THE BOOK--Playing The Percentages In Baseball

Wednesday, May 18, 2011

Derek Jeter: History Told You So

There are two things I learned from MGL early-on in my internet postings:
1. Regression
2. Put your money where your mouth is

Before reading MGL on BaseballBoards.com, I never knew about regression and I always figured that the observed stats really represented the player.  Regression is all about inferring something true, based on observed samples.  You observe this, you observe that, and then you have to ask: “what does this tell me about the entity involved?”.  You may observe Joe Mauer hitting alot of home runs in half a season, but what does it really tell you about Joe Mauer?  What does observing Chipper Jones hitting .420 really tell you about Chipper Jones?  And so on.  I think we all kind of dance around the idea at first that there’s some luck, but to actually have a process to figure out how much luck was pretty exciting for me.

The other is that if you get to the point that your forecast requires you to bet on it, all of a sudden, you aren’t so sure of yourself.  Now you start to think about bad luck, you start to be risk aware.

So last autumn, I forecasted Derek Jeter.  What I did first was rely on historical comparables.  I didn’t just invent a formula.  I looked at what’s actually happened with actual players.  It was a study really of human beings, not some theoretical exercise.  And, in terms of betting on a forecast, I gave a pretty wide-range:

Then we have the final eight remaining players—more than 40 percent of the entire group—who had no value left.  Overall, 80 percent of these 19 players have a value of nothing to $25 million, and 20 percent have a value around $75 million. That’s an average of $25 million, which—removing what he means to the Yankees from the equation, which is intangible and can’t be measured—seems to be what his next deal should be.

Jeter still had a 20% chance entering 2011 of blowing out most people’s forecasts of him.  But, there was also a 40% chance that he’d be useless.  This was a high stakes gamble.

But when you see the “analysis”, you see alot of cherry picking.  In the off-season, you had alot of people arguing that he was really on that 75MM$ value side.  And they staked their position on that.  And now, you have alot of people arguing that he’s almost worthless.  And, they cherry pick the data, seeing “trends” starting in 2009 (a “trend” that disappears if you start in 2008 instead).

You have to look at the idea, and infer what it really means.  And you have to be able to bet on it.  That means regression, and that means uncertainty levels.

And given we now have more data of him, instead of our pre-season forecast of him having a 40% chance of being useless, and 20% chance of being a superstar, maybe we change that to 50% / 10%.  Instead of being worth 25MM$ over 3 years, maybe he’s now worth 15MM$ over the remaining 2+ years.  (This paragraphs uses numbers for illustration purposes only.)