THE BOOK--Playing The Percentages In Baseball

I have not read the article yet, but I gave a lot of information to the author.

The worst fielding 1B is (should be) around 1 to 1.5 wins below average and the worst fielding LF is around 1.5 to 2 wins below average.

Anyone beyond that should be moved to another position or DH’d.

Also, Manny’s -32 last year of course does not represent his true fielding talent.  I told the author that that is around -20 runs (I think I have Manny projected at -21 per 150 in 2007).

So yes, he should be at DH or 1B, which would “increase” his overall value.

I also told the author that Ortiz at 1B is probably no worse than -15 (although I suppose that there are some players who absolutely cannot play the field - i.e., they would be worse than -15 at 1B), so they would probably be better off with Ortiz at first, Manny DH’ing and someone else in LF.

Whether that would indeed be better than what they have now depends on what kind of new LF they could get, how much they would have to pay him, etc., versus who they use at 1B.

Of course there are many other issues involved.  You can’t just move a player (defensively) at will when that player is a highly-paid superstar.  Think Jeter, Griffey, etc.  Also, maybe Manny would experience a large DH penalty hitting-wise if moved out of LF.  Ortiz probably does not have a DH penalty (i.e., he would not hit better if playing 1B - maybe he would hit worse).

The question of whether Manny should be responsible for where he plays, in terms of his defense-neutral value, is a tricky one I think.  Not so clear cut as Tango makes it to be (he thinks he should not be penalized).

In any case, his overall value is what it is as long as he is playing LF.  If he played another position for the Sox or for another team, it would be different.  The Sox and any other team are obviously welcome to compute his value at any position they may think he might play.  That is one of the beauties of “numbers analysis.” As long as you understand what you are doing and you have as much information as possible, you can draw the correct conclusions given any number of contexts (where Manny plays, your other personnel, etc.).

The author sort of misquoted me in the last paragraph and provides some misleading information.

I told him (the author) that the worst 1B we see in MLB are around -15 runs per 150 and the worst LF are around -20.  That -15 is probably a shade high and the -20 is probably a shade low, but that is the right ballpark.

First of all, those numbers are what we typically see at the far end of the defensive talent spectrum for those players that are allowed to play.  In terms of estimating what a player might do who has never played a particular position before, and especially one who is a DH (and rarely plays any position at all), those numbers might be too conservative.  If I were to play LF in MLB, I would be -60 or -70 runs (or maybe worse), but certainly not -20 runs.

The other problem with that last paragraph is that he is mixing apples and oranges by saying that Ortiz at -15 would be only half as bad as Manny at -32 (last year’s UZR).  I esplained to the author that the -32 was a sample number and did NOT represent Manny true fielding talent (and his projection for this and future years), and that the -15 was the practical downside limit in fielding talent for a 1B but certainly not the one-season sample limit. 

So he should have said something like Ortiz at 1B would be 5 runs better than Manny in LF (I told the author that -20 or -21 was Manny’s true talent), and not 15 runs better as he says or implies.

MGL,

Thanks for hanging tough in the dialogues here and at BTF.  The sensible bottom line that people seem to be missing, but which you’re trying to express, is that when you have a one-season sample, you should not get yourself in an uproar over whether one zone measurement puts the guy at -15 and another at -30, for both theoretical and practical reason.

The theoretical reasons are (i) ‘random’ measurement error, which usually settles down when you have a couple of consecutive full-time seasons, and (ii) different (and oftentimes equally sensible) approaches to aggregating and context-adjusting zone data, which in the majority of cases don’t result in terribly wide discrepancies.

The practical reason is that whether the guy is at -15 or -30, he has a problem, and the fact that teams now have the means, thanks to systems like yours, to identify these problems is incredibly useful.  Many, many (pun intended) fielders of yore were allowed to play positions they had no business fielding, because management couldn’t say with confidence they were objectively bad and even approximately how much it was costing the team.

One of the other things that should have been brought up in the article (and I am an e-mail ‘friend’ of Dan) is that even if Manny is at -30, he’s not breaking any records.  Tom Tippett’s 2003 defense article said it is quite common for the spread in plays made at any position in any single year to be +/-30 or even 40 plays, based on his zone data.  Dewan’s Plus/Minus had Jeter at -39 groundouts in 2005, which is, as you and I and Tango know, is about 30 runs lost.

For what it’s worth, I saw Manny at Yankee stadium a couple of years ago, and he misplayed two balls that should have been outs into extra bases.  (No errors were charged.) As I was looking at him, I was reminded of Bill James’ essay about Greg Luzinski--he really had no idea where a ball hit over his head was going to go.  Someone at BTF said that Manny is good at balls hit in front of him.  Since Fenway cuts off so many balls behind him, maybe he’s helped by The Green Monster!

MGL:

FWIW, when Manny was first acquired by the Red Sox in 2001, he played the first 54 games as the DH, before making his LF debut on June 4, 2001 at Yankee Stadium.  Through June 3, Manny was hitting .381, with .458 OBP and .710 SLG for a nice 1.168 OPS.  At the time, a lot of people were reluctant to get him into the field because he was raking it so well as the full-time DH… so, he might be a good candidate for a full-time DH, though that slot is filled right now in Boston…

I thought that Fenway had the weird effect of making Manny’s fielding runs in UZR much lower because players are expected to do better, however the actual damage (in real runs) Manny does is less than 32 by far.
Also David Ortiz (or at least Hafner) is kept off first not just for fielding reasons, but also for health reasons.

"when you have a one-season sample, you should not get yourself in an uproar over whether one zone measurement puts the guy at -15 and another at -30, for both theoretical and practical reason.”

I tend to disagree. From MGL’s posts on the BTF thread, Manny played about 137 games in Fenway/LF in 2005-06. Using one methodology he was -61 over that period, using another methodology he was -27. (Those are absolute figures, not per 150.) If I work for the Red Sox, I want to know which number to trust! Do you think he would still be on the team if they believed the -61 figure?

More broadly, how much noise in our defensive stats is caused by inaccuracy in measurement? I can believe that Fenway is a special case, but how many players each year are being mis-estimated to the tune of 15 runs?

Finally, what about differences in BIP distribution? That comes from both sides: Manny fields only balls hit off Red Sox pitchers, and almost all other LF in Fenway field balls hit by Red Sox batters. This could introduce bias into the defensive stats, which I’m okay with as long as the magnitude is small, but do we really know that?

You are correct that we have two sources of error, one based on recording the observed data, and the other based on the sample size.  As well, “positioning” is implicitly accepted as a skill of the fielder, and the manager/coaches are not involved at all.

When it comes to Fenway and a few other OF parks, I’d be wary of whatever adjustment is made.  In these cases, I’d rather have Fenway as its own line (i.e., have one line for overall, and one line for Fenway, and not just for Manny, but for all OF).

Number Six,

I wasn’t aware of the two-year 34 run discrepancy.  That is something that needs to be looked into.  Also, I would be concerned if there were more than a couple 15-run single season discrepancies across all fielders and positions that couldn’t easily be chalked up to obviously unusual circumstances / park effects such as Coors and The Green Monster.  When I’ve written about UZR I’ve also tried to compare its results with other zone systems, such as Tom Tippett’s, David Pinto’s, and plain Zone Rating denominated into runs in the way I proposed in a February 2005 article, as well as Dial’s method proposed several months later.  No system is perfect.

I thought--but I could be wrong--that the whole point of zone systems is that they _can_ handle weird batted ball distributions.

The good thing about this Manny article is that it highlights the challenges of using zone data, which is rich but also proprietary and complicated.  There really should be several people working with it in different ways to make sure there isn’t a big mistake being made somewhere.