.

I ask because BaseRuns seems to be closer to actual record than Pythag in the majority of instances.

[TheLeviathan]

 

In any fashion, a .77 correlation to real world wins is significantly stronger than any standard stats. For example, batting average correlates to wins at .27.

 

Yeah, I wasn't saying it was more or less of a correlation than any other stat, but it seemed as though that particular analysis worked out conveniently to illustrate the point being made moreso than if you had picked out a different set of teams.  Largely because those two teams fell almost perfectly in line with the 68 wins + WAR formula, which I understand is not generally going to be the case.

 

Even if one were to acknowledge that WAR correlates better than most any stat (with defense) it still seems like that argument was a bit ham-handed in making a point that we've hashed out in every imaginable way thus far.  So wasn't trying to relight that fire, just struck me as a bit of a convenient comparison.

[jay]

Even if one were to acknowledge that WAR correlates better than most any stat (with defense) it still seems like that argument was a bit ham-handed in making a point that we've hashed out in every imaginable way thus far. So wasn't trying to relight that fire, just struck me as a bit of a convenient comparison.

Yep... but Levi.....

 

It's still 130 runs!!!!

[TheLeviathan]

http://memecrunch.com/meme/1XHZT/oh-no-not-again/image.jpg

[jay]

http://memecrunch.com/meme/1XHZT/oh-no-not-again/image.jpg

Strangely, my occasional doppelgänger in years long gone by. But, more Speed or Matrix than Bill and Ted.

 

In a twist of seriousness, thanks for the kind words earlier and a thread where we could dig into the merits a bit more than the usual pre-determined bunker lines.

[USAFChief]

Here's a macro look at WAR correlated to wins - http://stats.seandolinar.com/predicting-baseball-wins-with-war/

The correlation factor (r^2) of .77 is quite strong. That correlation is exceeded by a team's pythag record, but that can be expected to at least some extent given that pythag is capturing the sequencing that leads to runs while WAR intentionally remains agnostic. I've seen other studies call the correlation closer to .8 and we can find higher or lower in a given season.

In any fashion, a .77 correlation to real world wins is significantly stronger than any standard stats. For example, batting average correlates to wins at .27.

As a side note, the correlation between WAR and real wins has been found to be stronger with the current state defensive components included. This is a big plus in the rationale for why the "creators" include it.

What's the correlation between pitcher W's and team wins?

 

I kid, I kid.

[KGB]

And I prefer to look for the truth, whatever that may be.  I don't decide who is better and then cherry-pick stats.  If I'm wrong, I'm wrong.  In our discussion, you preferred not to value RE24 based on a few examples, but instead said that those who SAW the Twins play didn't think Mauer was better offensively. I just used my eyes and old time stats and said I saw something different. I showed how different people's eye tests might show different things for a variety of reasons.  I posted earlier some of those reasons. Perception based on biases, for one.

 

I like Dozier better than Mauer.  I have no reason to look for reasons to say Mauer is better.

 

The truth is what we are all looking for - here's also what I said -

 

"One thing is certain about numbers, you can make a case for just about anything. They can be helpful tools but you still really need to evalute players play during a game."

 

I don't dislike Mauer, but I don't think he was better than Dozier offensively last year. I just think stats can be as misleading as your eyes if used incorrectly.

[Otto von Ballpark]

I'd also like to echo jay's appreciation of this thread.  I've learned new things from all sides, even if I've stubbornly butted heads with a few folks too.

[Willihammer]

concur. This thread is an all-timer IMO. Great reading.

[Brock Beauchamp]

Yeah, it has been a great thread all-around. Lots of debate, lots of opinions, lots of angles from which to examine the situation.

[USAFChief]

concur

[Squirrel]

 

concur

Yes

[Brock Beauchamp]

no

[Mike Sixel]

 

And as far as the RBI stat, let's look at this very common example.  Guy gets on base (by walk, HBP, single). Then next guy gets a single and this allows the guy on 1B to go to 3B.  The next guy flies out to the OF and the guy from 3B scores.

 

Whoever made this RBI stat has decided that the only guy who made an out deserves the credit for the RBI.  He's a run producer.  The guy who scored gets the much less hyped stat, run scored.  The guy who put the guy on 3B with his non-out-making-single gets ZERO credit for any of it in the traditional box score.

 

This is the example I always use. The guy who hit a routine fly ball gets an RBI, but they guys who did the real work either get a lesser stat (Tim Raines anyone.....), or no stat. 

 

A guy could hit 500 routine fly ball outs, but if there is a guy on third with less than 2 outs 100 times, he's a run producer somehow.*

 

*yup, that would never happen, but it is the logic used in the stat......

[Hrbowski]

The infield will handle itself, but the outfield and catchers are questionable.

[ashbury]

 

This is the example I always use. The guy who hit a routine fly ball gets an RBI, but they guys who did the real work either get a lesser stat (Tim Raines anyone.....), or no stat. 

 

A guy could hit 500 routine fly ball outs, but if there is a guy on third with less than 2 outs 100 times, he's a run producer somehow.*

 

*yup, that would never happen, but it is the logic used in the stat......

Alternatively, the example shows that the skill of hitting a medium depth flyball has a certain value at times, moreso say than a popup to the catcher, and should be rewarded proportionately.

[Twodogs]

I wonder what Boston thinks of their defense?

[Willihammer]

I pulled the inside edge fielding data from fangraphs for all players (no min) in 2014, and compared the Twins totals to the averages. Excel sheet here

 

If my math is right, the Twins defense missed 40 plays that an average team would have made.