Debating WAR

[Willihammer]

Joe Posnanski just tweeted this:

 

 

If Gordon wins because of a defensive bonus from playing LF, that has to be the death of WAR as a meaningful comparative statistic.

 

He is not good enough to stick at 3B but he can do well in LF and earn a bunch of dWAR, what a bunch of nonsense. A slightly above mediocre CF or 3B is so much more valuable than an elite LF, because it is hard to find even an adequate (or replacement level) defensive player at those positions. The replacement level players for LF are the 70% of MLBers who don't play there because they are good enough defenders to play elsewhere, not the lugs (ahem, Willingham) who they put out there because they can't play anywhere else.

I don't see why that would be the death of WAR. Maybe the positional adjustment is wrong or out of date but that could be easily changed. As it is, Gordon already has a full Win to make up on CFers or 3Bs

[drjim]

I don't see why that would be the death of WAR. Maybe the positional adjustment is wrong or out of date but that could be easily changed. As it is, Gordon already has a full Win to make up on CFers or 3Bs

 

I looked at this closer. Gordon gets about half of his value (2.5 WARish) from his defense. According to the WAR adjustment, this would mean is the equal defender to a 3.5 WAR defender at 3B right? But there is three years of data showing Gordon was about a 0 WAR defender at 3B (08-10). So he can move down the defensive spectrum and pick up 2.5 WAR in value, that is a nice gig.

 

I was happy to see that the guys on Effectively Wild and Keith Law in his current MVP ballot didn't take this WAR value too seriously.

 

But to answer your statement, the positional adjustment misses the point in my mind. The replacement level threshold is the issue, and trying to quantify defense into the same numbers as offense leads to even more issues. I thought there was value in WAR in comparing across positions, but looking more closely since this thread started I'm now more skeptical if it even has value in that.

[Mike Sixel]

Read this, and comment, please:

 

http://www.fangraphs.com/blogs/so-lets-talk-about-alex-gordon/

[Mike Sixel]

Why can't a guy be a terrible 3B, but an excellent OFer? I don't get that argument at all.

 

So, player X is a great OF, but you are sure he'd be a better SS or C? I'm confused by this argument.

[jay]

I thought there was value in WAR in comparing across positions, but looking more closely since this thread started I'm now more skeptical if it even has value in that.

I wouldn't recommend making that determination based off an outlier.

[drjim]

Why can't a guy be a terrible 3B, but an excellent OFer? I don't get that argument at all.

 

So, player X is a great OF, but you are sure he'd be a better SS or C? I'm confused by this argument.

 

I never made that argument.

 

My argument is that many of the people currently playing 3B or SS or 2B or CF would be a better LF than Gordon. He benefits because he is compared to the worst defensive players in baseball as his baseline "replacement" level player, while players who stick at other positions are compared to better defensive players as their baseline "replacement" level player.

 

The notion of a replacement level player in LF should be expanded to include the many players in baseball who don't have the luxury of being moved to LF and have to stick at a position that is much tougher on the defensive spectrum.

[Mike Sixel]

But they did not play LF......how would you know how good they are in the OF? Why complicate this wi th hypotheticals? People don't like defensive measures, and now we should add hypothetical outcomes?

[Willihammer]

Read this, and comment, please:

 

http://www.fangraphs.com/blogs/so-lets-talk-about-alex-gordon/

I also invite Gordon skeptics to read this: http://www.fangraphs.com/blogs/the-value-of-alex-gordon-not-using-his-arm/

[Willihammer]

I never made that argument.

 

My argument is that many of the people currently playing 3B or SS or 2B or CF would be a better LF than Gordon. He benefits because he is compared to the worst defensive players in baseball as his baseline "replacement" level player, while players who stick at other positions are compared to better defensive players as their baseline "replacement" level player.

 

The notion of a replacement level player in LF should be expanded to include the many players in baseball who don't have the luxury of being moved to LF and have to stick at a position that is much tougher on the defensive spectrum.

If you are going to do that, should you also expand the notion of replacement level 3Bs and CFs to include the players who are stashed in LF because their bats are too good to keep out of the lineup? The defensive threshold may be lower in LF but players out there are expected to hit better. Overall, a -10 run penalty seems to be in the ballpark.

[drjim]

Read this, and comment, please:

 

http://www.fangraphs.com/blogs/so-lets-talk-about-alex-gordon/

 

This article didn't say anything I didn't already realize about WAR, how it's measured, and how it might apply to Gordon specifically.

 

I should say, to be clear, that I think Gordon to be a very good baseball player and obviously an elite LF (relative to the other LFs), I don't argue any of that. I just believe there are many players who currently play in MLB that could make a similar switch and do similarly if not just as well, but are playing much more important positions.

 

This article did nothing to improve on the inherent problems of UZR (it admits them), and it really doesn't say much about the problems of applying defensive value when comparing a wide spectrum of baseline "replacement" players or the problems of turning defensive numbers into the same value as offensive numbers.

[drjim]

But they did not play LF......how would you know how good they are in the OF? Why complicate this wi th hypotheticals? People don't like defensive measures, and now we should add hypothetical outcomes?

 

Fair point, WAR complicates enough and already uses enough hypotheticals.

 

And the key point, as far as value is concerned, is that these potential LFs wouldn't necessarily have to be quite as good as Gordon, but good enough to raise the hypothetical "replacement" LF defense above the level of slug that it currently is at. This would lower his defensive WAR value substantially.

[drjim]

If you are going to do that, should you also expand the notion of replacement level 3Bs and CFs to include the players who are stashed in LF because their bats are too good to keep out of the lineup? The defensive threshold may be lower in LF but players out there are expected to hit better. Overall, a -10 run penalty seems to be in the ballpark.

 

This is a fair point, and would limit somewhat the amount of players that could shift to LF, but probably not so much that it would hard to find 30, 50, even 100 (or more) that would be in the ballpark range of providing enough offense to be an acceptable LF bat. Enough that, even if they aren't quite as good as Gordon defensively, enough would be that it would significantly raise the floor of the "replacement" level defender.

[Mike Sixel]

drjim, your argument has merit, but then you really need to expand things. What about defenders that are awesome, but can't hit at all? And so never play. Should they be in the calculation somehow?

 

I think people are over thinking this. This is not a measure used to keep airplanes in the air. It is a measure that attempts to quantify the relative value a player contributed (not will contribute, but did contribute) to their team. I guess if you just take it as an indicator of value, I don't understand how it isn't that. Are the stats less than perfect? yes. Are they so imperfect that they actually mislead us in terms of direction? I don't think so.

[old nurse]

no one has ever rationally argued that WAR is precise to the decimal point. The reality is that WAR has always been best used for grouping players of similar levels of contribution, not for arguing that a 0.1 WAR difference means that Player X is having a better year than Player Y. No one actually argues for using WAR as a precise tool to measure minuscule differences

 

That is from the article. If decimal places are not valid, why did they include them? Fangraphs own argument on WAR is that it is an inprecise tool only usefull to give you a grouping of players. Click on the leaderboard. There as the batting leader is Alex Gordon Listed there as the number one player. Not even listed by offensive WAR. They go by total WAR. It becomes hard to accept the author's defense that WAR is not precise yeat Fangraphs themselves is using it a a precise tool.

[TheLeviathan]

Read this, and comment, please:

 

http://www.fangraphs.com/blogs/so-lets-talk-about-alex-gordon/

 

That entire article was basically arguing "Yeah, WAR is wrong here" and then tried to figure out how wrong it was by talking about what a very small sample size is really weighting the UZR component.

[Brock Beauchamp]

no one has ever rationally argued that WAR is precise to the decimal point. The reality is that WAR has always been best used for grouping players of similar levels of contribution, not for arguing that a 0.1 WAR difference means that Player X is having a better year than Player Y. No one actually argues for using WAR as a precise tool to measure minuscule differences

 

That is from the article. If decimal places are not valid, why did they include them? Fangraphs own argument on WAR is that it is an inprecise tool only usefull to give you a grouping of players. Click on the leaderboard. There as the batting leader is Alex Gordon Listed there as the number one player. Not even listed by offensive WAR. They go by total WAR. It becomes hard to accept the author's defense that WAR is not precise yeat Fangraphs themselves is using it a a precise tool.

They're not using it as a precise tool. They're outputting data and sorting that data numerically. They could include no decimal points or they could calculate WAR to the twentieth point... That has no bearing on how it should be used.

[notoriousgod71]

It's not even just Alex Gordon. When Ben Zobrist and Jason Heyward are in the top ten it's tough to take it seriously.

[old nurse]

They're not using it as a precise tool. They're outputting data and sorting that data numerically. They could include no decimal points or they could calculate WAR to the twentieth point... That has no bearing on how it should be used.

But they are  using it as a precise tool to rank the batters in their listings. They are misusing their own tool.

[snepp]

It has to sort somehow, and they're not forcing anybody to use it down to the decimal. What should they do with them, round everyone up or down by a half a win then sort them alphabetically?

[old nurse]

I do like this from b-ref's om bwar

We present the WAR values with decimal places because this relates the WAR value back to the runs contributed (as one win is about ten runs), but you should not take any full season difference between two players of less than one to two wins to be definitive (especially when the defensive metrics are included).

[glunn]

We could re-enact some of those old BYTO discussions on UZR/WAR, but we probably wouldn't make it past page 1 before someone would have to close it down.

Seriously, you must have been reading my mind.  I love you guys but cannot understand the passion about statistics.  Be glad that you have an exponentially larger choice of statistical measures than existed in the old days and keep in mind that there will always be lots of room for (respectful) debate.

[notoriousgod71]

It has to sort somehow, and they're not forcing anybody to use it down to the decimal. What should they do with them, round everyone up or down by a half a win then sort them alphabetically?

 

 

I'm going to just start rounding all statistics. Therefore a 4.49 ERA is the same thing as 3.51.

[Craig Arko]

I'm going to just start rounding all statistics. Therefore a 4.49 ERA is the same thing as 3.51.

And pi is equal to three.

[jay]

I'm going to just start rounding all statistics. Therefore a 4.49 ERA is the same thing as 3.51.

If you still believe in ERA, then a debate about WAR is going to accomplish nothing.

[Mike Sixel]

People do understand that most statistics come with error bars and ranges, right? Most statistics used in the world to do compkex analysis are estimates.

[Brock Beauchamp]

But they are  using it as a precise tool to rank the batters in their listings. They are misusing their own tool.

 

That's a simple sorting mechanism. Don't read too much into it. That's how databases work.

[drjim]

Here is Law in his chat today:

 

I agree with Jeff Passan's take on Twitter the other day: We do not and should not have the same level of confidence in the precision of defensive metrics than we do in that of offensive ones. If the margin of error on fWAR for a player whose value is largely defensive is really +/- 2 wins, you might as well flush the stat down the toilet. (I don't know that that MOE is accurate, though.)

 

 

I would be curious if they can figure out what the MOE is and what an acceptable MOE would be. If the best solution is to not take the exact number too seriously...

 

Probably my bias coming through but I really think WAR is taking a beating this year.

[TheDean]

People do understand that most statistics come with error bars and ranges, right? Most statistics used in the world to do complex analysis are estimates.

 

Couldn't agree more. Very few things deal in binary yes/no.  Baseball fans have it easy with out/no out, hit/no hit, on base/not on base.  If people don't want to step outside the comfort zone of these binary statistics, no one is forcing them.

 

That said, nearly every science and statistical analysis cannot work on yes/no.  Everything is gray area and uncertainty (thousands of non-baseball examples out there, all of which are probably off-topic).  As a baseball example, I really liked the homerun vs. out comparison in the Fangraphs article you linked.  The result is binary (homerun/no homerun), but, if we want to apply this binary result to something more subjective (Davis's skill at hitting the ball), then we run into something that necessarily requires error bars.

[Mike Sixel]

Imagine trying to make decisions on how to change a car based on statistics, if we were so unwilling to accept standard deviation and margin of error and all the many other things people are saying here about statistics in general (since the argument is that if you aren't sure a stat is "true", it isn't usable).

[Craig Arko]

Imagine trying to make decisions on how to change a car based on statistics, if we were so unwilling to accept standard deviation and margin of error and all the many other things people are saying here about statistics in general (since the argument is that if you aren't sure a stat is "true", it isn't usable).

Let us not forget the humble confidence interval, either.