Against My Better Judgment

This morning there were three new comments in my email for recent posts on this blog. This doesn’t sound like a lot, but for this site it is a remarkably high number, and so I figured something must be up. And what was up is that it(my MVP post) was linked at BTF.

So I went over there and took a look at the thread. And against my better judgment, I will respond to some of the criticisms of my post. This is against my better judgment because I do not intend to get into the business of going around the internet looking for people commenting about my work and responding to it, unless they actually have a point. I don't think that much of what is said in the BTF thread really has a point, or in any case is misdirected. But there are some fairly egregious misrepresentations of the analytical structure I am using, and so I will make an exception. I don't want to let the misconceptions define what I am doing.

First, let me briefly describe that analytical structure. My RAR numbers come from comparing a player to a replacement-level HITTER at his position. This approach inherently assumes that the total offense + defense value of each position is equal. This assumption is one that reasonable people can certainly disagree with; in fact, I don’t agree with it 100% myself. However, it is by far the easiest way to go about constructing an analytical system.

There was a very good thread at the Tango/MGL/Dolphin Inside the Book Blog about some of the issues involving positional adjustments, and I suggest that you read that if you want to delve into this further.

The replacement level player is defined as one who performs at 73% of the positional average (equivalent to a .350 OW%). I personally believe that this baseline is too low (for my views on replacement level in general see the “Baselines” essay on my website). But it has been, or at least was for a long time, the standard level used by sabermetricians. Bill James eventually settled on .350 as did the Big Bad Baseball Annual. Perhaps there has been some movement away from this standard; Baseball Prospectus’ VORP now uses close to 80% at each position (which I think is probably a better choice then 73%), but on the other hand Clay Davenport’s structure defines replacement as a .230 EQA which is equivalent to a .350 OW%. Anyway, I used 73% in this analysis. If you want to quibble with this, be my guest. But keep in mind that my analysis and ranking of players would change if I used a different baseline.

Now a little bit about defensive value. There are multiple competing defensive metrics out there, few of which have published comprehensive 2006 values yet. Therefore, even if I could settle on which metric to use, it is simply not feasible to include defensive values in the numbers on my site. This is why I throw in defense in terms like “approximately ten runs”, and it seems to be an afterthought. I do not disavow the value of defense, not by any stretch.

But remember that once you have gone down the offensive positional adjustments route, you are essentially assuming that the defensive value differences between positions are equal to the offensive value differences between the positions. Again, feel free to disagree with this assumption, but it is hardly one that is unique to me or one that is not used by other sabermetricians (in the linked thread, MGL to some extent defends this line of analysis, and it is exactly what the Pete Palmer analytical system does, as well as any VORP+defense measure that BP may publish. So it’s not like I’m some lone crackpot out there on this front.)

And once you have determined that the defensive value gap between positions is equal to the offensive value gap between positions, and applied an offensive position adjustment to correct for this, you are now operating in a world where +5 runs above average playing shortstop has equal incremental value (that is, when added to our offensive, position-adjusted RAR) to +5 runs above average playing first base, or any other position.

Keeping these points in mind, I will quote from a few posts on the BTF thread, and show what these misconceptions are.

#1 RMc Presents
“Er...isn't being 10 runs below average worse than being at zero? Jeter costs his team ten runs -- about one win -- on defense. Ortiz doesn't cost the Red Sox anything on defense. How can this be a wash?”

This is in reference to me saying that Jeter is probably something like -10 runs defensively. Yes, Ortiz has a big 0, but remember that my structure has assumed that total value of 1B = total value of DH. This, as other BTF posters rightfully point out, is not true and is illogical. Therefore, a DH should be penalized an additional amount for the loss in flexibility or whatever you want to call it of him being a DH. Ortiz being a DH everyday prevents Manny from DHing when he’s a little tired for instance. Where I will disagree with some of the BTF posters later is that they seem to think that this means a given DH could never be more valuable then a given 1B. They may not have stated that, but that seems to be the implication. I would say that a DH is probably costing you 5-10 defensive runs versus a first baseman. And so Ortiz does have a cost, and yes, it could very well be a wash with Jeter’s cost (recognizing that Jeter, even as a poor shortstop, is a plus defensively, but he is a cost relative to an average shortstop, which is what the analytical structure demands that we consider).

#19 The Yankee Clapper
“Putting aside the question of Jeter's exact defensive value, it is ridiculous to say Ortiz is a neutral value on defense. His defensive value is much less than Jeter's.”

Absolutely agreed. But remember again that the positional adjustment has already incorporated much of the defensive value gap. If you just compare Ortiz and Jeter to a replacement level hitter, regardless of position, then Ortiz is +81 to start out with and Jeter is +64.

#26 Gaelan
“This is a preposterous ballot. How can Ortiz be second and Ramirez be left off the ballot when Ramirez has better offensive rate stats, has more defensive value. The only advantage Ortiz has is in playing time and that's not enough to make up the difference let alone vault him so far ahead.”

Ramirez does have better offensive rate stats. Ortiz does have more playing time, which the poster kind of glosses over as if it’s not really that important. But it is. We’re talking about a difference of 128 plate appearances. If we just compare Ortiz and Jeter to a replacement level hitter, regardless of position, Ortiz is +81 and Ramirez is +59. So Ramirez has to be 22 runs more valuable defensively then Ortiz in order to close this gap.

Now let’s put in the positional adjustment. It’s now +70 to +64 in favor of Ortiz. So now it’s only 6 runs he needs to make up. But I don’t think he can do it. Ramirez, according to MGL, is truly a dreadful defensive outfielder, projected to be about -15 runs. Even if we dock Ortiz a full 10 runs for the difference between being a first baseman and being a DH, he gains 5 runs on the deal.

Now it is true that on the Red Sox, if Ortiz was a better left fielder then Manny, he would play, so Manny clearly has more defensive value within the context of the Red Sox. But on a league-wide basis, this is not true. Playing Manny in left field may be the right move for Boston, but it is a bad move for a hypothetical team. I choose not to define value in terms of the personnel of the team surrounding him, because there are way too many factors to consider and way too much uncertainty in putting tangible values to those factors. If you don’t like my way of defining value, fine, but I think that my conclusion is perfectly reasonable within that framework.

#26 Gaelan (cont.)
“In general treating firstbasemen and DH as equivelant is a huge mistake since they aren't the same. It results in overestimating the value of DH's and underestimating the value of 1B. For instance the only way to dismiss Morneau's candidacy is to treat him as a DH which is ridiculous.”

To paraphrase my new friend, this contention is preposterous. I agree that an average 1B is clearly more valuable then an average DH. But to claim that this results in underestimating the value of a 1B is just false. First of all, including DHs in the pool of 1B for the purposes of calculating the offensive positional adjustment actually REDUCES the magnitude of the adjustment, because the group of players who actually play 1B outhit the group of players who actually DH. The explanations for this have, in all likelihood, next to nothing to do with the inherent differences between playing 1B and DH, and everything to do with the personnel actually picked to fill those roles (DHs are often old, half-crippled, players nursing injuries, etc.).

So if you just calculate my RAR v. position for a 1B and a DH, yes, the 1B will get the short end of the stick. That’s why it’s appropriate to penalize a DH another five or ten runs. But it makes no difference, none whatsoever, in comparing the 1B to a shortstop.

As to the claim about Justin Morneau, as I discussed at length in my post, he, by any reasonable sabermetric approach, is not a contender for the top spot on the ballot. Would it be reasonable for someone to list him eighth or tenth? Sure. But it is pretty darn easy to dismiss his candidacy for being THE MVP out of hand. Morneau was ninth in the league in RC. He was twelfth in RG. And while he trails four DHs in this category, he also trails a left fielder, two right fielders, a catcher, two shortstops, and a third baseman. That doesn’t sound like an MVP to me, and we have not yet included the position in the numbers we are looking at. Do that, and he ranks sixteenth. Even if you throw out the four DHs, as apparently Gaelan would have you do, he is twelfth, behind a bunch of guys who play more demanding defensive positions then he does. Unless Justin Morneau is something like +25 runs defensively at first base, he has no claim to being anywhere near the top of the ballot.

#26 Gaelan (cont.)
“And how the hell does Rodriguez get anywhere near that ballot. He's not in the top ten in any total offensive metric, he's bad defensively and he certainly shouldn't get any credit for intangibles.”

I don’t know what this guy defines as a “total offensive metric”, or what stats source he uses, but I have ARod as sixth in RC. He is ninth in position-adjusted RAR(PRAR). The guys who are near ARod in terms of PRAR are Manny Ramirez, Jermaine Dye, Vlad Guerrero, Jim Thome, Victor Martinez, and Vernon Wells. We can throw out Thome by Galen-logic since he’s a DH. Martinez is a bad defensive catcher. Ramirez is a huge defensive negative. So perhaps Dye, Guerrero, and Wells are better, relative to their positions, then ARod is. ARod had a poor season defensively, from all accounts. But without most defensive metrics available yet, I don’t like to infer too much, and even if those guys are better then ARod, which they may well be, he is still in the top 15.

#30 One Alou
“And what makes this ironic is that although BPro realise that a replacement DH is a league average hitter, for the purposes of VORP they set the replacement level for DHs absurdly low (far lower than at first base, left or right field). Which is why Ortiz and Hafner have such absurdly inflated VORPs. And why, in turn, sabertypes who should know better put Ortiz and Hafner absurdly high on their ballots.”

Again, as I already stated, yes DHs deserve an additional penalty versus my numbers. However, I am not using VORP, regardless of the similarities in the method, and I never claimed to be using VORP. If I was using VORP, I have enough decency to say so and not pretend like I am presenting BP’s work as my own. And my replacement level for DHs is equal to that of 1B, but higher then that of the corner outfield spots and any other position.

#36 Gaelan (quoting the segment of One Alou’s post reprinted above)
“This is so true. Hence the ballot full of shortstops and DH's but no corner outfielders or firstbasemen.”

If he would have scrolled down the page, he would have seen my NL ballot. Of course there are no DHs on it, but there are three firstbaseman…in the top five! The only shortstop appears at tenth. The reason why my AL ballot is full of shortstops and DHs is because no corner outfielder or firstbaseman was among the top 10 players in the AL. If it was systematic bias, it would be hard as heck to have them go 1, 2, 5 in the NL, now wouldn’t it? I’m surprised somebody didn’t criticize me for giving the short shrift to up-the-middle players in the NL.

Let’s look at the AL list of RAR, with no position adjustment, except I will tell you what position each guy in each spot actually does play. The leader is a DH. Second is a DH. So far we have DH, DH. Expanding this to the top ten we have DH, DH, LF, DH, RF, SS, RF, CF, 3B, C. As you can see, no first baseman were among the top ten hitters in the AL regardless of position! It’s darn hard to make an MVP case for first baseman who can’t crack the top ten in hitting RAR.

And of course the corner outfielders who do make the list are going to be hit by a big position adjustment, relative to the shortstops and centerfielders and catchers, etc. LF and RF are much less valuable defensively then any position other then 1B/DH. The identities of the corner outfielders above are Ramirez, Dye, and Vlad, by the way. Ramirez was dreadful defensively, and I have already acknowledged that Dye and Vlad could easily be in the 7-10 range. But you just as easily could choose not to have them in there, and have them as the ones just off the ballot, which is of course what I have done.