Optimizing the Yankee Lineup

Yesterday Matt penned a piece calling for the demotion of Derek Jeter from the leadoff spot in the Yankees lineup.  This led to a wider discussion about the future of Derek Jeter, whether he deserved the benefit of the doubt, and whether it was reasonable to expect a bounceback in the coming months.  At the time, I opined that neither Jeter nor Gardner should be batting in the leadoff slot, and it got me to thinking about lineup optimization.  Today I’ll use Baseball Musings’ Lineup Analysis tool to determine the optimal order of the Yankee lineup.  By way of reference, you can read all the research behind the tool at Beyond the Box Score (part 1 here, part 2 here) and another piece by Sky Kalkman here, as well as over at Catfish Stew here.

I used current, up-to-date OBP and SLG data first.  Today’s current Yankee lineup, the lineup that I would expect Girardi to employ in Game 1 of the ALDS (Jeter, Swisher, Teixeira, Rodriguez, Cano, Posada, Berkman, Granderson and Gardner), is projected to yield 5.715 runs per game.  However, the Lineup Analysis tool indicates that the following six lineups would yield the most runs per game of any possible permutation:

The current Yankee lineup is on pace to produce 926 runs.  This optimized lineup would yield about 5.81 runs per game, or 941 runs over the course of the season, an increase of 15 runs.  Similarly, here are the five worst possible lineups Girardi could employ:

These lineups would yield about 873 runs over the course of the year, a decrease of 68 runs from the optimized lineup and a decrease of  53 runs from the current lineup.  ”But Stephen R.”, you could rightly ask, “the current OBP and SLG for those nine batters aren’t necessarily what we should expect for the remainder of the season!  Teixeira is back in form, Rodriguez is underperforming but could rebound, and Berkman seems to be heating up slowly.”  Fair point.  Accordingly, I’ve run the same calculations using ZiPS’ rest of season projections.  Here is the current lineup with the projections as the inputs:

This does seem awfully conservative on Swisher, Teixeira, Cano, Posada and Gardner, and fairly aggressive on Rodriguez, Berkman and Granderson.  I would probably bet the Over on all of Swisher, Teixeira, Cano, Posada and Gardner, and I might take the Under on Rodriguez and Berkman.  Granderson is a Push to me.  Regardless, this lineup yields 5.623 runs per game, which is down from the 5.715 runs per game the current lineup with current stats was yielding.  In other words, ZiPS expects a slight downturn in offensive performance from the Yankee lineup. Using the Lineup Analysis tool with these projections, we find the six best possible lineups to be:

The six worst lineups are:

Getting back to Sky Kalkman’s piece about lineup optimization, it’s important to note that the most important slots in the lineup aren’t the slots that you’d normally expect.  Kalkman puts it well, so I won’t attempt to rephrase it:

Another way to look at things is to order the batting slots by the leveraged value of the out.  In plain English (sort of), we want to know how costly making an out is by each lineup position, based on the base-out situations they most often find themselves in, and then weighted by how often each lineup spot comes to the plate.  Here’s how the lineup spots rank in the importance of avoiding outs:

#1, #4, #2, #5, #3, #6, #7, #8, #9

So, you want your best three hitters to hit in the #1, #4, and #2 spots.  Distribute them so OBP is higher in the order and SLG is lower.  Then place your fourth and fifth best hitters, with the #5 spot usually seeing the better hitter, unless he’s a high-homerun guy.  Then place your four remaining hitters in decreasing order of overall hitting ability, with basestealers ahead of singles hitters.

ZiPS and the Lineup Optimizer like Berkman, Teixeira and Rodriguez as the best three hitters going forward, with Swisher trailing close behind.  It also sees Granderson/Cano, Posada and Gardner as the three worst hitters going forward.  I would be very surprised if Cano, Posada and Gardner do not clearly outperform their rest-of-season ZiPS projections.  That said, given what we know about lineup optimization and my personal expectations of how the Yankee lineup will perform, I’d construct the lineup as follows: Berkman, Swisher, Rodriguez, Teixeira, Cano, Posada, Jeter, Granderson and Gardner.

Ultimately, the differences between the most statistically-optimized lineups and the ones drawn up by a smart guy like Joe Girardi are pretty thin.  Last night was the 106th game of the year; over the remaining 56 games the total run difference between the best lineup and the absolute worst lineup that could be invented by man is about 17 runs.  Given that Girardi will never bat Teixeira, Rodriguez and Cano 7-8-9, the Yankees ought to be safe from calamity.  It’s a fun discussion, but as Kalkman put it in his conclusion, “Finally, stop talking like the lineup is a make-or-break decision.”  This rings especially true when considering a team with as many high-caliber hitters as the Yankees.