THE BOOK--Playing The Percentages In Baseball

Friday, April 18, 2008

How to calculate Loss Shares

Bill, Where can I find how to calculate loss shares?
Asked by: Anonymous
Answered: April 17, 2008

I haven’t written up an explanation for the method yet.

Here’s one way:

Let’s use Derek Jeter as an example.  Let’s presume that there are 48% of Game Shares allocated to the offense.  With 162*3 game shares available, that gives us 233 for the offense, or about 26 per full-time player.

Jeter has played roughly 12 full seasons, so we expect to have around 312 game shares.

Bill James is reporting Jeter with 219 win shares and 86 loss shares, which added gives you 305 game shares.  Close enough for such quick rounded numbers that I’m using.  You’d really want to figure Jeter’s number of PA divided by the total number of Yanks PA and multiply that by 233.

As for defense, presuming 17% of Game Shares goes to fielders, and (I forget exactly how much SS claim to the fielding is… I’ll just use 16% for now), then a full-time SS gets 486*.17*.16= 13 game shares.  At 12 seasons, that’s 156 game shares for fielding.  Bill James is reporting Jeter with 81 game shares for fielding.

You will note that 48% + 17% = 65% game shares allocated to nonpitchers.  As you know, I believe that nonpitchers should be getting 57% of value, and therefore, you need 57% of game shares to go to them.  If we reverse engineer that, and if 48% is set in stone, then that means 9% of game shares goes to fielding.

So, redoing our calculations: 486 * .09 * .16 = 7 game shares for the SS.  Times 12 seasons gives us 84 game shares.  Fairly close to James’ 81 game shares.

What we’ll find therefore is that James is now going to give 43% of win shares to pitchers, as opposed to his original 35%.  I think James has seen the light, and is engineering his model around this 57/43 split that I’ve been advocating.  And that he will admit that he was severely penalizing pitchers (especially starting pitchers).

I just want to repeat: all of this is purely Linear Weights, and all can be explained by a Linear-based value-added model.  Truly, this may be Bill James, at least implicitly, agreeing with Pete Palmer’s original framework, and just disagreeing on Pete’s implementation.  The entire problem with Palmer’s implementation was the “single number”, a problem that James shared with his original Win Shares.  Had Palmer simply also presented “game weights”, and Bill had originally also presented “game shares”, then we would see that they were both looking at exactly the same thing.  And James’ misgivings of Linear Weights for going on twenty years was simply unfounded.

Let’s see if this is what James has done.  If so, then as Emperor Palpatine surely thought upon telling our protagonist that Padme died at the hands of Anakin: Transformation Complete.