« averaging six, winning half | Main | Book Report: McSweeney's Enchanted Chamber of Astonishing Stories »

Warning: Inside Baseball Inside

So, I played a fair amount of APBA growing up. APBA, for those who don’t know, is a simulated baseball game based on statistics, similar to Strat-o-Matic or Pursue the Pennant. The players are field managers, mostly, setting lineups and controlling things such as pitcher changes, pinch-hitters, and sacrifices. In the version I played in the late seventies and early eighties (a dice, cards and boards version) the manager also made all the decisions about steals, trying to score from second on a single, throwing to second or home, and so on. Player autonomy was very limited; I imagine in the modern computer versions you can give players the green light to steal when the chance is more than n, or even take into account players who make a habit of trying to stretch hits. Anyway, the point is that I managed some dozens of games before I was bar mitzvah.

The correlation between a misspent youth simulating ball games via dice or computer and a stathead take on baseball strategy seems to be high enough to accept the plausible causality. I’ve heard it said that a GM should, before hiring a manager, insist that the candidates play a 54-game sim season with the previous year’s team, to learn whether the fellah has any idea how to manage. And, you know, I have some sympathy with that. A lot of the insights of sabermetrics are well illustrated by that kind of sim. The value of OBP, for instance, or the dangers of giving up outs, either by bunting or by attempting risky steals. The frequency of the Big Inning (in a high-scoring environment), which makes that first run less important. The order of the lineup being less important than you’d think. The goofiness of the Save. The ideas behind the Run Expentancy charts (not to mention the Win Expectancy book we’ll see someday).

The problem, though, is that although it would be nice for the manager to be aware of all this stuff, it doesn’t prepare the manager for the actual game very well. Well, and there’s another problem, which is that very few managers are hired or valued for their in-game skills, but rather for their communications skills. The manager’s primary job, it seems, is as communicator, both as the team’s voice in speaking with broadcasters and as the front office’s voice in speaking with players. And, I suppose, as the players’ voice in speaking with the front office, although I suspect that is not high on the employment criteria. Other aspects of the communicator’s job are equally important, mostly keeping players happy and productive and having the good clubhouse atmosphere so beloved of curly-haired sportswriters. Some of that stuff really does have to do with winning games, by the way; I’m convinced that Dusty Baker (to choose an instance at random) had a skill to keep players on a winning streak, that is, to parlay a few wins into a hot month by kicking up players’ expectations of themselves, and keeping the players thinking they can win any game, even when they are down by four or five runs. It’s hard to know if that is real, but it’s what I perceived when Mr. Baker led the Giants. Of course, I’m convinced that Mr. Baker is a magnificent hitting coach who performed that job throughout his tenure as manager, and that helped the team more than the streak thing. But this has become a digression.

Anyway, the tricky thing to keep in mind is that in the middle of an actual game the manager always knows more than the statistics can tell him. In the specific case of the Run Expectancy chart, the manager knows the base-outs condition, but also, for instance, whether the pitcher has already thrown a hundred pitches. Whether the batter is fast or slow, and whether the runners are fast or slow. Whether the first baseman is a three-hundred-pound, eight-hundred-slugging block of immobility. Whether the batter can beat out a bunt, or at least come close enough to maybe force a wild throw. The manager knows (or can know) the batters OBP, and thus the likelihood he will make an out (although this is clearly misleading in many ways), and also the OBP of the next batter, and the next. The manager knows, or can know, not only the general information about platooning, but the specific splits of the batter and the pitcher, as well as the guys on the bench and in the pen.

Well, and all that may be included in a really good sim, but then the manager knows more than that. He knows where the wind is blowing, and whether the infield is in shadow. He knows if the batter has been drinking. He knows if guy on deck has the flu. He knows if the pitcher seems to be having trouble with his mechanics. He knows if the batter is smarter than the pitcher, or if the pitcher is smarter than the batter. None of that is going to be in the sim, and it would be disastrous for any manager to ignore all of that specific information to adhere to what the Book or the Odds tell him to do.

Sadly, of course, much of the specific information that the manager knows in any given situation is false. He may “know” that the batter hits worse if he’s been drinking, or that he hits better, depending on the manager’s feeling about drink rather than the batter’s actual state. He may “know” that the pitcher is weakening, when in fact he has been unlucky. He may “know” that a batter is clutch, when he has only been slightly better in the clutch than his normal stink-on-ice level. Or worse, of course; the manager may be remembering a memorable clutch hit and forgetting many clutch ground outs. The sim will not help the manager distinguish between what he knows that’s so, and what ain’t so, particularly if the sim is not accompanied by a tutorial in risk analysis, or for that matter statistical assessment (to identify the lessons that really are in the sim). So. The lessons of statistical analysis can overwhelm good decisionmaking. Of course, you could argue that this is only true if the lessons are learned incorrectly. True, true. And yet, there is little chance of them being learned correctly, or at least within a correct framework.

My point? Just that it seems to me that Felipe Alou’s use of the bullpen smacks to me of a stathead manipulating APBA cards. Yes, there is often a slight statistical edge to be gained by going through three pitchers in an inning. That edge may even (possibly) justify the cost in less flexibility in later innings. But unlike APBA cards, pitchers have good days and bad days, and even if you can’t predict which day it is for each of your relievers, I have to think it’s a bad idea to go through three or four relievers every day, hunting for that one that whose dippidy-doo is dippidy-don’ting today.

chazak, chazak, v’nitchazek,
-Vardibidian. War

Post a comment

Please join in. Comments on older posts will be held for moderation. Don't be a jerk. Eat fruit.