Bill James Guide to Baseball Managers, The (19 page)

In real cases, however, the data is very different. Four National League teams that year actually got better performance out of their number-two hitters than their number-three hitters. For the St. Louis Cardinals, on the other hand, the number-two hitter was usually Ozzie Smith, and the number-three was Gregg Jefferies. His batting average was 54 points higher than Smith’s, his slugging percentage 129 points higher.

Further, in real situations, other factors frequently will operate to accentuate those differences. The .280 hitter at bat may actually be a .260 hitter if the opposition has switched to a left-handed pitcher, while the .340 hitter on deck may actually be a .350 hitter. A player with a 1-in-30 chance of hitting a home run in theory might have a 1-in-15 chance if the game is in Colorado, but 1-in-40 in St. Louis. If the wind is howling in from right field, it might be half that. A hitter may be nursing a minor injury, which might make him vulnerable to a double play.

So the variation in performance between batting-order prototypes is not representative of the factors at play in a real-life situation, and therefore what is true in the model might very probably not be true in many real-life situations.

Perhaps an easier way to explain why this argument is unconvincing is simply to apply the same logic to another strategic option, that being the intentional walk. In Palmer’s model, of course, the intentional walk would always be a bad play, since it would always increase the expected runs for the opposition.

No one believes that having runners on first and second with one out is a better deal (for the defense) than having just a man on second with one out. But if Barry Bonds is at the plate, and Rikkert Faneyte is on deck, you still might order the walk. The advantage you’re going for doesn’t reside in the inherent situation; it resides in the identity of the hitter. The same might well apply to the sacrifice bunt.

Another issue raised by the Palmer/Thorn and Dewan/Zminda studies is that the bunt might “fail.” Palmer and Thorn concede that the bunt, in theory, might be a good play in certain situations if it works—for example, with runners on first and second, no one out, and the home team behind by one run in the seventh inning:

The STATS
Scoreboard
study uses a similar method, relaying the results when the sacrifice bunt succeeds, and when it fails.

This approach, however, assumes that there are two possible outcomes from a bunt attempt—a successful sacrifice bunt, or a failure. Suppose that we use Palmer’s data and Palmer’s method, but we ask this question instead:
If the batter attempts to bunt for a hit, and gets the sac bunt as a consolation prize, how often must he succeed in order for it to be a good gamble?

The answer, as it turns out, is “not very damned often.” Using Palmer’s data, the expected runs for the team with a runner on first, no one out, are .783. With a runner on second, one out, the expectation is .699. But with runners on first and second, no one out, the expected runs increase to 1.380.

Asked this way, how often does the bunter have to “succeed” in order for the bunt to be a good play? Answer: 12.3% of the time. If he can get a base hit one time in eight, and get the sacrifice bunt the other seven times, the team can expect to score more runs with than without the bunt.

A good bunter, like Brett Butler or Otis Nixon, can bunt for a .300 or .350 average with the bases empty, maybe higher. Steve Garvey, who ran about as well as Bill Clinton, used to bunt for a hit ten times every year. If you send the runner from first, the defense isn’t going to get a force out, although there is some risk of a pop fly double play. The question is, can the batter bunt .123 with a runner on first? If he can, the bunt is probably a good play.

Is it fair to evaluate the bunt by considering a base hit and a successful sacrifice to be the only two possibilities? It is probably as fair as it would be to consider the only possibilities to be a sac bunt and a failed bunt. One cannot evaluate the impact of the sacrifice bunt, without dealing with the full range of possible outcomes.

Well, what is the full range of possible outcomes? In two editions of the
Scoreboard
, STATS published lists of how many times everything happened in major league play. In 1990, for example, there were 469 foul fly outs, 844 runners thrown out advancing, and 31 cases of batter obstruction. There were 20 times when a runner was caught stealing, but safe on an error.

Although this data is not precise in all the details we need, it is useful for modeling the problem. There appear to be seven outcomes of a sacrifice bunt attempt which are relatively common. Those are:

• a base hit
  
• an error
  
• a fielder’s choice/all safe
  
• a fielder’s choice/failed bunt
  
• a successful sacrifice bunt
  
• a forceout
  
• a pop out
  
• a double play
  

The STATS data doesn’t give full details on all of these; for example, it doesn’t say how many times a player may have bunted for a hit in a possible sacrifice situation. We must assume that this did happen, but those occurrences are buried in the general category “singles,” and we don’t know how many of them there might have been. This is an important omission, and there are other things which are also unclear. There is a category entitled “Bunt Pop Out,” but it isn’t clear whether this includes only sac bunt attempts, or perhaps also attempts to bunt for a single.

Anyway, this is the best data we have, so let’s work with it. We can estimate, based on the STATS data, that the frequencies of these events are:

About 19% of bunt attempts fail, with about 10% of those (2% of the total) being double plays. The 5% figure for base hits is just a guess.

I have a computer program which simulates baseball games, sort of like a perpetual APBA game. Suppose that we take a team, a kind of average team, and we run them through a hundred simulated seasons in which they don’t bunt.

Then we build the information on the chart above into the simulation program, and we put the bunts back in, and we run them through 100 seasons with sacrifice bunts. Will they be better with the sacrifice bunts, or without?

The realistic bunt information isn’t all that easy to put into practice, because you can’t put things in on a random basis. I worked it out as best I could and ran the simulations. The studies don’t show what I thought they might show. I thought the teams might perform better with the bunts than they did without. If they had, then I could argue that the sabermetricians who have studied this before were just wrong, and that the sacrifice bunt really
is
a good play if you use it well.

What I have instead is a mixed picture. I ran three teams through 100 seasons each with and without sacrifice bunts, a total of 600 simulated seasons. All three teams did better without the sacrifice bunt than with the bunts.

But looking carefully at the details of the study, it still appears to me that a case can be made on behalf of the bunt. While all three teams did better overall without the bunt, two of the three actually did better
in home games
when the bunt was in. Team C, for example, had a .521 winning percentage without the bunt, which was the same at home and on the road. When I put the bunts in, their winning percentage went to .515 on the road, but .524 in home games.

Suppose, then, that I just shut off the bunt in road games; what would we have then? Then we’d have .524 at home, and .521 on the road—a better overall record with the bunt (in half the games) than with no bunts at all.

And if you think about it, one of the oldest saws of managing is what? Play for a tie at home; play to win on the road. Never play for a tie on the road.

Isn’t that really almost the same thing? What does the expression “never play for a tie on the road” really mean, in practice? It means “if you’re behind and you’re on the road, don’t bunt.”

So it may be that what is indicated by the study is not that the bunt is a poor percentage play, but that I
didn’t do a good enough job of instructing the computer when to bunt
.

Well, that’s easy to understand. I had the computer bunting about fifty times a season, and I put the bunts in on a kind of a commonsense basis—don’t bunt with the three/four hitters except in unusual circumstances, bunt mostly with your best bunters, don’t bunt with the middle of the order coming up in the early innings, don’t bunt when you’re two runs behind, etc.

The computer produces bunts in
reasonable
circumstances, but there’s also a random component to the decision. It is extremely difficult to teach a computer to think like a major league manager, as anyone who has ever tried it can verify. It seems likely to me that if the computer had done a better job of picking when to bunt, it would probably have posted a better winning percentage with the bunt than without it.

There are other details in the study which suggest the same conclusion. The program keeps track of the number of runs scored per inning, sorted by who the leadoff hitter was. While the number of runs scored dropped overall, the runs scored in the innings started by the seven, eight, and nine hitters did increase in all three studies. Again, this suggests that if the computer had done a better job of picking when to bunt, the results very well might have been different.

So what am I really saying here, other than that I don’t know how to tell a computer when to bunt? Am I saying that the 1940s managers, who bunted 100 times a season, were making the right decision?

Certainly not.

Am I saying that Palmer and Thorn were wrong when they said flatly that the bunt was a bad play?

No, I can’t say that they were wrong.

What I’m saying is that I think it’s an open question.

One way to phrase the question would be to ask for a number. How many times are there, during a season, when a team
should
bunt?

Billy Southworth thought that the answer was somewhere near 150.

Gene Mauch thought that it was over a hundred.

Earl Weaver thought it was more like 30.

Palmer and Thorn’s work would suggest that it was near zero.

What I’m saying is that I simply do not know. The answer, dear class, is rolling in the grass. I don’t think the right number is zero, and I doubt that it’s near zero, but I don’t know what it is. Having thought about the issue at great length, having worked hard to analyze the math involved, I can only tell you that there is no definitively correct mathematical answer at this time. Earl Weaver may have been right; Billy Southworth may have been right. Maybe each of them had the right answer for his own team. The rest of us need to keep an open mind.

In ’47, full of woe

The Pittsburgh Pirates finished low

But in ’48, under Billy Meyer,

They’ll very likely finish higher.

—Max Salva,
Baseball
magazine,
March 1948

The Pirates did indeed finish higher under Billy Meyer—in fact, more than 20 games higher. Billy Meyer, after almost forty years in professional baseball, was an overnight sensation, a genius for a day.
The Sporting News
named him the Major League Manager of the Year.

A small, singles-hitting catcher, Meyer played in one game with the White Sox in 1914, then spent two years as a third catcher for Connie Mack. That failing, he was sold to Louisville of the American Association, where he would spend eleven seasons.

Joe McCarthy became manager of the Louisville Colonels in 1919 and set about building a powerhouse. The Colonels won the minor league World Series in 1921. In 1925, after another 102-win season, McCarthy was hired to manage the Cubs. Billy Meyer inherited the Louisville manager’s job, probably at McCarthy’s recommendation.

For one season they carried on as before, winning 105 games in their first season under Meyer. In 1927 the team fell onto hard times, losing 103 games in 1927, 106 in 1928. Meyer was fired, and at the same time released as a player. He was thirty-six years old, and had played for nineteen seasons.

Meyer returned to his home in Knoxville, Tennessee. There had always been another game tomorrow, another contract next year; now he was thrown back on his wits. Out of baseball for three years, he looked back on his years as the Louisville manager and realized that he had done nothing to discipline himself, let alone his players. He had made good money and wasted it. The depression hit, and money was suddenly scarce.

Meyer looked to get back in baseball. He got married in the spring of 1932 and got a job as a minor league manager, at Springfield in the Eastern League. It wasn’t the American Association, but he was back in the game. He had the Springfield team in first place with a record of 53–26 when, on July 17, the league disbanded.

The depression had hit baseball, too.

Meyer was hired to manage Binghamton in the New York-Pennsylvania League, another step down.

At the pace of modern justice, he began to work his way back toward the majors. He took over a seventh-place team in Binghamton, and dragged them to a 34–28 record the rest of the season. He won the pennant there in ’33, his third minor league pennant, counting Springfield. The league adopted a split schedule in ’34; Meyer’s teams won half the pennant each year in ’34 and ’35.

In 1936 he went to the Pacific Coast League, where he managed the Oakland Oaks. After two years of indifferent results in the PCL, he ran into George Weiss at the winter meetings. Meyer asked Weiss about a place in the Yankee system, and Weiss agreed to ask Joe McCarthy about him. McCarthy gave Meyer an enthusiastic recommendation, and Meyer was hired to manage the Kansas City Blues in 1938.

The highest minor league classification at this time was Double A. The Yankees had two Double A teams, one at Newark and one in Kansas City. Both teams were extremely strong. Meyer managed those two teams for the next ten seasons—Kansas City for four years, Newark for four years, then Kansas City for two more. In the ten years his teams won four pennants and finished second four times, posting only one losing record. His Kansas City teams of 1939–1940 are among the best minor league teams of all time, finishing 107–47 in 1939, and he played a role in the development of many major league stars. Like McCarthy, he was a disciplinarian without being a screamer, a student of the game with a perpetually active mind.

In the 1990s a manager with such a record might get a shot at a major league job within a couple of years. In the mid-1940s there were more minor league teams, fewer major league teams, and the major league teams strongly preferred star players. Meyer was fifty-six years old when Pittsburgh finally hired him for the 1948 season.

The Pirates, a strong franchise until 1945, had slipped badly in 1946–1947, losing over 90 games each year. Ralph Kiner came to the team in 1946, and, as a rookie, led the league in home runs, with 23. Hank Greenberg the same year led the American League in home runs and RBI, with 44 and 127. Greenberg talked seriously about retiring, which forced the Tigers to make him available. The Pirates put together a big-money package and acquired Greenberg to bat behind Kiner. They had visions of first place.

They finished last. Billy Herman, taking over the team in 1947, had recommended that the team trade Kiner, saying that they had Greenberg to bat cleanup now, and he didn’t believe Kiner was really that good. Kiner hit .313 with 51 homers. Maybe Herman was right. As Branch Rickey told Kiner in negotiations the following spring, they could have finished last without him.

Anyway, Greenberg retired, and Meyer was hired to fix the mess. The flaws of the 1947 team are in retrospect obvious. They had no left-handed hitting, no speed, and a group of old pitchers who couldn’t throw strikes.

Meyer put into the lineup several players who had been hanging around the league for years, not playing much. Stan Rojek, Ed Stevens, Danny Murtaugh; they weren’t any better than the guys they had replaced, but they balanced the lineup left/right, and they were hungry. The team stolen base total more than doubled. Bob Chesnes, purchased from the Pacific Coast League, became the Pirates’ best pitcher of 1948, at 14–6. The Pirates finished 83–71, just 8½ games out of first place.

Billy Meyer was a genius for the day. The next year, the Pirates fell into their old ways. They loaded up their roster with a bunch of guys who had been good before the war—Ernie Bonham, Kirby Higbe, Walt Judnich, Hugh Casey, Dixie Walker. Murtaugh and Rojek, a brilliant double-play combination in ’48, fell off badly. Chesnes slipped to 7–13. A trade with Brooklyn, a short-term winner in 1948, turned into a long-range disaster. The Pirates slipped to 71–83 in 1949.

It got worse. They went 57–96 in 1950, lost 90 more in 1951. By 1952 the Pirates had become one of the worst teams of all time, losing 112 games, against only 42 wins. They finished 54½ games out of first place.

How did the Pirates sink so low? What happened to Billy Meyer, after he managed so well in 1948? Why was he unable to solve any of the Pirates’ escalating problems in the following seasons?

There was a story in
Sport
magazine, ten years later, that the Pirates team rebelled against Meyer after he mishandled a question at a postgame press briefing in midseason, 1949. He had given a player a hit-and-run sign, and the player ran into a pitchout. Asked about it after the game, he allowed the press to draw the conclusion that the player had run on his own. A writer wrote something about the incident, and the player felt betrayed. The team lost respect for Meyer, and he was never able to regain their confidence.

Well, that’s just a story. Meyer’s first team, the Louisville team of 1926, played well for one season, then collapsed. His Pirates did the same. The 1952 Pirates, at the end of Meyer’s leash, were unnecessarily bad because Branch Rickey, who had been hired to rebuild the Pirates, wanted to start by taking a long look at some young players who were obviously not ready to play in the major leagues. Meyer was relieved of his command at season’s end—the first time he had been fired since 1928. He would be dead within a few years.