As Phil Garner goes all conventional and suggests that Preston Wilson will bat "fifth or sixth," other people around the web are trying to take a less flippant approach to lineup building.
And if you thought leading off with Brad Ausmus, as Stephen Stein's Lineup Toy suggested, was wack, wait til you get a load of this. . . .
It begins, as many of these counterintuitive sabermetric-type things do, over at Beyond the Boxscore. Writer Cyril Morong quite reasonably suggests that the relative importance of OBP and slugging for any particular position in the order might vary. For example, it's no great stretch to imagine that onbase percentage has a lot more to do with a leadoff hitter's success than slugging does, or that the cleanup hitter must balance the two skills more evenly.
But by doing some fancy math I don't understand (including something called "Markov Chains"), the author claims to be able to put a numerical value on these percentages.
From there, it didn't take long before an enterprising A's fan applied Morong's formulae and constants to a real lineup with an eye towards grading it. And then Ken Arneson at the Baseball Toaster wrote a perl program that would use Morong's work to systematically grade ALL lineup possibilities, all 362,880 of them.
It was at that point that I became aware of these efforts at lineup otimization, and with much help from Astro Annie and her perlgenius son, I've been able to run the program for the Astros.
With the Lineup Toy, I'd basically been interested in three versions of what the Astros might run out there April 3 vs. the Marlins. Biggio Ensberg Berkman Wilson Lane Everett Ausmus and Oswalt are the eight etched in clay, with any of the following three players taking the last slot in the order: Bagwell Taveras or Burke.
Here's the top five permutations for each option. The numbers at the left are runs scored per 162 games:
With Bags at his 2005 level
816.42 Berkman Ensberg Bagwell Lane Wilson Biggio Everett Oswalt Ausmus
816.23 Berkman Ensberg Bagwell Lane Wilson Everett Biggio Oswalt Ausmus
815.68 Berkman Ensberg Bagwell Lane Biggio Everett Wilson Oswalt Ausmus
815.55 Berkman Ensberg Bagwell Biggio Wilson Lane Everett Oswalt Ausmus
815.47 Berkman Ensberg Bagwell Lane Biggio Wilson Everett Oswalt Ausmus
With Willy T
799.08 Berkman Ensberg Taveras Lane Wilson Biggio Everett Oswalt Ausmus
798.88 Berkman Ensberg Taveras Lane Wilson Everett Biggio Oswalt Ausmus
798.34 Berkman Ensberg Taveras Lane Biggio Everett Wilson Oswalt Ausmus
798.21 Berkman Ensberg Taveras Biggio Wilson Lane Everett Oswalt Ausmus
798.13 Berkman Ensberg Taveras Lane Biggio Wilson Everett Oswalt Ausmus
With Chris Burke
797.61 Berkman Ensberg Burke Lane Wilson Biggio Everett Oswalt Ausmus
797.42 Berkman Ensberg Burke Lane Wilson Everett Biggio Oswalt Ausmus
796.87 Berkman Ensberg Burke Lane Biggio Everett Wilson Oswalt Ausmus
796.75 Berkman Ensberg Burke Biggio Wilson Lane Everett Oswalt Ausmus
796.66 Berkman Ensberg Burke Lane Biggio Wilson Everett Oswalt Ausmus
Hmmm. Besides reminding us (and maybe Uncle Drayton) that we should be really really hoping that Bagwell can come back to play even at his diminished levels of last year, the first thing you notice is that the formulae on the whole would rather see Berkman lead off. In fact, taking just the Bagwell version of the lineup, the top 18 versions of the lineup suggest that the Astros place their best slugger (and best onbase guy) at the top of the order.
I have to admit: I'm pretty skeptical. Other points these lists bring up, I think, are intriguing. Batting the pitcher eighth, or one of your better OBP guys last, for example, are ideas that make a sort of sense if you play with them for a little bit.
But leading off with your best OPS guy? I dunno.
I think it was a couple years ago where Felipe Alou--probably during some postgame aguardiente drunk of his--had suggested the idea of leading off with Bonds down the stretch simply to maximize Bonds' at bats, and beyond the fact that such a ploy would only make a difference one game in nine (right?) the gambit does make some sense.
If you're trying to get Barry Bonds some home runs.
But if you're just trying to put runs of any sort on the board, seems like instead of maximizing at bats for Lance, you'd want to maximize men on base per at bat.
Yeah I know my <gasp> conventional thinking is showing, but why would you use Roy Oswalt to set the table for Lance? If the people in front of him don't get on base, how is this gonna maximize runs? To see how reactive the formulae were, I replaced Oswalt's and Ausmus' SLG and OBP's with 0.00 and ran the thing again. Surely that would kick Lance out of the leadoff slot.
But it didn't. The top lineup (I used Taveras as player # 9, for no particular reason) when I radically emasculated the bottom of the order in that way went like this:
580.15 Berkman Ensberg Wilson Lane Biggio Oswalt Everett Ausmus Taveras
580.15 Berkman Ensberg Wilson Lane Biggio Ausmus Everett Oswalt Taveras
579.78 Berkman Ensberg Everett Lane Wilson Oswalt Biggio Ausmus Taveras
579.78 Berkman Ensberg Everett Lane Wilson Ausmus Biggio Oswalt Taveras
579.36 Berkman Ensberg Biggio Lane Wilson Oswalt Everett Ausmus Taveras
In other words, when the leadoff hitter is guaranteed to be batting with the bases empty 2/3 of the time (and with two outs 2/3 of the time), it still suggests that the leadoff guy be your best slugger.
I don't buy it.
But maybe I'm missing something?
These files are 25 megs each, and are 360,000 lines long, but if you want to take a look at the output files, I'll have them up for a while here: