Sunday, January 15, 2017


Firemen and Closers

The 70’s weren’t a great time in America in many ways, but they were an amazing time for relief pitchers.  This was the age of Goose Gossage, Bruce Sutter, Al “The Mad Hungarian” Hraboski, and the two greatest moustaches in baseball, Sparky Lyle and Rollie Fingers.  They were known as “firemen” back then, because their job was to put out fires.  When the game was on the line, and it looked like the team was about to blow a lead, the fireman would come in and shut it down.

That period didn’t last, though, and it’s not because starters went back to throwing complete games.  They didn’t; in fact, complete games are rarer now than they were then.  It’s because of statistics and specialization.  Some statistician invented a measure called the “save,” which is awarded to a relief pitcher who throws the last three innings or fewer (usually fewer) and preserves a lead of three runs or less.  Over time, the “save” went from descriptive to prescriptive.  Now, most teams have a relief pitcher they call the “closer,” who is only used in “save” situations.  Game on the line in the fifth inning?  Not a “save” situation -- call in someone worse.  One run down, and the other team is about to put it out of reach?  Not a “save” situation -- call in someone worse.  And so on.  You don’t “waste” your closer on a non-save situation, even if the game is on the line.  Closers close.

Reading The Only Rule Is It Has to Work, by Ben Lindbergh and Sam Miller, made a great counterpoint to Michael Lewis’ latest, The Undoing Project.  Lewis’ book is a profile of Amos Tversky and Daniel Kahnemann, who collaboratively started the field that eventually became behavioral economics (and, as Lewis rightly notes, formed the unacknowledged theoretical underpinnings of his own Moneyball).  Lewis’ book is an accessible and even touching story of the collaboration the two men forged, and a strong introduction to their key contributions.

As Tversky once put it, rather than studying artificial intelligence, they studied natural stupidity.  They looked at the cognitive equivalent of optical illusions -- non-random, predictable errors that people make, even when they should know better.  

Behavioral economics made its way into baseball via Bill James and sabermetrics, and acquired popularity through Moneyball.  Theo Epstein used sabermetrics to break a longstanding curse and deliver multiple World Series championships to Boston.  

Lindbergh and Miller’s book brings that stuff down to earth.  They’re a couple of stats guys and podcasters who, through circumstances not entirely clear, got a chance to run a minor league team for the 2015 season -- the Sonoma Stompers -- to see what would happen if they could put their theories into practice.  Which meant, among other things, that they had to convince an actual team comprised of actual players to abandon the “closer” model and go back to the “fireman” model.  The book is a co-written, first-person account of their season.

I groaned in recognition repeatedly.  Though they were the “general managers,” they weren’t the actual manager of the team.  The manager of the team was a veteran player who doubled as a manager, and who had his own sense of how things should be.  Lindbergh and Miller clashed with the player-manager repeatedly, but never more heatedly than over the role of the “closer.”

Lindbergh and Miller had what they considered an airtight mathematical case for the “fireman” over the “closer.”  They even had a specific pitcher in mind, Sean Conroy.  (Conroy achieved fame that year as the first openly gay active player in professional baseball; some of the funniest moments of the book are the descriptions of other players’ well-meaning attempts to be supportive.)  But the player-manager, Feh, wasn’t having it.

When the team was on a hot streak, Feh argued against the need for change.

When the team was on a cold streak, Feh argued that change would look like weakness.

When the team was just humming along, Feh argued...well, he said it best:

“Guys, listen: I’m not going to fucking take the closer out.  Every game this week has shown me that you need a fucking closer.  We lost one and then they just lost one and I feel like those games are supposed to make the fucking point.  How demoralizing is it for the fucking starter to give it up because he doesn’t have somebody shut the fucking door.  You guys don’t fucking -- I can’t have this talk anymore.”  He walks away, yelling at the sky.  “This is Baseball 101.  This is just Baseball 101 because you haven’t fucking played it.”  (p. 193)

It’s all there -- unrepresentative small sample, high emotion, pathos, and ad hominem.  Everything except an actual reason.

Shortly thereafter, they finally fired Feh and hired a manager who was willing to implement their system.  That particular change went well, though the economics of low-minor-league baseball inflicted a certain entropy on their roster as the season progressed.  

On paper, the superiority of the “fireman” to the “closer” is clear.  But getting veteran players to see beyond what they’ve always done requires a lot more than math.  The struggles are the same, whether the numbers are earned run averages or pass rates.  For academics who aren’t put off by baseball, I really can’t recommend Lindbergh and Miller’s book highly enough.  

As an academic who loves baseball, I need to put it on my list of books to read when I have time, probably after I retire from working in the academic trenches.

I must compliment you on burying the lede in the penultimate sentence.

Why not "learning rates"? That is what a baseball W-L record measures.

Which pass rate, the one for a gateway course or the one for the next course in the sequence? I've seen my share of cases where the pass rate in 101 goes up but the pass rate in 102 goes down. Worse, I've seen where the pressures to pass more students keeps 102 the same so that 103 becomes a total disaster. Pass tates are OK only as long as you have a fixed assessment and the assessment measures what matters for the next course in a pre-requisite chain.

That said, one of the challenges when reading the edu-lit is making sure some innovation works at scale and not just in the hands of a gifted teacher who pioneered it.

If you really want to assess a college, watch its transfer and transient students. Do your transfers pass all their classes in their first post-transfer semester? Do transients from a nearby 4-year school do better or worse in classes at your college? Those are what interest me.
FWIW, the "save" was not created by a statistician, but by a journalist, Jerome Holtzman of the Chicago Tribune. Had it been created by a statistician, it would almost certainly be defined very differently.
The "save" was not invented by a statistician, but a sportswriter.
Post a Comment

<< Home

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