The Futility Infielder

In response to Tuesday's piece about the Dodgers, reader Andy Vogel piped up in the comments to call me on the following statement about Cesar Izturis: "batting him first or second, as Dodger manager Jim Tracy did all year, is a pretty awful idea." Quite reasonably, Vogel asked the following:

How do you square this statement with data suggesting that batting order doesn't make much of a difference in team scoring? Is it just the extra at bats for Izturis you want to avoid, or is there something more to it? I'm not saying Izturis should bat leadoff, but I'm interested in your take.

My quick response was that it's still a bad idea to throw outs away when it's not necessary. Dropping a player one spot in the lineup takes away about 20 plate appearances over the course of a season. Dropping a guy from 1-2 to 7-8 means eliminating about 120 PA. Those 120 PA get redistributed to guys who are, in all likelihood better hitters than the one dropped, especially when we're talking about a guy with a career OBP below .300. It's not a huge amount but it does add up, especially in a low-run environment like the Dodgers typically play in.

Ignoring the fact that complicated research has been done on this problem, I decided to take a quick stab at modeling this via my spreadsheet. I grabbed the Dodgers' 2004 splits by batting position (1-9) from ESPN.com. For convenience's sake, I gave these positions names based on which player hit there the most or whose stat line they most closely resembled on the team, more or less. Here's what the Dodgers had:

1 Cesar  .276  .333  .394

2 Jason  .291  .349  .466

3 Milton .255  .335  .421

4 Adrian .306  .384  .563

5 Shawn  .296  .355  .492

6 Juan   .245  .304  .409

7 Alex   .248  .311  .427

8 Dave   .191  .284  .297

9 Pitch  .178  .219  .258

Yeeech. In firing off my first response to Andy, I neglected to consider just how bad the Dodger hitting was at various spots in the lineup. The #8 spot, much of it taken up by light-hitting catchers Dave Ross and Brent Mayne, was far worse than "Cesar" in the leadoff and "Juan" (Encarnacion) in the #6 spot. Of course, there's the automatic out in the pitcher's spot, which will stay at #9 because neither Jim Tracy nor I resemble Tony LaRussa.

I then set about creating a new batting order using the same nine "players." Each batting slot's rate stats (AVG/OBP/SLG) and per-plate-appearance frequency of events were held constant, but the totals were scaled up or down based on the proportion of plate appearances between old and new lineup positions. That done, I added up the team's totals and used a linear run estimator (a simple version of Bill James' Runs Created) and compared the new lineup to the old one. I lacked hit-by-pitch and sacrifice data, but that stuff tends to come out in the wash anyway.

Here is the new lineup:

1 Jason  .291  .349  .466

2 Milton .255  .335  .421

3 Adrian .306  .384  .563

4 Shawn  .296  .355  .492

5 Alex   .248  .311  .427

6 Cesar  .276  .333  .394

7 Juan   .245  .304  .409

8 Dave   .191  .284  .297

9 Pitch  .178  .219  .258

I simply dropped Cesar down to #6, moved Juan down to #7, and then shifted everybody up to the next available slot -- not an incredibly scientific method, but hardly as disconnected from reality as a lineup that leads off with the top two sluggers. Then again, batting the keystone duo of Alex and Cesar fifth and sixth is no great shakes either.

Adding it all up, this "team" has almost exactly the same totals -- three more homers, most notably -- and saves themselves literally a couple of outs. For my trouble, they gain a quarter of a point of OBP and one-and-a-third points of SLG. By the Runs Created formula all of this adds up to the whopping total of...

2.09 runs.

That's it. Two measly, stinkin' runs. I tried more complicated run-estimation formulas -- a technical Bill James as well as Extrapolated Runs, neither of them exacly appropriate because of the missing data -- and the most I added was another 0.2 runs. Of course, the gains would be more if you buried "Juan" in the landfill of some coastal state -- wait, the Dodgers actually tried that one -- and found a catching tandem that could hit. That move alone could easily gain you ten times the number of runs my suggested lineup adjustment might reap.

The bottom line is that it's far more important to have the right players out there than to spend a lot of time worrying about their optimal order. That said, if better options than a leadoff hitter with a career OBP below .300 exist, they should be taken, because that's more times your top hitters come up with men on base. It's still elementary.

I would be remiss if having gotten such a meager return on my inquiry didn't mention more rigorous studies which tried a lot harder, only to come up with essentially the same answer. In The Numbers Game, Alan Schwarz recounts valiant attempts by proto-sabermetricians Earnshaw Cook (whose estimate yielded a whopping 11-run difference) and Art Peterson (whose FORTRAN game simulations yielded "negligible" differences).

More recently Mark Pankin took a swing at the problem using a mathematical concept called a Markov Chain model coupled with some strong baseball reasoning ["1) Getting on base is everything. To much lesser extent, home run hitters should not lead off. Stolen base ability is irrelevant"]. The maximum improvement he found was a total of 16 runs, with most of the teams within 10 runs, about one full win. Nothing to sneeze at if it's the difference between golf and baseball in October, but otherwise small potatoes.

So there you have it. The next time I'm tempted to rail about batting order, I'll hold my tongue, or kvetch about why the Yankees even signed Tony Womack in the first place, let alone allowed Joe Torre to put him at the top of the... wow, I'm feeling queasy already.