THE BOOK--Playing The Percentages In Baseball

Thursday, April 15, 2010

Morgan Ensberg steals home

Ensberg recounts this tale:

With 2 strikes on Wes Rachels, (our second baseman and eventual College World Series MVP) I took my lead at third base and got close to the baseline trying to give myself the smallest distance between myself and home plate.  ASU’s right handed pitcher was in the wind up and I was looking at his left foot.  I knew his left foot would be the first movement he would make, alerting me to go.  He stepped back, and I took off!

“Go…go….I’m going to make it!  Slide! Slide!”

I watched my left foot cross the plate as the catcher tagged my left knee.

“Safe!!!!  SAFE!!!” The umpire yelled.

I immediately popped up to the standing position and gave a huge fist pump and yelled as my teammates came and almost tackled me.

Very cool I thought.  Stealing home has a low breakeven point with 2 outs, which means that if you think you have a decent chance of making it, go for it.  Then my jaw dropped:

Wes Rachels lined the next pitch down the third base line for a 2 run RBI. Gillespie’s brain put us in that position and it was the right call.  We won that game 21-14 and clinched the 1998 National Championship.  All on a decision based on taking advantage of an opportunity with an average runner.

21 to 14!!  Wow, that is the complete opposite of small ball.  When the run environment is that high, you don’t want to play small ball. 

One of my proudest contributions to the community is the Markov calculator.  It is perfect, as in perfect.  (With the constraint that you can’t make any outs on base.) Baseball is a very simple thing to model, programmatically: runners move base to base, until 3 outs are made.  And it’s made much easier if you think recursively.

What the calculator does is that it gives you a baseline, a starting point, to answer any run- or win-related question in baseball.  If you’ve never used it, then you are missing a basic tool in your toolset.  It’s a hammer or screwdriver.  It’s that valuable.  If you don’t think so, you are wrong.  Set aside your opinion, because that’s the only way it’s going to change.

If you simply hit CALCULATE, you will see that’s the default run environment (4.9 runs per game).  It also gives you an RE chart.  With bases loaded and two outs (presumably there were two outs, but Ensberg didn’t say), the RE is .819.  If he’s out, that’s the cost: minus .819 runs.  If he steals, and presumably the other two runners remain on base, that’s a run plus .478 of run expectancy for the runners on base and future batters, for 1.478 runs.  That’s a gain of .659 runs.  So, you can see how you need a bit more than a 50/50 shot for breakeven here.  When you are right, you gain .66 runs.  When you are wrong, you lose .82 runs.  So, you need to be right 55% of the time to break-even.  But that’s in a 4.9 runs per game environment.

When a game ends up 21-14, I will presume that the run environment in college is much higher.  Let’s say that the expected run environment is 10 runs per game for each team.  Go back to the Markov calculator, and change AB to 27.  That makes it a 10.4 runs per game environment.  The RE chart now tells us that bases loaded 2 outs is 1.233 runs.  So, if Ensberg is out, he’s costing his team 1.23 runs.  If he is safe, that’s a run plus .782 of run potential left, or 1.782 runs.  So, he gains +.55 runs if he is safe.  The breakeven point is 69%!

The reason that the breakeven point is so high is that he already had a good chance of scoring, as did the other runners on base.  Getting thrown out not only kills him, but decimates the runners already on base.

Was it a good play or bad play?  Well, it all depends on how successful Ensberg thought he could be.  If he thought he’d be safe at least 70% of the time, then it was a good play.  Did he think that?  Let’s ask him.