Do lineups matter?
Posted by disappointmentzone on 27 July 2006
In the last five years (and perhaps going back further) the conventional wisdom among stat-heads has been that lineups don’t matter much. More specifically, that the order in which a manager bats his players does not matter over the course of the season; the best lineup is worth an additional win per year, maybe two. In other words, if a monkey picked a lineup it would be about as productive as if Joe Torre picked it.
But in the last year there have been two major contributions in the area of lineup analysis (it’s a fairly small area) that may change how we think. The first contribution was by Cyril Morong, who did a regression analysis to determine the relative value of OBP and SLG%. Then Ken Arneson wrote a script that would compute all the possible lineups of a group of nine players based on their OBPs and SLG%s and then sort them into a hierarchy based on runs scored. Dave Pinto, in turn, used Arneson’s script to create an optimization tool he was kind enough to put on his website for everyone to use.
Using this optimization tool I was able to test my hypothesis that managers would be better off “platooning the lineup” than following the patterns all managers seem to follow (for a full summary, see the previous post). What are the conclusions?
(Feel free to scroll to the bottom of the post for the final values, but I’m going to write everything out first because it’s pretty interesting)
First I should explain how the optimization tool works. For each of the nine spots in the lineup you provide two inputs: OBP and SLG. The tool then spits out the most optimal lineups in runs produced and the least optimal.
So for the Indians I input this lineup and everyone’s OBP and SLG%:
The optimal lineup would score about 5.922 runs per game. The least optimal lineup would score about 5.488 runs per game, for a difference of .434 runs per game or 70.3 runs per season. Here are the handedness of the lineups, most optimal first, input lineup second, least optimal third:
L, L, S, L, R, R, R, R, R.
L, R, R, L, S, L, R, R, R.
R, R, L, R, R, S, L, L, R.
The most optimal lineup puts the switch-hitter in the middle of the three left-handed hitters and the five right-handed hitters, thus creating a run of five right-handed hitters or four left-handed hitters.
Of course, the problem with just inputting the season statistics of each player is that the lineup is unable to configure itself based on the handedness of the pitcher. Thankfully there are split stats that allow us to look at the production of each player against a particular handedness of pitcher. Using those statistics as inputs we get the following:
Against right-handed pitchers (same list of players as above) the optimal lineup scores 6.166 runs per game versus 5.591 runs per game for the least optimal, for a difference of .575 runs per game or 93.2 runs per season. Against left-handed pitchers (Perez now batting in place of Broussard) the optimal lineup scores 6.150 runs per game versus 5.437 runs per game, for a difference of .713 runs per game or 115.5 runs per season.
A team will hardly ever face pitchers of only one handedness in any game. Let’s assume the starter is right-handed 65% of the time and works six innings, after which there is a 50-50 chance of right- or left-handed relief pitcher and the lineup will produce only an average number of runs. Using the optimal lineup against right-handed starting pitchers will result in 391 runs through 100 games, or 21 more runs than the Indians have scored against right-handed pitching this season in 100 games. Furthermore, if the starter is left-handed 35% of the time and works six innings, after which there is a 50-50 chance of a right- or left-handed relief pitcher and the lineup will produce only an average number of runs, using the optimal lineup will result in 210 runs through 100 games, or 26 more runs than the Indians have scored against left-handed pitching this season. Through 100 games, then, the optimal lineups will score about 47 more runs than the lineups the Indians have used this season. Projected to a 162 game season, using the optimal lineups will result in about eight additional victories (assuming 10 runs equal one victory) per season. That’s the difference between 81-81 and 89-73, which is pretty substantial.
So lineups do matter, and they matter a considerable amount. What about “platooning the lineup”? Well, when I entered the split statistics of each player into the optimization tool the result in both instances was higher run production than just using the aggregate totals: 6.166 and 6.150 versus 5.922. Both split-stat lineups also featured long strings of same-handed batters (L, L, R, R, R, R, S, R, R against right-handed pitchers; S, L, R, R, R, L, R, R, R against left-handed pitchers). Based on this group of Indians, there is strong evidence to suggest that platooning the lineup would result in more victories and more wins.