THE BOOK--Playing The Percentages In Baseball

I looked at this a while ago with a sim that I had built.  The sim was very basic; it didn’t include baserunning (it used very rudimentary and static baserunning rules), didn’t include handedness, etc.  I did a “Star and Scrubs Lineup” and a “Even Steven Lineup.” The former had a one great player and a bunch of average ones and the latter had players that were exactly the same.  I matched the lineup by avg/obp/slg, which is obviously not as good as lwts, but like I said, this was a while ago.

I found rpg differed by .01, the run distribution was pretty much identical, and moving the star up and down the lineup resulted in a change of <0.5 wins over a whole season.

It was a pretty rudimentary study, but i think it showed that the effects of lineup “balance” are secondary at best.

Sal, could be.  As I said, my experiment, for various reasons, definitely didn’t answer the question, although it may have shed some light on it. Just a simple Markov lineup simulator should be able to answer the question though. Common sense tells us that since basic lwts is based on average baserunner opportunities, that if, for example, you have a slugger in your lineup, that if no one is on base for him, the value of his slugging will be less (than lwts suggests), and if, for example, you have a high on base guy in your lineup, but there are no sluggers, the value of his on baseness is going to be less (than lwts suggests).

But those are extreme examples (no one on base for the slugger or no one to drive in the high on base guys).  In a “real” lineup, it probably doesn’t make much difference.  Probably the thing that matters more than a “balanced” lineup in terms of each player’s OPS or lwts is to have a balance of on base guys and slugging guys.

That is probably what I should look at. How does a team of high on-base guys with a total lwts of X compare to a team of balanced on-base and slugging with the same total lwts or to a team of slugging guys only also with the same total lwts.

Of course when I say total lwts, that lwts is computed by using our normal, static lwts values for each event (s,d,t,hr,bb,out), which would be incorrect for the imbalanced lineups.

Here is a link to a study I did on balance

http://cyrilmorong.com/BalanceBTN.htm

it also has links to studies by Gassko and Woolner. I think my evidence was mixed, based on actual teams and regressions. I think Woolner found some slight positive benefit from balance

Cyril, thanks.  Although I am not crazy about using regression for things like this when you can (and should) easily (and more definitively) answer the question using Markov or actual lineup simulators.  Regression is nice in addition to the sim method just in case there is something in real life that the sim is not capturing.

Would comparing RP27 (runs per 27 outs) be more accurate?
vr, Xei

You’re welcome. What I wonder about is whether or not teams should take balance into account when making personnel decisions.  It may be hard to figure that out. But I think what we might look at first is how many more runs the most balanced team would score compared to the least balanced, holding overall hitting constant (I think you were tyring that or something like that). Then if at least potentially there is a big difference, it might be worth it for a team to sign a guy who is not quite as good as someone else but keeps the team more balanced (or maybe sign two guys who are close instead of one really good guy and one really bad guy even though their combined numbers are the same). But commentators often say a team is more balanced and that gives them an edge when we don’t know that for sure or how big the edge would be.

Also, do you need to be balanced in both SLG and OBP or just one? In your simulation, were their intentional walks? Did the unbalanced team get more and did it affect scoring? In the extreme, balance might matter. If you have two guys who hit a HR every time up and 7 replacement hitters, it might make sense (and I am just guessing here), for the opposition to always walk the two HR guys and you might not score many runs.

Can you run a simulation with, say, 2 great hitters, like Ruth and Bonds, and 7 of the weakest plausible hitters and see how many runs they score compared to a team with 9 identical guys and the same overall team LW (or OBP & SLG)?