27 replies to this topic

[Brett]

I like to think I'm a stats guy, but, for whatever reason, I've never been able to get into WAR (Wins Above Replacement), which is supposed to be "THE STAT." It is designed to a universal stat that reduces every player to a value (above the mythical "replacement player"), which can be compared to every other player. Pretty handy.

But, as I said, it just hasn't looked ... right to me. WAR frequently tells me that Player X is better than Player Y, when a review of their stats tells me that just can't be right. I have tended to be content to let other people fight the battle when I say things like "meh, WAR doesn't tell the whole story." Mostly because there are only so many hours in the day, and the vast majority of mine are spent obsessing about the Cubs' roster and GM search. I don't like writing about things that I don't understand to the nth degree.

Well, someone took up the cause and laid out a beautiful discussion and take-down of WAR, based on an even more disliked advanced stat: UZR (Ultimate Zone Rating). The latter is supposed to be the best of the advanced fielding stats, but is wildly, wildly prone to inexplicable swings - Marlon Byrd was one of the worst defensive center fielders in 2009 in Arlington, and then suddenly (at age 32), he was one of the best defensive center fielders in 2010 in Chicago. I just don't buy those kind of absurd, inexplicable swings in something that, by age 32, is a relative constant.

Ok, enough throat-clearing. Here's the post from It's About the Money. Warning: it's long and dense. Totally worth it if you're wading into the advanced metric stuff, though.

An exemplary conclusion paragraph:

While I admit the difficulty of building a model that accounts for the effect a pairing like Braun/Fielder or Pujols/Holliday has on the rest of the lineup, this is one area in which I find the conventional wisdom to be irrefutable. While I applaud WAR (and other metrics) for aiding in our appreciation of defense and baserunning, it’s beyond asinine to conclude that Ellsbury is twice as valuable as Fielder. Too often WAR is used as a means of comparing oranges to apples. One of the things that makes baseball great is the diversity of the fruit basket. WAR give incredible weight to scarcity of shortstops, but no weight to the scarcity of pitcher-intimidating, strategy-altering cleanup hitters, which I see as a form of reverse discrimination.

[MichiganGoat]

I'm am not someone who remotely understands these advanced stats (too much math for my taste), and I've tried to use WAR to justify what I'm seeing in other stats and on-field, but as Brett says it never really seems to add up. I'll look forward to reading through the article and understand the errors and merits in these types of stats.

[KyleNovak]

Most excellent post Brett.

I couldn't agree more about UZR and other defensive-related stats.  

Let's face it.  Data accumulation and analysis is a very real and important part of baseball.  Many sabermatricians believe that defense is the "new frontier" in the search to get an edge.
Defense numbers are still primitive and raw compared to what they will become.

I wouldn't be surprised if the following isn't A ) already in use or B ) in the planning stages of many teams:  

A series of HD-cameras that give team analysts the precise location of every ball hit and the distance each fielder had to move to make the play.  Charting the distance, trajectory, and velocity on every single throw that an outfielder makes when trying to throw out an advancing runner or when an infielder makes a throw across the diamond on a force play.  Having on-field stopwatches that time every baserunner's speed from home plate to first, first to second, and so on.  Comparing the conditions of the playing field as well as the layout of the infield grass in each stadium (and the effect it has on grounders).
You can take that data and compare true defensive ability across your own team and across the league.  For example, easily look at where all of the balls in play were hit and by whom.  Where and how were any errors committed?  By this kind of analysis, you can determine what constitutes an "average" fielding performance (range to each side, fielding accuracy, throwing accuracy, "dig or scoop" accuracy, particularly at first base)  A true fielding advantage can be assessed when you clearly see that Player A actually prevents more baserunners than his counterpart Player B simply because he has the ability to make tougher plays as noted by the data, even if he has a few more errors than Player B.  It is this way that plays can be considered "routine" while others are considered "tough".  A successful bare-handed charging play made on a weak infield dribbler hit on the grass by Jacoby Ellsbury is going to be a better indicator of fielding prowess than a slow chopper hit right at the fielder by Jim Thome, even though they get both get counted as a groundout in the box score.  It can also take away certain visual falsehoods, such as the whole "crowd sees me dive for a ball and cheers me yet doesn't seem realize I HAD to do it to make the play because I'm slow/getting older and lost a step/or constantly get a bad jump" bias.* 

It's going to be interesting to see how fielding metrics evolve in the next handful of years.  Whatever the case, I think we can agree that data supported by VISUAL evidence is going to be key.              

In 2011, I can look at the hitting statistics of any player and get a clear, unbiased snapshot of how they perform without ever having to see them play an inning.  Power numbers.  Patience.  Contact rate.  Any piece of data you want is available.  That is why I really like WAR, particularly the offensive variety, as it makes me dig a bit deeper into the reason they ended up with that particular rating.  Player A had a 7.5 WAR last year?  Seriously, how was that the case?  Then I start to look around their advanced stats, maybe noticing a high OPS with RISP, great isolated power numbers despite some bad hitting luck (low BABIP compared to league average), possibly even noticeable success stealing bases, and so forth.  A high offensive WAR means you can that find facets of a player's offensive game that were clearly very valuable.   

That being said, WAR has its issues.  There is no universal way to calculate it and the defensive contribution to the score brings us right back up to the first point of this post.  How do we even calculate good defense?  I agree about the "apples to oranges" argument with the Ellsbury and Fielder debate being a great point.  Baseball-Reference has a "position adjustment" section in its WAR calculation.  Why should a first baseman be penalized for his position and what arbitrary number was agreed upon as said penalty?  Should Ellsbury be considered more valuable just because his numbers are so much higher than other center-fielders?  Fielder is compared to Albert Pujols and Joey Votto (along with a large number of other NL first-basemen who are solely responsible for raising the league average hitting statistics) while Ellsbury (along with Curtis Granderson) pretty much stands way above the much closer-to-average crop of center-fielders in the AL.  Using WAR isn't necessarily going to make it easy comparing players across eras either.     

WAR is just another piece you can use to help justify an argument or support your facts, just like any other statistic.  I wouldn't want to base an opinion solely on WAR alone.


*Example:  Jim Edmonds was a good center-fielder, but was often accused of getting a poor jump on a large number of flyballs, thus resulting in him having to dive to make what would have been a much easier play had he gotten a better jump.  While the dives were impressive to say the least, it often overrated him as a fielder. (Web Gems on SportsCenter are definitely guilty of this). This same criticism has been leveled at Andruw Jones when he was in Atlanta.  While I am choosing neither to confirm/deny the criticism of either player, this method of fielding data collection could very well provide evidence in similar cases. 

 

[Brett]

I think Kyle may have just netted the first Board-to-blog post promotion. I gotta bookmark this badboy.

[TWC]

He'll be in good company up there with the Brady Bunch discussion.

[Brett]

He'll be in good company up there with the Brady Bunch discussion.

Bleacher Nation: Where everything and everyone is welcome, for better or worse.

[King Jeff]

I'm am not someone who remotely understands these advanced stats (too much math for my taste), and I've tried to use WAR to justify what I'm seeing in other stats and on-field, but as Brett says it never really seems to add up. I'll look forward to reading through the article and understand the errors and merits in these types of stats.

I thought you were a math teacher for some reason.

[TWC]

I'm am not someone who remotely understands these advanced stats (too much math for my taste), and I've tried to use WAR to justify what I'm seeing in other stats and on-field, but as Brett says it never really seems to add up. I'll look forward to reading through the article and understand the errors and merits in these types of stats.

I thought you were a math teacher for some reason.

I got an awfully good chuckle out of that one Jeff.

[MichiganGoat]

I'm am not someone who remotely understands these advanced stats (too much math for my taste), and I've tried to use WAR to justify what I'm seeing in other stats and on-field, but as Brett says it never really seems to add up. I'll look forward to reading through the article and understand the errors and merits in these types of stats.

I thought you were a math teacher for some reason.

I got an awfully good chuckle out of that one Jeff.

The idea of me teaching math has me laughing more than any of you

[T C]

ObstructedView (or, mb21 more specifically) basically wrote the same thing the other day, and he basically wrote the same thing about UZR and how it is completely and utterly broken. Which it totally is, and by the transitive property, so is WAR. Theres no way for WAR to be valuable when it includes a useless piece of data like UZR.

But as far as O-WAR, or WAR with the defensive component removed, I love it, and I dont have an issue with the fact that it "underrates" big, slow first basemen and sluggers (which it totally doesnt, by the way. WAR absolutely LOVES Joey Bats and Albert Pujols and the best bats in the league). But the idea that I think youre missing here is the idea of "over replacement". Yea, Fielder isnt loved by WAR, buts thats because there are so many firstbasemen that hit like he can. Off the top of my head, Pujols, Gonzalez, Teixeira, Votto are all at his level, and then theres a myriad of others that are a lot closer than most people think. His "replacement" player has a much higher wOBA than, say, Troy Tulowitzki's or Ben Zobrist's "replacement". The idea is that its great to have a great player at a scarce position, cause it provides a much stronger competitive advantage over your opponent than having a Fielder over Mike Morse.

[NormB]

Why is it troubling that there is so much variation in year to year defense? We see the same year to year fluctuations in pitching and hitting.
And I think it was touched upon, but Jacoby Ellsbury was twice as valuable as Prince Fielder in 2011 because WAR is based on position. Jacoby is about 10 Wins better than a replacement center fielder, Prince is about 5 WAR better than a replacement first baseman. Give me Jacoby and the 15th best first baseman over Prince and the 15th best center fielder any day of the week.

[King Jeff]

Why is it troubling that there is so much variation in year to year defense? We see the same year to year fluctuations in pitching and hitting.

Starlin Castro's UZR dropped from -2.7 to -8.7 from 2010 to 2011. He committed 2 more errors last year in 300 more innings, there is no way his defense got that much worse. Tell me where that much variation came from.

[Brett]

Why is it troubling that there is so much variation in year to year defense? We see the same year to year fluctuations in pitching and hitting.

Starlin Castro's UZR dropped from -2.7 to -8.7 from 2010 to 2011. He committed 2 more errors last year in 300 more innings, there is no way his defense got that much worse. Tell me where that much variation came from.

Yeah, there's fluctuation in hitting and pitching, but nothing like the kind of swing I pointed out in Byrd (which is commonplace in UZR). Further, it's just my opinion from playing and watching baseball that defensive ability is likely to be more consistent from year to year than hitting or pitching.

[King Jeff]

Why is it troubling that there is so much variation in year to year defense? We see the same year to year fluctuations in pitching and hitting.

Starlin Castro's UZR dropped from -2.7 to -8.7 from 2010 to 2011. He committed 2 more errors last year in 300 more innings, there is no way his defense got that much worse. Tell me where that much variation came from.

Yeah, there's fluctuation in hitting and pitching, but nothing like the kind of swing I pointed out in Byrd (which is commonplace in UZR). Further, it's just my opinion from playing and watching baseball that defensive ability is likely to be more consistent from year to year than hitting or pitching.

I agree, a few percentage points fluctuation would give me more confidence in the rankings, but the way that they swing now just doesn't give me year to year confidence in the stat. I would think your fundamentals and positioning would help you improve defense the more you play, and errors usually at least support that theory. UZR seems to somehow change drastically from year to year for certain guys, which leads me to believe that it weighs some stats too much. I like using it to compare players at each position, each year, but it doesn't really accurately indicate the level of a players defense over the course of their career.

[NormB]

Why is it troubling that there is so much variation in year to year defense? We see the same year to year fluctuations in pitching and hitting.

Starlin Castro's UZR dropped from -2.7 to -8.7 from 2010 to 2011. He committed 2 more errors last year in 300 more innings, there is no way his defense got that much worse. Tell me where that much variation came from.

Yeah, there's fluctuation in hitting and pitching, but nothing like the kind of swing I pointed out in Byrd (which is commonplace in UZR). Further, it's just my opinion from playing and watching baseball that defensive ability is likely to be more consistent from year to year than hitting or pitching.

Jacoby Ellsbury Hr's.
Adam Dunn's OPS
James Shields ERA

Those swings aren't as big as the UZR jumps?
I'm no UZR supporter by any means, but I think they jump around just as much as a players AVG, a players HR's, a players RBI's, ERA, Whip, etc....