Much was made last season of some of Willie Randolph’s questionable methods in lineup construction, particularly concerning the placement of David Wright, Carlos Beltran, and Jose Reyes. Expect more of the same this year if Willie bats Paul Lo Duca second and Wright lower in the batting order as many assume.
But what should the lineup look like? Over at Beyond the Boxscore, Cyril Morong wrote a two part column last week detailing the value of on-base average and slugging percentage at each lineup position. The results are very interesting and have been widely discussed recently.
There are several findings which can be very counterintuitive at first, as Dan Scotto mentioned on Sunday. Perhaps the aspect that will grate on baseball traditionalists the most is that the number three slot should not be the most hallowed slot in the lineup, reserved for the “best hitter.” Instead, it’s a spot where you might want to stick one of your weaker hitters in the lineup, or at least a guy who doesn’t quite fit anywhere else.
The number two hitter really becomes the prototypical third hitter, as Dan writes: “The 2-hitter should be the lineup’s most balanced hitter, a good combination of OBP and SLG. David Wright fits the bill here, as does the player I chose, Chase Utley.” Oddly enough, Wright doesn’t make it as the Mets best OBP/SLG combo, as you’ll see below.
Even though these ideas might not be accepted by traditionalists, they won’t necessarily be embraced by statheads either; much of this is new to the sabermetrically oriented also. Many of them, myself included, have always believed it best to concentrate those who make the fewest outs (those with the highest OBP’s) in one spot, in an effort not to isolate the team’s best hitters. Instead, this analysis suggests the opposite; spreading the outs around will be most conducive to scoring runs.
This irritated me a lot at first too, but eventually I was able to reconcile the thought. While placing the best hitters in a knot will give you a very good chance of scoring in the first and maybe the second inning, it gives you a terrible chance of scoring in the third inning, and may cause difficulty as the lineup starts to make more “awkward” breaks (getting the 8,9,1 hitters instead of the 1, 2, 3). Basically, instead of throwing away an inning, you give yourself a slightly higher chance throughout the game. It also works out a little better on the occasions that your better players fail to get on base.
Really, I think that is what makes this sort of lineup construction work the best: by spreading things out a little more, you’re not as reliant on the best case scenario. You have a better chance even if the optimal trio of hitters doesn’t get it done.
Luckily Ken Arneson and David Pinto have made it easier to see specifics. Pinto has provided a tool on his website that allows readers to insert any nine players with their OBA’s and SLG’s, and then provides the number of runs per game that lineup will score. It also gives the best and worst lineups that can be constructed from those nine players. Using this tool, I supplied the ZiPS projections (Note: I really wanted to use PECOTA projections, but due to a computer disaster wasn’t able to; all I had on my other computer was ZiPS. Apparently, commentor JZach used PECOTA projections for his own experiment.) for the nine Mets most likely to play in the Mets lineup, alternately using Xavier Nady and Victor Diaz in right.
Here’s the lineup I think Randolph is most likely to use, alongside the lineup I think most others would somewhat realistically like to see:
Jose Reyes Jose Reyes
Paul LoDuca Carlos Beltran
Carlos Beltran David Wright
Carlos Delgado Carlos Delgado
David Wright Cliff Floyd
Cliff Floyd Nady/Diaz
Nady/Diaz Paul LoDuca
Kaz Matsui Kaz Matsui
Pedro Martinez Pedro Martinez
Willie’s lineup will score eight runs or so less than the other lineup over the course of a season, regardless of whether Nady or Diaz gets all the playing time. It’s not a huge difference, but every run counts.
Here’s what the optimal lineup would look like:
A couple of items of note. First of all, there are two models available with Pinto’s tool, which give different lineups for the Mets. One uses a 1959-2004 sample size, and the other uses just 1998-2002 data. I used the 1959-2004 model for three reasons: it used the larger sample size, produced a more “traditional” lineup, and predicted a smaller difference in how many runs individual lineups would score. The smaller sampled model had a disparity of fifty or so runs between Willie’s lineup and the optimized lineup, which seemed a little too effective considering the players involved weren’t changing at all. I think the other model is just a “safer” choice. This lineup scored 13 more runs than the “Other” lineup and 21 more than “Willie’s.” Also, Matsui’s projections predicted a .339 OBP, but if he repeats last year’s numbers, he and LoDuca swap spots.
The best hitters are hitting one-two, and third is something of a falloff, if not as severe as the other model. It picks up in the four-five spots, but six represents the second worst hitter in the lineup. Seven generally does a little better in the on-base department while eight a little better in power. The worst hitter still hits ninth (the more drastic model bats the pitcher eighth). It’s not too different from the usual sabermetric argument; the better hitters generally hit first, but they don’t strictly.
Either way, it’s going to take a very progressive manager to use a lineup like the one above, and it won’t be Willie. Though, it would be fun to see what would happen if someone tried.