Lineup Optimization and Multi-Run Homers
Why do some teams hit multi-run homers while other teams struggle? The relationship is not as simple as: better OBP, better rate of multi-run homers. I recently dug through sevens season of WPA logs and determined the baseball gods are not totally logical.
Observing the variation is one thing, but to ascribe it all to purely noise is another. Teams can control their runs per home run rate through constructing rosters and lineups predisposed towards greater home run efficiency. So we can’t consign variations to the random luck spittoon until we’ve more specifically assessed what’s happening in the lineup.
In the previous article, I briefly outlined what I called the Giancarlo Problem — where a team’s best OBP and best HR-rate are located within the same player. The Giancarlo Problem can result in deceiving team-wide statistics. So in this second venture, we are going to examine three dimensions instead of two: 1) OBP, excluding home runs, 2) home run rates, and 3) lineup positions.
Let’s begin by examining the span of 2007 through 2012, this time splitting the data by batting order position:
So, in order to maximize runs per HR, a manager wants the highest gray line in front of the highest black line. In the AL, we see a relatively clear Giancarlo Problem at No. 3 — the best OBPs are the best HR hitters. The AL No. 3 hitter has a strong 3.4% HR-rate and a .320 OBP, less HR%.
Logically, managers have posted their best OBP-sans-HR hitters at No. 1 and 2. In the NL, the No. 8 hitter gets the ol’ “a-pitcher-is-behind-you, like-hell-I’ma-pitch-to-you” boost.
I think one of the broader observations available here, to me at least, is that the No. 1 AL hitter should be hitting No. 2. Personally, I’d rather have the best OBP right in front of the best HR hitters, even if that means giving the first PA to a slightly inferior leadoff hitter.
This proposition is far from revolutionary. This is in fact the exact proposition Tom Tango and MGL make in The Book’s chapter regarding lineup optimization. Put one of the best hitters at No. 2. Put the other at No. 4 and then No. 3.
As we saw in the previous post, the 2012 Rays had one of the worst runs via homer rates of the last seven seasons. Their 1.45 runs per homer rate was tied for 10th worst in the selected period. What caused this discrepancy? How could a team widely considered among the smartest in organized sports have such an incredible inefficiency?
Here’s the outlay of their 2012 production:
Desmond Jennings, Elliot Johnson, Jeff Keppinger, and Will Rhymes had on-base percentages over .380. Not for the whole season, of course, but when batting No. 7, this quartet managed a batting average better than .300 and OBPs nearing All-Star levels. Together, their 200+ PA at the No. 7 slot helped elevate the Rays No. 7 slot to the second-highest OBP in the batting order.
Putting the second-best OBP together in front of the lowest HR-rate on the team was never in Joe Maddon’s designs.
Moreover, a low OBP from B.J. Upton and Carlos Pena — two hitters with occasional flashes of boom-power — went power crazy at the No. 2 spot and got lost on their road to their career OBP. This is harder to consign to bad luck, though, as both Upton and Pena had shown signs of OBP decline (and plate discipline issues) over a season or two at that point.
Consider additionally the 2012 Giants, a team that had the highest runs per HR rate since 2007, excluding the presently active 2013 season. These World Series champion Giants averaged 1.74 runs per HR, and look how their production appeared via lineup:
The 2012 Giants got a .338 OBP (less HR%) from the No. 2 hitters (Ryan Theriot, 294 PA; Marco Scutaro, 222 PA; Melky Cabrera, 105 PA for the top three). That positioned this team for solid home run production. Theriot (a career .341 OBP) had a .316 OBP in total last season, so his strong OBP in the No. 2 spot may have been a function of chance or matters of clustering. But other than that, batting Melky, Marco, and Ryan (based on his career numbers at least) was a smart move by manager Bruce Bochy.
At the same time, though, Bochy had Brandon Belt batting No. 6 more than any other one player, and Belt’s strong OBP and low HR-rate in 2012 went perhaps underutilized. But who was Bochy to guess that Belt would have a leadoff hitter slash after hitting 9 homers in 209 PA in his rookie season?
These two teams show both a bit of talent (on Bochy’s prudent use of the No. 2 slot) and some noise (the randomness of the Rays’ No. 7 slot).
The study of R/HR is a fertile field, and I hope to keep the inquiry moving. So please do add your thoughts and critiques and dream journal observations below.
Where do the Red Sox rank this year in terms of multiple run homeruns/homerun? They have the “Giancarlo problem” but also have a lineup that has been pretty balanced as far as OBP and power numbers so aside putting Nava in front of Middlebrooks lineup maximization seems less of an issue here. I’d also be interested to see if Fenway’s high “doubles factor” aides the Multi run homers.
Where do the Red Sox rank this year in terms of multiple run homeruns/homerun? They have the “Giancarlo problem” but also have a lineup that has been pretty balanced as far as OBP and power numbers. So aside from putting Nava in front of Middlebrooks lineup maximization seems less of an issue here. I’d also be interested to see if Fenway’s high “doubles factor” aides the Multi run homers.
Sorry I did not see the link above till I went back and re-read it. Anyways thanks for the article. I find lineup maximization fascinating.
“Observing the variation is one thing, but to ascribe it all to purely noise is another.”
It’s due to having a poor measurement basis.
“consign variations to the random luck spittoon”
It’s an accepted way of admitting ignorance.
Well that’s the attitude. No one would have ever figured out the square root of 64 without the concept of counting numbers. Gotta start somewhere, no?
I like that you bring in batting order. For years, I’ve thought that baseball innings should move out of pencil&paper thinking, where every inning is treated identically (I guess newswriters cannot count past 3 outs) and every batter was the same. Looking at the number of doubles when two men are on base to estimate the odds of scoring is GIGO when treated simply by number of outs in some random inning. Your data clearly shows the variation in scoring between batters #9 and #4.
I’ve never bothered to work it out completely, but I know there are clear variations in game outcome depending on the inning. So using lineup position and number of outs in the Game (i.e. to 27) seems a smarter basis. The time to ‘get even’ runs out, and your measurement basis should reflect that.
I thought the Book advocated putting the best three hitters in the #1, 2, and 4 positions, while the #3 and 5 hitters were to be your fourth and fifth best hitters?
Yeah, sorry, I wasn’t trying to rewrite The Book, but rather trying to focus on the 2, 3, and 4 spots. Sky did a great job of breaking down The Book’s recommendations here:
http://www.beyondtheboxscore.com/2009/3/17/795946/optimizing-your-lineup-by
I’m merely saying: In the pursuit of multi-run homers, a higher OBP in front of boppers is the key. Sometimes that means putting a normally No. 2 guy (strong OBP, but maybe a bit more power than a typical leadoff hitter) at leadoff and then put the OBP monsta at No. 2.
Lineups might also be optimized on OBP and slugging percentage (instead of strictly home runs). A home run is only one way of driving in runners. The 2013 Tigers might have such a high R/HR ratio this year because that is their best option for driving in runs. A hundred years ago most home runs were inside the park home runs, and Ty Cobb had a .535 slugging percentage on only 4 home runs. Clearly the Tigers have changed a lot in the last hundred years. What I have been wondering is if there is a lineup optimization routine that applies equally well to both the 2013 Tigers and the 1913 Tigers, or if the key components in lineup optimization vary from era to era.
Another option might be to use OBP and ISO for lineup optimization.
What JKB said. I love those graphs, but yes, HR are only one way to drive in runners on base. Obviously you would prefer someone who hits lots of doubles and triples and only a modicum of HR than a guy who hits lots of HR but few doubles and triples.
In fact, since a triple always drives in a run, wouldn’t you want to redo the whole thing using triples and HR percentage and not just HR percentage (OBP-HR is fine)?
Since doubles usually drive in a runner, why not include that too?
Very complicated issue this lineup thing is.
And by the way, after the fact, a team that has a high runs per HR is likely to have hitters who had high OBP bat just before hitters with high HR rates, and vice versa. That does not mean that managers made good or bad decisions. For that, you would want to look at pre-season projections.
A good Monte Carlo is your best way of breaking down lineup optimization.
Good idea – if I want to design a Monte Carlo simulation for, say, the Rays, I need to account for two things (that I can think of right now): 1) historical platoon rates for each player, and 2) lineup composition – there are around 15 batters on the Rays and 9 lineup spots, so there are a number of different combinations that can be used. I also need to figure out how far back in time to go for historical rates, and how to weight recent performance higher than past performance (to account for adjustments over time).
I agree with the construction of the lineups you are proposing, but Pena was not on the decline for a couple years in a row in 2012. In 2011 he posted a .357 OBP to go along with a 121 wRC+. I will give you 2010, but 2011 was a very good year as well as every year before that beside that 2010 year.