2009 AL MVP Showdown: Greinke vs. Mauer vs. Mr. X, or, the Race for Second Place, or, Ken Rosenthal, Avert Your Eyes
by devil_fingers on Oct 15, 2009 7:31 AM PDT
in Sabermetrics
38 comments
Much of the debate about the AL MVP this year (well, aside for the silly aspects of it) has focused on position players such as Joe Mauer, Kevin Youkilis, Derek Jeter, and dark horse candidate Yuniesky Betancourt. But along with a few others, I've argued (for a while now) that Zack Greinke probably deserves to win it. In light of my post on catcher defense earlier this week (andin the comments following), some people understandably thought that Joe Mauer's impressive +4.3 showing put him over the top, if he wasn't already.
Well, I hate to go Ken Rosenthal on you, but Mauer isn't the obvious choice. And by that, I mean that (on my WAR calculations) he obviously isn't the right choice for AL MVP. What isn't obvious is whether or not he's even number two...
Not This Again!
Okay, that last sentence was pretty rhetorically charged. The neat thing about Wins Above Replacement (WAR) is that people can do it in different ways -- using different pitching stats, team-adjusted linear weights or not, league adjustments, etc. What I want to do here is simply to "show my work" for the position players to back up my claim that, at least on my numbers (some of which are borrowed from FanGraphs for convenience), Greinke is the clearcut choice, and #2 isn't.
Again, Rally or others may have something different on their calculations, depending on their fielding stats, how Greinke's RA and defense comare, etc. (although my catcher defense results probably aren't that different from what his will be).
I've already shown my work for Greinke in an earlier post and the various posts on pitcher value scattered throughout this site. I want to show that I extend every courtesy to Mauer -- the same adjustments for the AL, the same run-to-win conversion, and including everything.
For Greinke, I concluded that he was worth 9.52 Wins Above Replacement this season. It's important to keep in mind that although he only saved ~85.2 runs above replacement, that pitchers effect their run environment, so they get a customized run-to-win conversion so that they are valued properly.
I'm going to use FanGraphs linear weights and positional adjustments, as I'm fine with those and they are park adjusted. However, I don't think they adjust for AL/NL differences in replacement level yet. I also think they use one run-to-win conversion for all position players. The latter isn't that big of a deal, but since I used the AL run-to-win conversion for Greinke (customized, of course), I should use it for Mauer, too.
Valuing Mauer
Let's take a look at Mauer's Value section at FanGraphs. He was an incredible +56 hitter, and gets a +5.4 positional adjustment for the time he spent at catcher and DH. For playing time ("replacement"), FanGraphs currently does +20 runs per 600 PA (I believe). To get the league adjustment right (as we did with Greinke's win %), we'll change that to +25 runs (should be 2.5 Wins, but runs is actually more generous to Mauer and more straightforward). Mauer has 606 PAs this season, for 21.64 replacement runs (I know two decimal places is silly, but I did it for Greinke...). He also gets +4.3 runs for fielding, as noted yesterday.
But, as I commented in the earlier thread (to some consternation), his baserunning wasn't great this season. People were surprised, but check out Baseball Prospectus's EqBRR for the Twins. We have to take out the stolen base runs since FanGraphs linear weights (aka wOBA) already includes that. So for Mauer, we get -3.7 EqBRR - -0.8 EqSBR = -2.9 BsR.
So, add it all together to get RAR: +56 Hitting + 4.3 Fielding - 2.9 Baserunning +5.4 Position + 21.64 replacement = 84.44 Runs Above Replacement.
That's great, but we need to convert to wins. The formula I use is from Tom Tango: (lgRA*1.5)+3. Yes, there's a PythagenPat version that is more accurate, but this is close enough and is more easy to integrate with the dynamic run-to-win conversion, and FanGraphs uses this, as well. For the AL, the lgRA is 4.75. so the Run-to-Win voncerson is about 10.13.
84.44 divided by 10.13 = 8.34 Wins Above Replacement for Joe Mauer in 2009. That's outstanding. It would be like Albert Pujols 3rd best season (off the top of my head).But it's more than a win below the 9.52 we got for Greinke. Hey, as Buzz Aldrin might say, second comes right next to first...
But wait...
Mr. X
Mr. X. also plays in the American league. Mr. X was 40.2 batting runs above average in 2009. He was a +25.9 fielder. Although he isn't a great basestealer, he was still +0.4 as a baserunner. He played all over the place, and had a -1.1 positional adjustment. He had 599 AL PAs, meaning 21.39 replacement runs.
+40.2 Hitting + 25.9 Fielding +0.4 Baserunning -1.1 Positional +21.39 Replacement = 86.99 Runs Above Replacement, or 8.59 Wins Above Replacement.
So who is this Mr. X? Derek Jeter? Kevin Youklis? Um, no, it's Ben "Zorilla" Zobrist.
Final Thought
I look forward to see what Rally (and hopefully others!) come up with after the season. I'm just giving you my take Sure, here's lots of ambiguity on the fielding stats for these guys, and they're pretty close.
But according to these numbers, the debate about Mauer should be whether or not he or Zobrist should be the runner-up to Greinke.
Of course, in a way, all three players deserve to win the MVP. But in another, more accurate way, Greinke deserves to win it.
0 recs |
38 comments
|
Email
Print
Comments
Good stuff, per usual.
I, too, put Greinke #1 on my MVP ballot.
NL was a bit trickier between Timmy and Pujols. I went with Pujols, but either would have been fine — Timmy dominated the pitchers more than Pujols dominated the hitters, for what that’s worth.
Beyond the Boxscore Not a member? Sign up.
by Sky Kalkman on Oct 15, 2009 7:43 AM PDT reply actions 0 recs
Given the error bars around our estimates of offensive and defensive contribution, I would say that the odds are that Mauer was better than Zobrist. (IOW, it is more likely that Mauer was +56 hitting than it was Zobrist was +26 fielding. This is not talking about true talent, just the uncertainty of our estimate of actual performance.)
As for Grienke – the uncertainty around estimates of replacement level really hit home when you want to compare hitters to pitchers.
by cwyers on Oct 15, 2009 8:52 AM PDT reply actions 0 recs
I've read Tango's explanation on replacement for pitchers, but I always felt like they were a bit overestimated by WAR
Colin, can you point out a discussion which considers these uncertainties?
Marlin Maniac, a Florida Marlins blog
Come attend Intro to Sabermetrics 101!
Check me out at Beyond the Box Score as well.
by SFiercex4 on Oct 15, 2009 9:15 AM PDT up reply actions 0 recs
Not as such.
I’ve mused about the topic before. Let’s just say the consensus about pitcher replacement level is not there to the extent it is for hitters. For instance, a few popluar rep-level metrics and the win percentage for a replacement level team:
- BPro’s WARP – .230
- Fangraphs WAR – .290
- Rally’s WAR – .330
Here’s the funny thing. They all basically agree on what a replacement level position player looks like – average defense, hitting about about 70-80% of league average. So what does that tell you about replacement-level pitching?
by cwyers on Oct 15, 2009 9:45 AM PDT up reply actions 0 recs
Do we need to know what replacement level is? Why can't we use runs above average?
Or would that cause problems because an average pitcher isn’t worth what an average hitter is worth?
by lookatthosetwins on Oct 15, 2009 9:42 PM PDT up reply actions 0 recs
That's a large part of it.
The other issue is playing time. An average baseline makes Mauer look a lot better compared to the other hitters, because it doesn’t properly account for the time he missed.
by cwyers on Oct 17, 2009 11:45 AM PDT up reply actions 0 recs
I agree about Zobrist
although the same would apply to Mauer’s defensive performance — and for both, the uncertainty could go both ways (although I would agree that the further you get from average, the greater the uncertainty is when it comes to defensive performance)
As for uncertaint about replacement level pitching… well, I’m not sure. Expand on that. I will say that I’ve been surprised by who has been above replacement level when I’ve run the a big SQL for this stuff, so maybe Zack is being underestimated…
It’s not like there’s a few runs between Zack and Mauer, it’s more than a win.
I'm not a sabermetrician, but I do play one at Driveline Mechanics.
Can't get enough of me? Check out my Twitter feed.
by devil_fingers on Oct 15, 2009 9:19 AM PDT up reply actions 0 recs
In this case, where you set the replacement level is everything.
Tango uses a .380 Win% rep-level for starting pitching, Rally uses something more like .410. (You can see a discussion here.) I lean more towards Rally’s definition. And we are certainly talking about “more than a win” differences here. So you have to agree on pitcher rep-level before you can converse about Grienke versus Mauer, or you’re just shouting past each other.
by cwyers on Oct 15, 2009 10:00 AM PDT up reply actions 0 recs
Well, I'm using .370 Win% for the AL, as Tango has it
If I put replacement level up to .410 for Greinke, he’s at 8.50, that’s equal (well, slightly higher) than Mauer. If he adjusts for league, what is it for 2009 AL? What is it for AL hitters?
I'm not a sabermetrician, but I do play one at Driveline Mechanics.
Can't get enough of me? Check out my Twitter feed.
by devil_fingers on Oct 15, 2009 10:42 AM PDT up reply actions 0 recs
Blurg
Look, I realize that UZR is the best measure we have currently, but a) it doesn’t mean that it’s definitive and b) it doesn’t mean that you can look at one season of UZR and say "ah ha! Now I know exactly how good someone is defensively.
Look at Dan Uggla’s UZR fluctuations. Did Jacoby Ellsbury all of a sudden become a putrid fielder? Because if he drank something, then so did David Wright, Nick Markakis, Adam Jones, Victorino, and a host of others who apparently aren’t worth much defensively because UZR says so. I’m also sure that the Phillies are happy to find out that Ryan Howard is a better fielder than Pujols and Teixeira defensively.
Pointing to WAR as THE ANSWER is a bit misguided. Do you really think that Ben Zobrist saved 18 more runs than Mauer did defensively? Really?
And second, if Pedro didn’t win MVP in ’99 or ’00, then no pitcher ever will.
by Galt on Oct 15, 2009 9:20 AM PDT reply actions 0 recs
propose a better answer
moreover, there’s a big difference between discussing true talent and recent performance
I'm not a sabermetrician, but I do play one at Driveline Mechanics.
Can't get enough of me? Check out my Twitter feed.
by devil_fingers on Oct 15, 2009 10:44 AM PDT up reply actions 0 recs
there's not necessarily a better answer. but that doesn't mean yours is rights, good, or even adequate
the cy young award is for the best pitcher in a league. the MVP, while not an award for the best position player, is not even really an award for the best player, period. there’s an awful lot of subjective tomfoolery that comes into play, and thinking you can boil the essence of what one award means, into a single number, is just an effort to exalt yourself as some all-knowing, righter than thou, pundit, when, in fact, your trying to account for everything between the numbers leaves out a huge factor of what exists without.
performance is not linear; neither should the evaluation of it be linear.
human beings, who are almost unique in their ability to learn from the experiences of others, are also remarkable for their apparent disinclination to do so.
by variablesdont on Oct 17, 2009 7:44 PM PDT up reply actions 0 recs
Well said!
Agreed. The MVP award is NOT exclusively for the best position player – and inarguably NOT the award for the BEST player either. Many times over the past several years each of us can remember instances where this award has been given to the ‘wrong’ player and we will use any and all set(s) of stats to ‘prove’ our case.
Clearly though there does seem to be an unwritten rule in effect among the voters that the MVP designation is solely reserved for positional players as long as you can reward such a deserving pitcher with the Cy Young. No pitcher has won an MVP award since ‘The Eck’ did in ’92…that was quite a while ago.
by BrokeBatMountain on Oct 21, 2009 12:08 PM PDT up reply actions 0 recs
I've heard this a milliont imes
and I agree with d_f. If we’re not using that, what are we using? Colin mentioned that UZR is less certain in terms of actual performance value for one season than wRAA perhaps is, but most people tacitly know this and discount a bit in their heads.
And again, there’s a difference between measuring true talent and actual performance. A guy hits three homers in his first game. Did that not happen just because you know that rate won’t stick? People perform better or worse than their true talent level all the time. Saying defensive performance should always be consistent is as bogus as making that claim for offense.
On a more explanatory note, I think people have issues with defensive numbers because they’re just that, one or two numbers in total. If we see a guy performing well early in the year, but his BABIP is well above expected for his batted ball profile (or whatever other measure floats your boat), we can jump in and say “that may not stick” and explain it with some reasoning. Similar with pitchers, using BABIP and strand rate. But with defense, we only have a breakdown between range, errors, and other factors that we can’t easily explain. As a result, we expect those numbers, which like wOBA run totals aren’t going to be consistent if tracked on a time scale, to stay consistent.
Given the distribution of batted balls sprayed around the field, there’s no way it wouldn’t affect fielders on the short term. And given that fielders don’t get nearly as many opportunities to field as hitters do at the plate from year to year, there’s a larger error in those run totals. But because of the denomination of “season” feeling like a sufficient sample size, people then claim UZR has inherent problems in measuring true talent when all you can do when you look at it is measure production.
Sorry, long post, but I’ve had that on my chest for a while.
Marlin Maniac, a Florida Marlins blog
Come attend Intro to Sabermetrics 101!
Check me out at Beyond the Box Score as well.
by SFiercex4 on Oct 15, 2009 12:55 PM PDT up reply actions 0 recs
Better answer, etc.
But it’s not just ‘true talent’ vs. ‘actual performance’. As even MGL, the creator of UZR says, there is measurement error in single season UZR, so that 2009 UZR is NOT ‘actual performance’ for 2009, but a sample or estimate of actual performance, and it therefore needs to be considered just a data point of performance, which needs to be informed with other data, such as prior eyar UZR, etc., unlike offense, where we have pretty accurate measurement of performance with outs, singles, etc.
KJOK
by KJOK on Oct 15, 2009 2:10 PM PDT up reply actions 0 recs
I think the logical thing to do is take UZR and regress
Maybe not as much if we were looking for “true talent” but some none the less. If we really want to get into it, things like BABIP, etc. should be regressed also, but they are a lot closer to true performance than one season of UZR is.
by lookatthosetwins on Oct 15, 2009 9:40 PM PDT up reply actions 0 recs
well, if you do that,
then you’re not addressing the problem. what you’re saying is that if you don’t agree with the outcome, then it’s the computer’s fault. but if you don’t agree with what the number is telling you, then you probably just shouldn’t use it, instead of trying to water it down in ways that make it inconsequential.
human beings, who are almost unique in their ability to learn from the experiences of others, are also remarkable for their apparent disinclination to do so.
by variablesdont on Oct 17, 2009 7:48 PM PDT up reply actions 0 recs
what are you talking about?
I don’t agree with what the number is telling me? All statistics are estimates of performance. Some have a higher degree of error than others. By regressing, we get the best estimate we can.
by lookatthosetwins on Oct 19, 2009 8:11 AM PDT up reply actions 0 recs
Just because I don't have another option doesn't mean it's still valid
UZR does not measure “true talent” on a seasonal basis, but it also doesn’t measure “performance” either.
If UZR says that Ryan Howard was a better fielder than Pujols or Teixeira this year, then there is something seriously wrong with UZR. It is flawed and shouldn’t be relied upon.
When you combine the valid offensive measures with the flawed defensive measures, it results in a flawed overall RAR.
No way did Zobrist save 18 more runs defensively than Mauer this year. Not a chance.
by Galt on Oct 16, 2009 7:24 AM PDT up reply actions 0 recs
It seems like you're coming with preconceived notions
There’s a difference between accepting the measurement error of +/- 5 runs IIRC and simply claiming “it’s not right, because it doesn’t fit my accepted notions.” I agree with KJOK that we need additional data points to make defensive measuring on one season is “more accurate.” But to say “there’s no chance this happened” is just ignoring a legitimate, if inaccurate for a season, measurement. I would prefer to do this and include other measures like plus/minus, TotalZone, and the Fans Scouting Report, but for the moment UZR is the only one I have access to in terms of runs.
Zobrist 26 runs according to UZR. The error bars are +/- 5 (presumably because the buckets aren’t as clean as singles/doubles and such are), so it’s likely between 21 and 31 runs of production. Again, it’s just a data point, so take it with a grain of salt, but I dislike this concept of “screw it, this is definitely wrong.” Isn’t that what we’re trying to get away from with objective analysis?
Marlin Maniac, a Florida Marlins blog
Come attend Intro to Sabermetrics 101!
Check me out at Beyond the Box Score as well.
by SFiercex4 on Oct 16, 2009 8:02 AM PDT up reply actions 0 recs
Are the errors really +5 on either side?
Isn’t it much more likely to be 21 than 31?
by lookatthosetwins on Oct 16, 2009 9:27 AM PDT up reply actions 0 recs
Certainly.
And I think “10” is probably closer to the truth than “5.”
by cwyers on Oct 16, 2009 9:59 AM PDT up reply actions 0 recs
If Ryan Howard's batting average for a season is .350, and Albert Pujols is .320,
do you think that batting average is inaccurate?
"Of course Kolby Rasmus was going deep! That’s what Kolby Rasmus does! You don’t give Kolby Rasmus second chances!" -Kolby Rasmus
by hazel on Oct 25, 2009 6:17 PM PDT up reply actions 0 recs
Game calling
Catcher evaluation will never be complete until you include game calling. If you believe Mauer calls a great game as many people believe, you might give him a win there. You’re generating or losing outs for your team by how you handle the pitchers after all.
by willkoky on Oct 15, 2009 9:26 AM PDT reply actions 0 recs
game calling
I'm not a sabermetrician, but I do play one at Driveline Mechanics.
Can't get enough of me? Check out my Twitter feed.
by devil_fingers on Oct 15, 2009 10:44 AM PDT up reply actions 0 recs
That's a fantastic link.
Old, but fantastic.
They call their best player "Kung Fu Panda" and they complain that people aren’t taking them or the game seriously enough? -Nick
by mikev on Oct 20, 2009 3:13 PM PDT up reply actions 0 recs
I agree with Colin
Ben Zobrist’ UZR is almost certainly wrong in terms of how good he has actually been. Whenever a player puts up UZR’s that good, it’s most likley a combination of good defense and measurement error If you regress each players defense, which you probably should even for an award, Mauer comes out better. Of course that doesn’t include clutch hitting, which should probably be considered in some way. Mauer’s been -.97 clutch wins this year, while Zobrist’s been +.5.
by vivaelpujols on Oct 15, 2009 4:53 PM PDT reply actions 0 recs
i guess
I’m not sure how much regression, but i’m not dogmatically against saying Mauer was the 2nd most valuable player this season, which is clear on my post.
And is also different from many of the “Mauer was way better than Zobrist, I just know it” posts above.
I'm not a sabermetrician, but I do play one at Driveline Mechanics.
Can't get enough of me? Check out my Twitter feed.
by devil_fingers on Oct 16, 2009 8:32 AM PDT via mobile up reply actions 0 recs
I have to say
I haven’t watched Zobrist play much but his UZR is wrong and because he played multiple positions his chances per position would lead me to believe in small sample size problems in the data. I am also not sure on Mauers True value as a defensive player or his runs above replacement values. Either Greinke or Mauer would be a good choice for MVP but Zobrist doesn’t belong in the conversation.
All Truth Goes Through Three Stages 1.It is ridiculed 2.It is violently opposed 3.Finally, it is accepted as self-evident. kinesiologist
Baseball’s biggest busts Andy Marte.
by E5 on Oct 18, 2009 5:31 PM PDT up reply actions 0 recs
I don't understand why people don't like using UZR from multiple positions
its still a full year of data, whether it came from multiple positions or not. Using the total UZR and total postional adjusments should be the same whether they played the whole season at SS or switched. The problem is that one year of UZR is all ready a small sample, and needs to be regressed.
As far as his UZR being “wrong,” I believe MGL would respond to that by saying something like this: His UZR was not wrong. What is wrong is assuming that he is 26 runs above average because of it. When used correctly, UZR is never “wrong.”
by lookatthosetwins on Oct 19, 2009 8:17 AM PDT up reply actions 0 recs
I went and looked at the true talent level for Zobrist at 2nd base
Using a regressed weighted average to get the UZR values, I have Ben’s 2nd base uzr valued at: 6.5 UZR
Also I figured it out for his time at SS and got -1.2 UZR.
From previous research I did on matched season pairs of players, there is on average a 4.3 increase in UZR for a person going from SS to 2nd. It looks like Ben’s true value is some where between 0 and 10 with 5 being a good estimate.
Jeff Zimmerman - Protecting the world from RBI's and Wins from my mom's guest house.
by Jeff Zimmerman (TucsonRoyal) on Oct 16, 2009 8:57 AM PDT reply actions 1 recs
I'd really like to see someone come out with a WAR that uses something like this
and also regresses things like BABIP for hitters, HR/FB for pitchers, etc. I don’t think this is only useful for “true talent” because there is measurement error in all of these stats.
by lookatthosetwins on Oct 16, 2009 9:30 AM PDT up reply actions 0 recs
that's about what i got for zorilla's tt in my eqrlier posy on him
A bit worse. But i’m still not clear how you’re distinguishing true talent from “true value.” I guess maybe Barry Bonds should have won about 10 NL MVPS in a row
Briefly (on mobile), regression and BABIP stuff is good and indispensable for projection andtrue talent, but to try to eliminate the gap between true talent and actual value eliminates an important distinction, and also conflates the methods used to predict an uncertain future with a reconstruction of the past (which is ambiguous, to be sure, but in an importantly different way).
I'm not a sabermetrician, but I do play one at Driveline Mechanics.
Can't get enough of me? Check out my Twitter feed.
by devil_fingers on Oct 16, 2009 9:56 AM PDT via mobile up reply actions 0 recs
WAR, from a season, is a valuation of what that player did that year, nothing more.
There are too many variables that go into person WAR value that they may not be able to control.
A person Batting 1st in the lineup is going to see an extra 50 PA than the person batting second.
The leadoff hitter on the Yankees with see more PA than the lead off hitter for the Royals because the team will come to bat more in a season.
This difference can be seen this season when Mauer played in 14 less games than Zobrist, but since he batted 2/3 in the lineup compared to 3/4/5 for Zobrist which leads Mauer to have more PA than Zobirst
Jeff Zimmerman - Protecting the world from RBI's and Wins from my mom's guest house.
by Jeff Zimmerman (TucsonRoyal) on Oct 16, 2009 10:16 AM PDT up reply actions 0 recs
right
so would it be the closest thing we have to the value of the player in that particular season?
I'm not a sabermetrician, but I do play one at Driveline Mechanics.
Can't get enough of me? Check out my Twitter feed.
by devil_fingers on Oct 16, 2009 1:43 PM PDT up reply actions 0 recs
No, grit is by far the best followed close by believing in "The Process"
Jeff Zimmerman - Protecting the world from RBI's and Wins from my mom's guest house.
by Jeff Zimmerman (TucsonRoyal) on Oct 16, 2009 6:33 PM PDT up reply actions 0 recs
I don't believe so,
because I believe there is measurement error in all of these. What we are trying to measure with WAR, is what that player did compared to what a replacement player would do in his shoes. The problem is that we don’t know that, because not every player sees the same pitches, gets the same hops, etc. Batting stats have measurement error, and that should be accounted for. UZR has an even larger measurement error.
I do believe there is a difference between a projection of talent and a description of value, I just still think measurement error should be accounted for.
I’m not going to take the time to try to properly regress all of these things, all I’m saying is that things aren’t as cut and dry as some people make them out to be. Knowing what I know, I believe its more likely that Joe Mauer was more valuable than Zobrist than the other way around, but I definitely don’t know that for sure.
by lookatthosetwins on Oct 19, 2009 8:22 AM PDT up reply actions 0 recs
Comments For This Post Are Closed
Secondary Sidebar
Related FanPosts on Driveline Mechanics
FanPosts
Community blog posts and discussion.
Certain photos copyright © 2010 by Associated Press or Getty Images. Any commercial use or distribution without the express written consent of Associated Press and Getty Images is strictly prohibited.
