Posnanski piece on WAR and the Cy Young

[gunnarthor]

http://joeposnanski.com/porcello-v-verlander/

 

A few days late but a nice little read.  Pos points out that a lot (if not all) Verlander's WAR lead was because of how WAR calculated defense.  And Pos doesn't think those numbers accurately reflect what happened.

[mikelink45]

Great evaluation of the statistics.  This breakdown is really significant and really points out the problems we have as we take a single formula to be the most important evaluation tool.  A defense does change according to pitchers.  Weak contact makes it easier to field.  Fewer hard hit balls make it easier to field.  The defense of any team varies depending upon the pitcher and in this case the variation is out of the bounds of acceptable statistics. 

[nicksaviking]

Good read, and kind of damning. No doubt a defense can make a pitcher look better than he is but I agree with the above post, a pitcher can make his defense look better than it is as well.

 

I'm not a great defensive metric mind, but just one specific data point of reference we are familiar with: Torii Hunter in his last year in Detroit had all kinds of terrible defensive rankings base almost entirely on his -18.4 Range Runs component of UZR. Everyone was concerned when the Twins signed him yet the next year he was .3 in the Range Runs component and had a positive UZR. It's not like he got faster, more likely it was due to positioning due to the manager's decision or it is due to the ballpark.

[TheLeviathan]

So, so perfectly said:

 

And I wonder how many of them realized they were voting for a defensive adjustment. I love the concept of WAR, and I appreciate the efforts to make it better all the time. And I know the Baseball Reference people do not claim that it is the perfect statistic or that anyone should base their entire award ballot on it. But WAR does have real sway in the baseball commuinity. And in this case, I think it was pretty misleading.

 

WAR does have real sway and so frequently we don't look at the underlying factors and just quote the number (in increasingly smaller samples it seems) to make points without really knowing if the statistic is meaningful.  I love this article, thank you for posting it!

[spinowner]

I've always had a bit of trouble accepting WAR. I understand the concept and significance of trying to rate a player in terms of how he impacts the outcome of games but the problem is that it is not an objective statistic. Formulas vary based on the opinions of those who determine them. IMHO baseball is too complicated to definitively measure the value of the contribution of any individual player.

[Mr. Brooks]

I've always had a bit of trouble accepting WAR. I understand the concept and significance of trying to rate a player in terms of how he impacts the outcome of games but the problem is that it is not an objective statistic. Formulas vary based on the opinions of those who determine them. IMHO baseball is too complicated to definitively measure the value of the contribution of any individual player.

If you can't measure the value of any individual player, then how do teams know which players to keep, and how much to pay them?

 

I'm fine thinking WAR can be improved, but I don't understand not even trying.

[spinowner]

As I said in my first post I understand the concept and significance of WAR. The WARs that are out there can give us a general idea of who the best players are but we can usually get that from objective statistics. I just think that it's not possible to determine accurately or precisely how much of an effect individual players have on the outcome of games.

[gunnarthor]

Posnanski had a follow up piece on this as well.  http://joeposnanski.com/a-bit-more-on-war/

 

 

[TheLeviathan]

 

Posnanski had a follow up piece on this as well.  http://joeposnanski.com/a-bit-more-on-war/

 

This is a nice nutshell of my issues with WAR.  Good stat.  Nice when used correctly.  Easy, quick cheat to estimate value in many cases.

 

But the second you start using that stat to make compelling arguments across positions and in small samples you lose me.  And I feel like the response Joe got is so typical: "WAR isn't perfect, but I'm going to ignore all your concerns, keep using the stat anyway, and make sure I say it isn't perfect again just to make sure you know i know that.  But now i'm going to use it like it's definitive again..."  And around and around we go.

 

It totally misses the point that, in many specifics instances, it's absolutely inappropriate to use and still very flawed.  And while so many proponents acknowledge it's flaws, they still use it like they don't.  

 

 

[Brock Beauchamp]

I like where Posnanski is going with this but the argument that because run support can fluctuate for pitchers, so can defensive runs saved/lost is asinine and he knows it.

 

The problem with using team defensive runs is that it doesn't go granular enough... As Posnanski mentions, is the pitcher a flyball guy? A groundball guy? How does his defensive alignment compare? If you use overall team numbers, you can skew things out of proportion.

 

Let's say the Twins had this alignment:

 

Buxton, Rosario, Hicks in the outfield.

 

Vargas, Polanco, Escobar, Sano in the infield.

 

Put Phil Hughes out there. He loves that defense.

 

Put Kyle Gibson out there. He hates that defense.

 

Yet that defensive alignment, with some proper luck, might be above average overall so both Hughes and Gibson are dinged the equal amount because overall defense=good.

 

Obviously, one of those pitchers should be dinged for that alignment while the other should be given a medal for ever making it out of the fifth inning.

 

But Posnanski's comparison of runs scored - a trait that can fluctuate wildly based on a single game - is still really ****ing stupid. I expect better of him. It's lazy.

[TheLeviathan]

I think you may have gotten lost in the weeds Brock. He just picked an argument he knew he'd have full agreement on to demonstrate that sometimes more careful analysis is necessary to really differentiate performance than simply looking at one stat. It is absolutely possible for a pitcher to have exceptionally dumb luck around him, be it defense or offense. At least that phenomenon is real enough that in any one year sample it bears investigation.

[gunnarthor]

What really surprised me was the acknowledgement that a 1-2 win difference isn't definitive.  I thought that was much smaller - like .5 win.  If it's 1-2, it's almost useless. 

[Brock Beauchamp]

 

I think you may have gotten lost in the weeds Brock. He just picked an argument he knew he'd have full agreement on to demonstrate that sometimes more careful analysis is necessary to really differentiate performance than simply looking at one stat. It is absolutely possible for a pitcher to have exceptionally dumb luck around him, be it defense or offense. At least that phenomenon is real enough that in any one year sample it bears investigation.

It poisons the well for those of us who know better, though.

 

First, defense is far more stable than offense. Dozier might hit three bombs in a night and drive in seven runs and then not hit another homer for three weeks. But Byron Buxton, while he may flub the occasional play, is not going to be 10mph slower while running down a flyball on any given night. And defense can't be influenced by the opposing pitcher, who can and will routinely shut down an offense.

 

Second, there's the huge gap in data points. A starter, if healthy, only starts 32-ish games a season so we're already suffering from sufficient data points. Second, runs are much more rare than defensive putouts (something to the magnitude of 6-7x more rare per game than defensive putouts).

 

He could have just as easily made the argument without making such an obviously flawed comparison. Defense doesn't slump but offense does on a very regular basis.

 

If Posnanski can find me the defensive equivalent of a pitcher being on the receiving end of his offense posting a ten spot in five innings, I'd love to see it (and a single ten run outing is worth .33 run support over a full, healthy pitching season).

[TheLeviathan]

 

First, defense is far more stable than offense. 

 

Are we sure about that?  And, even more specifically, are we sure our measurements of it are?

[Brock Beauchamp]

Are we sure about that? And, even more specifically, are we sure our measurements of it are?

I don't see how defense could fluctuate as wildly as offense, especially when broken into single games.

 

But the measurements, those are absolutely in question.

[gunnarthor]

 

I don't see how defense could fluctuate as wildly as offense, especially when broken into single games.
 

I dunno.  I could see an IF behind a slow moving pitcher with control problems being a step slow on a play or two.  Anyone remember how Buerhle's defenses always seemed ready to go compared to, say, Pelfrey?  Bad weather or carrying an at-bat into the field could muck something up.  I don't know if we have enough evidence to suggest that a player's defense is the same, game to game.  Esp since the player might get 4 chances one night, 8 the next and 3 the day after.  

[Willihammer]

What really surprised me was the acknowledgement that a 1-2 win difference isn't definitive.  I thought that was much smaller - like .5 win.  If it's 1-2, it's almost useless.

1-2 runs he says.

[gunnarthor]

 

1-2 runs he says.

That would make more sense but Sean said wins.  It's in the article Pos links to.

 

http://joeposnanski.com/baseball-reference-war-a-response/

 

"We state clearly that we don’t find differences of 1-2 wins to be definitive."

[Willihammer]

A player's defense obviously varies game to game. The question is, how much variation can an entire team have over a ~200 inning sample?

[jimmer]

Posnaski's issue comes up when people use bWAR for pitchers. Its why fWAR uses FIP in its equation.

[TheLeviathan]

 

I don't see how defense could fluctuate as wildly as offense, especially when broken into single games.

But the measurements, those are absolutely in question.

 

I don't know, couldn't a pitcher just have crap luck over a season?  When you think about how much it could influence one pitcher, it wouldn't take as much as you think.  In 30 starts, even a handful of games in which there is bad luck with his D could make a sizable difference.  Especially if that bad luck isn't counted as errors. 

 

Or maybe he just doesn't get as lucky with heroic defense efforts either.  With defense, it only takes a play or two to cause a domino effect in a game.  Less so with offense.

[Mike Sixel]

 

I don't know, couldn't a pitcher just have crap luck over a season?  When you think about how much it could influence one pitcher, it wouldn't take as much as you think.  In 30 starts, even a handful of games in which there is bad luck with his D could make a sizable difference.  Especially if that bad luck isn't counted as errors. 

 

Or maybe he just doesn't get as lucky with heroic defense efforts either.  With defense, it only takes a play or two to cause a domino effect in a game.  Less so with offense.

 

And how would other stats catch this? 

 

The constant straw man attacks on WAR and those who use it are tiring. No one says it is perfect. No one says it is the only stat. No one says don't look at other things. No one says anything that is claimed over and over.....

 

As for the "objectivity" claim above, we all know that errors are not objective, heck, balls and strikes aren't all that objective, but let's stay just on errors for now. Do people really think ERA is helpful, and is the perfect stat (hahaha) given that it is somehow objective, given how errors are given out (or not given out)?

[gunnarthor]

 

And how would other stats catch this? 

 

The constant straw man attacks on WAR and those who use it are tiring. No one says it is perfect. No one says it is the only stat. No one says don't look at other things. No one says anything that is claimed over and over.....

 

 

No one says those things but nearly everyone believes those things.  I don't like WAR and I use it a lot to say Player A is better than Player B except in strange weird circumstances.  Hell, a whole generation of writers/bloggers have rallied around it to keep Jack Morris out of the Hall of Fame.  Writers constantly use it say why one person should have won MVP.  There is very rarely any more depth than that.  Esp in our Hot Take culture.

[Mike Sixel]

Chat today on Fangraphs:

 

12:32
Sirras: Sunday’s links included a link to Bill James’s discussion of the AL Cy Young. He feels that those voting for Verlander over Porcello solely because of his higher WAR are misapplying defensive statistics. Did you read the piece and do you have any thoughts to add?

 

12:34
Dave Cameron: The issue that people are talking about with B-R’s WAR is the exact reason why we chose a FIP-based WAR. I wrote about it a few years back (http://www.fangraphs.com/blogs/why-our-pitcher-war-uses-fip/), and I think it’s good that people are starting to realize the downside of doing things the way B-R does them. There are downsides to both approaches, of course, but the downsides of the FIP approach have been discussed a ton, while the downsides of doing it B-R’s way have been overlooked, I think.

[Mike Sixel]

 

No one says those things but nearly everyone believes those things.  I don't like WAR and I use it a lot to say Player A is better than Player B except in strange weird circumstances.  Hell, a whole generation of writers/bloggers have rallied around it to keep Jack Morris out of the Hall of Fame.  Writers constantly use it say why one person should have won MVP.  There is very rarely any more depth than that.  Esp in our Hot Take culture.

 

There were waaaaaaaaaaaaaaaaaaaaaaaaay more things than WAR being argued about Morris. Way more. So much more I can't type way enough.

[Teflon]

I'm sure somebody will discredit this, but if you line up the individual pitcher's Bill James game scores in 2016 from highest to lowest, Verlander is consistlenly higher per game.

 

 

[gunnarthor]

 

I'm sure somebody will discredit this, but if you line up the individual pitcher's Bill James game scores in 2016 from highest to lowest, Verlander is consistlenly higher per game.

 

I like that - game score has some issues (and favors strike out pitchers) but I like stats that look a bit more at individual games rather than a compiled season of stats.  

[TheLeviathan]

 

And how would other stats catch this? 

 

The constant straw man attacks on WAR and those who use it are tiring. No one says it is perfect. No one says it is the only stat. No one says don't look at other things. No one says anything that is claimed over and over.....

 

As for the "objectivity" claim above, we all know that errors are not objective, heck, balls and strikes aren't all that objective, but let's stay just on errors for now. Do people really think ERA is helpful, and is the perfect stat (hahaha) given that it is somehow objective, given how errors are given out (or not given out)?

 

It's not a straw man if it's true.  I even addressed your very argument above.  Yes, everyone throws their caveats around about what WAR is or isn't.....and then utterly disregard them in practice.  

 

It's like saying "Hey, it's a bad idea to get drunk any day but Saturday.  I realize it's not something we should do.  Now,about getting absolutely hammered on Thursday..."

 

At some point I stop taking the caveats seriously when it becomes abundantly clear you don't take them seriously either.  This Verlander vs. Porcello thing arose because people were quoting WAR without any context or inspection.  Just slapping the ol' WAR number down and saying "case closed!"

 

And that, sadly, is the root of the problem.  There is basically no effort whatsoever to hold WAR to those qualifications about it's weaknesses.  It's just used, constantly, even when not appropriate.  And in this particular debate, it appears to have been inappropriate.  Yet it took a voice in the wilderness to yell out "Um...hey...not so fast!"

 

If all the WARheads out there are as concerned and beholden to the flaws of WAR, you'd think it wouldn't have had to get that far....no?

[Mike Sixel]

 

It's not a straw man if it's true.  I even addressed your very argument above.  Yes, everyone throws their caveats around about what WAR is or isn't.....and then utterly disregard them in practice.  

 

It's like saying "Hey, it's a bad idea to get drunk any day but Saturday.  I realize it's not something we should do.  Now,about getting absolutely hammered on Thursday..."

 

At some point I stop taking the caveats seriously when it becomes abundantly clear you don't take them seriously either.  This Verlander vs. Porcello thing arose because people were quoting WAR without any context or inspection.  Just slapping the ol' WAR number down and saying "case closed!"

 

And that, sadly, is the root of the problem.  There is basically no effort whatsoever to hold WAR to those qualifications about it's weaknesses.  It's just used, constantly, even when not appropriate.  And in this particular debate, it appears to have been inappropriate.  Yet it took a voice in the wilderness to yell out "Um...hey...not so fast!"

 

If all the WARheads out there are as concerned and beholden to the flaws of WAR, you'd think it wouldn't have had to get that far....no?

 

I assume WARheads is some kind of term of endearment?

 

so, every time I use WAR, I need to type how many caveats and other things? Just curious....so if I use it, I use it correctly.....

 

[jimmer]

 

I assume WARheads is some kind of term of endearment?

 

so, every time I use WAR, I need to type how many caveats and other things? Just curious....so if I use it, I use it correctly.....

Probably best just to stop using WAR and go back to using wins, RBI, save, ERA and fielding % to figure out value for players.  You know, the perfect stats that need no caveats :-)