Bill James Guide to Baseball Managers, The (15 page)

What happened here, I think, is something that any schoolteacher can relate to. Sometimes a class just gets away from you. Sometimes a classful of kids will love you; sometimes they’ll hate you. You go in the first day, smile when they do good, look at them sideways when they act up. Sometimes they decide you’re fun; sometimes they decide you’re a pain. There’s not a lot you can do about it.

There is probably no subject within the province of managing which draws more comment than batting order. Whaddaya think, should Olerud be batting cleanup for Toronto, or is it time to slide Delgado in there? Why did Brady Anderson hit leadoff for the 1996 Orioles, when they had Robby Alomar? The Orioles wound up a few games behind the Yankees. Maybe, if Anderson hadn’t hit thirty-some homers with the bases empty, they might have won a couple of games along the way.

My local manager, Bob Boone, is the target of constant criticism because he shuffles his batting order from day to day. One day Jose Offerman will hit leadoff, the next day seventh, the next day second.

It drives the fans batty, but what does it actually amount to? How many runs can a manager cost his team by misaligning his batting order? Is it a big item? Is it one of the most important things a manager does? How many games, over the course of a season, can be turned by improper lineup selection? And what is the proper order? Bobby Bragan used to talk about leading off Henry Aaron, and actually did it a few times, on the theory that Aaron would get 50 more at bats a season batting leadoff than he would batting cleanup.

Some managers clearly think that the subject is important. Others seem to imply that it isn’t. Billy Martin, at least once, put the names of his players in a hat and drew them out at random, trying (successfully) to shake his team out of a slump. Some people think it is crucial to get speed at the top of the order. Casey Stengel often used leadoff men like Bob Cerv and Elston Howard who couldn’t get out of their own way, but did a good job of getting on base.

Let’s start with the broadest question: How much difference does it make? I programmed a computer to simulate games played by the Chicago Cubs in 1930, with this lineup:

The 1930 Cubs had a classic offense, one of the best ever. The leadoff hitter, Woody English, had 214 hits, drew 100 walks, and scored 152 runs. The cleanup hitter, Hack Wilson, hit 56 homers and drove in 190 runs. Many other members of the lineup had impressive hitting stats. The lineup above was more-or-less the team’s regular lineup, although Footsie Blair usually hit second, and Bell and Stephenson were in the lineup and out of it. The team scored 998 runs, one of the highest totals in major league history.

With that lineup in place, I then ran the team through 100 simulated seasons, 16,200 games, to see how many runs they would score.

Then I took the same set of players, and put them in what, it seemed to me, would approximate the worst possible sequence in which the players could be arranged. I had the pitchers bat leadoff, while Hack Wilson, with his 56 homers and .356 average, was assigned to hit ninth. I put Les Bell, a .278 hitter with few walks and little speed, batting eighth, so that Wilson would be following the weakest available hitter, and otherwise tried to place the better hitters toward the end of the order, but without putting them together so that their bats could interact with one another. This was the “illogical” lineup:

Then I ran that team through 100 seasons, to see how many runs they would score. The best order, and the worst order, you see, although (for reasons I explained before) we can’t really be sure what the best order or the worst order might be.

How much difference was there between the “correct” batting order, and the same players in an obviously irrational order? Surprisingly enough, very little. The “normal” batting order scored 99,766 runs in the 16,200 games, or 6.16 per game. The screwy batting order, with Hack Wilson batting ninth, scored 94,800 runs, or 5.85 per game. The difference is 5%, or 50 runs per season.

I am little reluctant to report this study, because I know that most of you aren’t going to believe me anyway. I am not the first person to study this issue in approximately the same way. Dick Cramer, twenty years ago, programmed a computer to simulate offenses. Dick constructed a lineup of six pitchers and three Babe Ruths. He then put this lineup in what seems like the most logical order (the three Babe Ruths hitting one, two, and three), and then in the most illogical order (the three Babe Ruths hitting three, six, and nine). By spacing out the three Babe Ruth lines among a string of unproductive hitters, he minimized the interaction between the only good hitters on the team.

I’m uncertain whether Dr. Cramer’s study was ever published, and I can’t quote specific data from it, but I know essentially what he found: the same thing I found, twenty years later. It doesn’t make much difference what order you put the hitters in. I know that you’re not going to believe me, because when Dr. Cramer told me about his study, I didn’t believe him, either. It’s counterintuitive. What do you mean, it doesn’t make much difference? We can all think of a thousand reasons why it
should
make a difference what order you put the hitters in.

Let me explain as best I can why it doesn’t. In the “good” model, Hack Wilson batted 64,237 times, hitting 6,082 home runs and driving in 20,818 runs, or .3241 RBI per at bat. (In real life he had .3248 RBI per at bat.) By moving Hack to ninth in the batting order, we took almost 8,000 at bats away from him, dropping him to 56,310 at bats. As a consequence of that, we cut him back to 5,490 home runs. By putting a weaker hitter batting ahead of him, we reduced his RBI per at bat from .324 to .294. Thus, we cut Wilson from 20,818 RBI to 16,558.

The thing is, though, that
you can’t move everybody down in the batting order
. When you move one guy down, you have to move somebody else up. And you can’t put everybody hitting behind the weakest hitters in the lineup. When you put Hack Wilson hitting behind one of the weakest hitters in the lineup, you have to put somebody else batting behind Riggs Stephenson (.367), and somebody else batting behind Kiki Cuyler (.355).

Of course, the loss on Hack Wilson is greater than the gain on the other hitters, because we moved the
best
hitter to the
worst
spot. We succeeded in eliminating 20% of Hack Wilson’s RBI—but 75% of those RBI were picked up by other hitters. And in the final analysis, the team lost only 5% of its total run production by having the hitters in the worst order that I could come up with.

With the “logical” lineup, the best innings start when the leadoff hitter leads off, and the second-best innings are when the second-place hitter leads off. In this study, Woody English led off in 31,251 innings, during which the team scored 27,849 runs, or .891 per inning. When the second hitter (Riggs Stephenson) led off, the average was .817 per inning:

There is a logic to that. Woody English not only led off the
best
innings, but also the
most
innings—in fact, he led off twice as many innings as any other player. On a typical team, the number five hitter is second on the team in innings led off. On this team that wasn’t true because they had so many great hitters that they tended to push those second-inning leadoff opportunities down further in the lineup.

With the “illogical” lineup, we got these results:

Now the players who lead off most often are the worst leadoff hitters, the pitchers. This creates fewer runners, and fewer runs.

But when you work through the math, the net loss in runs is small. Charlie Grimm now has Stephenson, Harnett, and Cuyler coming up behind him, rather than Blair and the pitchers, so the innings in which Grimm leads off now yield .692 runs per inning, whereas in the earlier study they yielded .545. Gabby Hartnett is not a natural leadoff man, but the innings in which Hartnett leads off now yield .722 runs/inning, rather than .530, because Kiki Cuyler is now batting behind him, and Hack Wilson is now the cleanup hitter in Hartnett’s innings. The innings led off by English, Cuyler, Stephenson, Wilson, and the pitchers are less productive than they were before, but the innings led off by Blair, Hartnett, Grimm, and Bell are more productive.

With the “good” lineup, this team scored 14,366 runs in the first inning (in 16,200 games). With the “bad” lineup, they scored only 9,147 first-inning runs—down 36%.

But in the
second
inning, the “good” lineup scored only 9,740 runs. The “bad” lineup scored 11,647—up 20%.

In the third inning the ill-constructed lineup is worse off by 13%, because the top of the order is up again.

But in the fourth inning, they’re up by 7%.

So if you add that together, what do you have? Down 36, up 20, down 13, up 7; that’s a total of down 22, for four innings. That’s a little more than five per inning. Actually, it’s a little more than that; it’s 9% through the first four innings—but 2% in innings five through nine. I’ve given you the numbers you need to run the math if you want to, but when you add everything up, the net loss in runs is only 5%.

Now, if the difference between a reasonable batting order and a completely unreasonable batting order is only 5%, what do you suppose the difference would be between two reasonable batting orders? That’s right: it’s nothing. You take any two reasonable batting orders for any team, put them on a computer and play a hundred seasons, and you’ll find they score just as many runs one way as they do another.

Of course, there is no legal requirement for you to accept the findings of the computer. This is an offensive simulator of reasonable sophistication. You might wonder, for example, whether the offensive simulator might miss the value of speed by some clumsy process of routinely sending all runners from first to third on a single, or no runners from first to third on a single. We have been assiduously gathering information on the play-by-play of major league games for more than ten years now, and we have good information on how often runners go from first to third on a single, and on how speed affects that outcome. That information is built into the model. The fast runners go from first to third on a single much more frequently than do the slow runners.

The model is sophisticated enough to specify which field the ball is hit to, and thus is sophisticated enough to send a runner from first to third slightly more often on a single by a left-handed hitter than on a single by a right-handed hitter, because the left-handed hitter is more likely to pull the ball into right field than is the right-hander.

At the same time, real baseball games remain vastly more complex than our statistical models of them. For that reason, it is quite possible that some portion of the difference we are attempting to measure here is eluding us. Many people would argue, for example, that when there is a fast runner on first the hitter at the plate is much more likely to see a fastball, and thus more likely to get a hit. If the first baseman has to hold the runner on, people will tell you, this increases the size of the hole on the right side, and thus increases the batting average of a left-handed hitter in that situation. Thus, they would argue, there is a synergistic interaction between speed and hitting. Our model does not simulate any such interaction.

Our model does not simulate any such interaction, because there is no evidence that such an interaction occurs. The evidence, in point of fact, is the opposite. If you look at a
real
base stealer, and you check the batting averages of the man who bats after him, you will normally find that that average goes
down
with a man on first base, whether the hitter is right-handed or left-handed. Why? Because the batter is often taking a pitch to allow the runner to steal, and for that reason is often hitting behind in the count, whether the runner attempts to steal, or whether he doesn’t.