Bill James Guide to Baseball Managers, The (21 page)

Our problem now is that we have Leo Durocher ranked even with Joe Altobelli. Al Lopez, a wonderful manager in anyone’s book, is now ranked even with you and me, since both of his pennant-winning teams lost the World Series. This extends the principle above to absurd proportions: everyone is a failure except the one guy standing at the end of October.

We’re on the right track now, but we need a system which recognizes levels of success. What are the things that a manager can do which would cause the rest of us to describe him as successful?

Well, if a manager has a winning season, that’s a mark of success—the lowest mark, but a starting point. Nothing below that can be considered an accomplishment. Winning a division title is beyond that, winning the league championship beyond that. I established a system which recognized six types of accomplishment for a season:

The first four of those are the four basic levels, and are scored as 1, 2, 3, and 4. If you win the World Series, that’s a four-point season. If you win the league but lose the World Series, that’s a three-point season. If you win your division but not the league, that’s a two-point season. If you finish over .500 but don’t win the division, that’s a one-point season.

To these four levels, we add one point if the team finishes 20 games over .500, and a second point if they win 100 games. Thus, a perfect season—a season in which the team wins the World Championship
and
wins 100 games—is worth six points. Winning the World Championship but with less than 100 wins is worth five points. Winning the World Championship but with a record less than 20 games over .500 is worth four points.

Or, on the other hand, winning 100 games but not even winning your division, like the San Francisco Giants did in 1993, that’s worth three points—one for having a winning record, plus one for being 20 games over .500, plus one for winning 100 games.

How good a manager someone is is a largely subjective question. How much success a manager has enjoyed is a relatively objective question. Is Sparky Anderson a good manager, or a lucky stiff who came along in the right organization at the right time? Everybody’s got an opinion, and there’s no way to prove one or the other. But either way, Sparky Anderson enjoyed a considerable amount of success as the manager of the Cincinnati Reds. All I’m trying to do here is to fix that objective observation in a number.

This system is the basis of the “most successful manager” notes which occur in each Decade Snapshot—the most successful manager of the 1870s was Harry Wright, etc. I buried the article about it here, in the middle of the book, in the vain hope that this would discourage reviewers from thinking that that was what the book was all about.

Anyway, let me defend my system and apologize for it. What I
can’t
tell you is that there is any compelling logic to this system, or that this is any reason why this system is right, and another system is wrong. Why should a team get one point for finishing 19 games over .500, and two points for 20? Why is a “perfect season” six points, rather than seven? It’s arbitrary.

I would argue, however, that

a) the arbitrary cutoffs are reasonable, and

b) any similar system, designed to measure the successes of major league managers, would yield almost identical rankings for the top 15 or 20 managers, regardless of precisely where cutoffs were established, or how different accomplishments were weighted.

So then, the 52 most successful managers of all time, through 1996, are:

The modern manager, in this system, has both advantages and disadvantages compared to earlier eras. The modern manager can earn points for winning his division, which were not available to managers pre-1969—but on the other hand, the number of teams competing for those “championship points” is greater now than it was then. The season is longer now, giving teams an advantage in the race for 100 wins—but there is more competitive balance than there was through much of baseball history, so that in fact there are
not
more teams winning 100 games a season, but fewer. My intention, in laying out the system, was to be fair to managers of all eras.

Anyway, of the top ten men on the manager’s list, all are in the Hall of Fame except Sparky Anderson, who I would assume will go in as soon as he’s eligible. Of the second ten, five are in, and five are not, but two of those who are in (Fred Clarke and Cap Anson) may have been selected largely as players. After the twentieth spot, all of the managers who are in the Hall of Fame were also outstanding players, with the arguable exception of Wilbert Robinson (and even Robby may have been selected as much for his playing contributions as for his managing).

So that answers, in one form, the question of what a manager has to do to get into the Hall of Fame: He needs about thirty points in this system. If Tony LaRussa has one more good year at St. Louis, he’ll probably be in the Hall of Fame. As it is, he’s on the bubble. Bobby Cox and Davey Johnson are in a similar position: They’ve accomplished a lot, but they’re not certain Hall of Famers at this point. In five years, they might be.

Another way that managers could be ranked is by comparing their won-lost records to their team’s expected wins, based on runs scored and runs allowed. I did that for contemporary managers in one of the
Baseball Abstracts
, and a gentleman named Bob Boynton did it on a wider basis in a recent SABR publication. The theory is that if the team scores 650 runs and allows 650 runs, but finishes 5 games over .500, the manager must have done a good job, whereas if they finish 72–90 he couldn’t have been said to have gotten the most wins out of what he had to work with. If they win the close games, they’ll come out better than expected.

Boynton studied this issue for 39 long-term and 65 short-term managers, comparing each season’s won-lost record to the expected won-lost record based on runs scored and runs allowed. He summarizes this relationship in a “D score”; if the manager done good, his D score is high.

His conclusion:

This conclusion is completely false; McCarthy’s teams did
not
“los(e) more relatively close games than they won.” The conclusion results entirely from a flaw in Boynton’s method.

Boynton based his win expectations on runs scored per game as opposed to opposition runs per game. If a team scored 1.000 runs per game more than their opponents, he expected them to post a winning percentage of .600. If a team outscored their opponents by 2.500 runs per game, he would expect them to post a .750 winning pecentage.

Within normal ranges of runs scored and runs allowed, this works well enough. But as the run margin increases, the value of each run decreases. What would happen if a team was outscored by, let us say, six runs a game? By Boynton’s method, they would have an expected winning percentage
less than zero
. If they outscored their opponents by five runs per game, he would expect them to win every game. If they scored more runs, winning every game wouldn’t be enough.

As a team begins to score more and more runs, it becomes impossible for them to win as many games as Boynton expects them to win. The 1939 Yankees, for example, outscored their opponents by 2.72 runs per game; thus, Boynton expects them to win 77.2% of their games. When they actually won “only” 70.2% (they finished 106–45), he gives them a D score of minus 70 (702 minus 772 equals minus 70). He believes they should have finished 117–34. “Although the 1939 season is an extreme case, McCarthy’s (record) indicates that the tendency to lose relatively close games was a consistent one throughout his career.”

In fact, though, the 1939 Yankees had a
good
record in close games—22–15 in one-run games, and 50–32 in games decided by three or less. The ’39 Yankees won 15 games by ten or more runs, a phenomenal accomplishment in itself, including three wins by twenty or more runs. Obviously, they weren’t going to
lose
games by twenty or more runs. They were 24–0 in games decided by more than six runs. These are normal expectations for such a team. If you studied minor league teams which scored seven runs a game, you’d find the same types of performance for them.

This problem doesn’t have much effect on anyone except Joe McCarthy, because no one else’s teams scored so many runs. But there is another problem. Boynton’s method predicts a .600 winning percentage if a team outscores their opposition by one run per game—ten to nine, or three to two. Obviously, this is not accurate. The expected winning percentage of a team which outscored their opponents ten to nine would be .552; the expected winning percentage of a team outscoring their opponents three to two would be .692. Walter Alston comes out looking great in Boynton’s system, because, in a run-starved environment, his run advantage means much more than projected.

I could have repeated the study with a better formula, I suppose, but I decided to pass. The theory, while intriguing, is just too speculative. It measures the manager
against
the individual accomplishments of his players, when in the real world, a manager has to work
through
the talent he’s got. Let’s take something really stupid that a manager might have done. Jack McKeon in 1974 with the Kansas City Royals

a) didn’t think George Brett was ready to play in the majors, and wanted to keep playing Paul Schaal, and

b) ruined Steve Busby’s arm by having him throw about 200 pitches in a game one time when his catcher was telling him Busby wasn’t right.

The mistake on Busby, the best young pitcher in baseball at that time, cost the Royals dozens of wins over a period of years, and the mistake on Brett (and Frank White—McKeon didn’t think he could play, either) could have cost the Royals countless runs. But how would either of these things have altered the ratio between runs scored, runs allowed, and wins?

If a manager does a good job of keeping his team focused on winning, how will that cause them to have a better ratio of wins to runs scored/runs allowed? It’s hard to see. If he tears the team apart with petty battles over irrelevant rules, how will this be reflected in the wins versus expected wins? I can’t see it. The theory of it is just too much of a reach.

There is another way of establishing “win expectation,” however. Another way to establish expected wins is to look at the performance of this team in previous seasons and establish from that how many games we would expect the team to win this year.

The best way I have found to establish expected wins for a team based on previous seasons is to combine four elements, which are the records of the team in the previous three seasons, and a .500 record. The record of the team in the last season is 50% of the load. The .500 record is 25%, since all teams have a pronounced tendency to drift toward .500. The records of the team the two previous seasons are each one-eighth, 12.5%.

For illustration, suppose that a team won 100 games last year, 90 games the year before that, and 70 games the year before that. How many games would we expect them to win next year?

We would expect the team, given that history, to win 90 games this year.

This system establishes expected wins very accurately for groups of teams. If you take a group of teams which would be expected to win 90 games by this method, they will in fact win 90 games, on the average. If you take a group of teams which would be expected to win 70 games, they will win 70. If you can find a group of teams which would be expected to win 100, they will actually win 100, on the average. The method works in 1910, and it works in 1995.

So then, we can use this to evaluate the performance of the manager, by comparing how his teams
actually
performed to how they would have been expected to perform, based on previous seasons. Let’s take the Texas Rangers in 1974, Billy Martin’s first year there. The team had gone 63–96 in 1971 (in Washington), 54–100 in 1972, and 57–105 in 1973. Given that history, and given 160 games in 1974, we would expect them to finish 63–97. Figured as follows: