August 25, 2004

Breaking Balls

Derek's Guide to Winning Baseball, an Occasional Series: Part 3

by Derek Zumsteg

Printer-friendly

I realized while screaming about a particular manager's penchant for particular by-the-book maneuvers that part of a manager's job isn't to make the right moves, it's also to make the wrong moves. Let me explain.

In games, systems for decision making only work until the other side figures out your system. In the RoShamBo Programming Competition for instance, programs play rock-paper-scissors against each other. The programs use varied approaches, statistical scoring, recent history, and all kinds of trickery. The least you should expect for your approach is a win in a third of all trials (33% win, 33% loss, 33% tie)--which is what you can get picking randomly. Systems with easy patterns are easily beat; rock every time takes only a couple of lines of code to detect and adapt to, for instance. But what if you intentionally act streaky and then when you've figured out your opponent's threshold for trying to take advantage of it, switch? What if they take advantage of your streakiness earlier? The more complicated your system gets, the more you may gain, and the more trouble you can get yourself into.

I want to add another analogy, this one from poker. If you play poker according to a system, no matter what the system is, what style it advocates, how successful the player who authored it, or how good the math behind it is, you will be beaten by better players consistently and soundly if you adhere to it rigidly. You'll walk into the Bellagio with a stake and a system, and leave blinking, jaw open, with a pit of anxiety and despair where confidence ruled hours earlier. This is because if you only raise when you hold certain cards, opponents will know you hold those cards when you raise, and know how to react. You must deviate from the system. There are optimal deviations, and the best teachers will try to show a player how to mix up their play enough that they can't be read and easily defeated. This is an important point in two ways:

You must be at least somewhat unpredictable.
You must intentionally make moves that aren't the best strict percentage plays to win.

Baseball has a book. Like the military, sports carry with them a huge amount of accumulated knowledge of how they have always been, what has worked and not worked, and all strategies and tactics grow out of this history. The world of statheads, with our databases and easy proofs of complicated concepts, is not that far down the road from the hunches of a manager who's been around the game all his life. In fact, the manager may have the edge.

Take a situation where there's a runner on first. The manager consults The Book, conveniently contained in bench coach form. Or if they're new school, he runs expectation charts, information on the pitcher, the catcher, the success rate of the runner, the opposing manager, the count, and the number of outs. The manager then decides it's a good idea to steal in this situation (1-2 count, one out, runner on first) to try and stay out of the double play.

The other manager scratches himself and calls a pickoff move, because every time he faces this other guy with the sheets of paper the dude steals on 1-2 with zero or one outs. Runner's out by five steps, and back in the dugout with his stat sheets. A manager who was only seconds ago entirely sure of himself is now angry his gamble didn't pay off.

In rock-paper-scissors, kids do this. Does my opponent favor the rock? I'll favor the paper a little, but give him enough scissors to keep him going...

In poker, players have different ways to avoid being predictable, choosing to bluff based on the suit of the card they're holding, how many chips are in the stack they're riffling, the minute of the hour, number of drinks on the table...all are seeds for a random number generator that makes them less predictable and more successful.

Take that manager with the sheets again. Even spot him a 100% success expectation on that steal, unless there's a pitchout, in which case the expectation drops to 25%. If the other manager knows he's going every time, he'll call a pitchout every time, and the actual expectation of that move drops to that 25% success rate. Which isn't enough to justify doing it. So you can't steal every time in that situation, because the more you take advantage of it, the smaller the expected gains. And yet there it is, the perfect situation to steal. It's like a forbidden fruit, only you're allowed to take a certain percentage of the apples before you start hitting the ones that tick off your betters.

By the same token, say the manager has a guy in a situation where he thinks he'll only succeed two-thirds of the time. There may be value in sending the runner. The minor harm the manager does on average every time he calls for that steal may be outweighed by increased pickoffs and pitchouts you'll get for the rest of that game.

There are other strategic concerns as well: Should you crank your base-stealing back enough to tone down your reputation for it so that when you really, really need one, they'll be surprised? Is that marginal advantage in an important situation worth not trying to take advantage of many smaller advantages you're presented with?

And similarly, say that you're the manager opposite a clueless guy who manages off the book. How often do you shear that sheep? If he's impossibly stupid, you could take advantage of him over and over, but other teams would catch on pretty fast as you ran up a tremendous record over that one idiot. Then they all beat him, he gets fired, and you've lost your advantage.

It's the ULTRA problem. Even when thousands of lives were directly at stake, the Allies had to pass up opportunities to take advantage of broken German codes because the long-term advantage of being able to use a diminished advantage that would survive longer was far greater (and even then, the temptation was so strong they nearly blew it a couple of times).

Managers are sometimes called field generals. And like real field generals, and poker players, and kids on the playground, to be truly effective they have to sometimes intentionally make the wrong decision to pursue a larger strategy. It's something to think about the next time your manager does something inexplicable, like have a guy steal third with two outs in the ninth when down by three. It could be that he's pursuing a grander strategy, the outlines of which you'll never see.