Saturday, October 18, 2003

Pattern Recognition: Manager's Friend
& Manager's Hiroshima  

An important aspect of managing both processes and people is internalizing and analyzing the historical performance of them in the contexts they played out in to try to better judge what to do now and in the future. A little under half of American managers do this, and they generally perform significantly better than the ones who don't. It takes talent, a decent memory, systemic thinking ability and persistence. It's work. It's essential to success in any competitive system.

The shorthand term I use for that work is "pattern recognition".

On the other hand, Grady Little met his Hiroshima in the deciding game of the A.L. Championship series because of pattern recognition. He took a chance on leaving his gutsy starter in when it was obvious to him and everyone else watching that his tank was empty & he was down to fumes. To most observers, including me, it looked like one of the most ill-timed brain cramps in the history of playoff baseball. How could that have happened?

Thanks to my buddy Steve Manes, we have a more rational explanation. He solved this riddle for us. Little was applying pattern recognition in the truest sense. Usually when we apply pattern recognition, we are using a partial analogy, or cobbling together a pile of parallel or somewhat-alike past events. Sometimes we luck out and have an exact or almost-exact duplicate of the situation. The more identical the past situation is, the more faith we put in our recognition, even though random factors can cancel it out, especially with people. Sometimes the identical factors we recognize don't guarantee the same outcome or even probability of outcome.

Duck & Cover

Grady Little was cursed by his good memory, because on the 58th anniversary of the nuking of Hiroshima, August 6, 2003, Pedro Martí­nez pitched a regular season game against the Angels that unfolded almost exactly like Thursday night's. The starter was breezing with six strikeouts and yielding five hits through five innings, and he was staked to a three-run lead. Parallel to the first Thursday homer to Giambi in the 5th, Pedro got rocked for a single run in the 6th of the Hiroshima Day on a pair of hard-hit doubles. In the Angel game, he came back and got the next two outs without any trouble, and in the Thursday game, he mowed down the next two Yanks with Ks.

Angel game: Hit hard for a run but quick recovery.
Yankee game: Hit hard for a run but quick recovery.

7th Inning
Angel game sequence:

-A Kennedy flied out to deep left center.
-B Molina singled to shallow right.
-R Quinlan struck out looking.
-D Eckstein doubled to deep left, B Molina to third.
-D Erstad grounded out to pitcher.

Yankee game sequence:

-H Matsui grounded out to second.
-J Posada lined out to center.
-J Giambi homered to right center.
-E Wilson reached on infield single to first.
-K Garcia singled to right, E Wilson to second.
-A Soriano struck out swinging.

The difference was Giambi's second home run, on the one hand indicating Pedro's additional vulnerability, but on the other hand, there was Enrique Wilson's "infield single" which wasn't an actual hit but a bad hop combined with an error on the first baseman (errors are politically hard to call for official scorers -- a great topic for another entry). But the parallels are striking. Pedro was stirred, not shaken, though he was clearly some pitches beyond his best command. And against the Angels, Little left him in for the 8th inning:

8th inning
Angel game sequence:

-T Salmon struck out swinging.
-G Anderson struck out looking.
-S Spiezio struck out swinging.

9th inning

-J Davanon grounded out to pitcher.
-A Kennedy safe at first on error by first baseman D McCarty.
-A Kennedy to second on fielder's indifference.
-B Molina popped out to shortstop.
-R Quinlan singled to center, A Kennedy scored.
-D Eckstein doubled to deep left, R Quinlan to third.
-D Erstad hit by pitch.
-T Salmon struck out looking.

It appears to both Steve and to me that Little remembered that Angel game sequence. Even without his best stuff, Pedro with a 4-1 lead whiffed the side in the 8th, faced adversity (the error and then the pilfered base without an attempt to stop it), gave up some hits, lost much of his control (the hit batsman), was way over his normal, post-injury pitch count limit (about 120 at that point), and still protected the lead with a dramatic strikeout. In brief, Pedro was gutty and even without his best stuff and way over a logical pitch count was able to protect a small lead. In the Angel game.

So when Little walked to the mound and asked Martí­nez if he could protect the small lead without his full control and now over his normal post-injury pitch limit and Martí­nez, predictably, said 'yes', Little went with a general understanding that Martí­nez is the best pitcher in baseball, his hopes (Management by Wishful Thinking) backed with what appears to be a single strong historical precedent (all the factors just cited) for this particular situation. And in the Yankee game, it just didn't work. The Bosox got Hiroshima-ed.

Little's pattern recognition didn't work out this time. Was he "wrong"? Tough call, sort of a lose-lose. If Pedro had had a little more good luck or the Yankees a little less good luck, the outcome could easily have reversed and Little's pattern recognition skills would have been applauded by many as a gutty move. Like the intelligence agencies' inaction before 911, it's really straightforward in retrospect to look at the data that was available and connect the dots, but it's much harder to do in the here-and-now. In retrospect, we can make strong arguments about how the situations were different: The Yanks are a better-hitting team than the Angels, Pedro was pitching on six days rest against the Angels, etc. But there are always differences in the baseline situations you use in pattern recognition, always differences people can use to hammer you with when the good (or not) decision you make doesn't work out

But either way, this particular Little Hiroshima is a fine cautionary example of one of the most dangerous tactical errors in the application of the pattern recognition skill: the closer to being identical the current situation is to one you've already experienced, the higher the confidence in simply duplicating the solution. With machines (at least those not running under a Windows operating system), this tends to be effective. With people and small samples, the illusion of identicality can overwhelm other, independent factors.

TIP: Use pattern recognition. Experiment with your people and your processes. But don't be fooled by a situation that appears "identical" but may not be.

This page is powered by Blogger. Isn't yours?

free website counter