THE BOOK--Playing The Percentages In Baseball

----"A couple of things. I’m treating the offense as if it was equal to the defense. We don’t know that it was.”

Isn’t that a rather large problem?

Just saw Patriot’s article. He seems to want to rescue pitcher W/L records because of the advantages of every league and park being at .500, etc. But I don’t think it’s a good trade-off. I think it’s better to simply look at the pitcher’s isolated performance, and make appropriate adjustments for lg avg and park.

But I’m sure Patriot is well aware of this, so I will keep an open mind as to wher he might be going.

I should probably clarify it then, because that’s not what I was going for.  I was just giving the good thing about pitcher wins and losses--as Tango did above, I figured there was no need to beat a dead horse about the myriad of problems with win-loss records.

I do think that, particularly at a career level, W-L record can be a decent stat (especially when team is accounted for) and should be a secondary consideration in evaluating a pitcher (with the primary consideration being RA and ERC and DIPS and all of that good stuff). 

And I’m not going anywhere special with this--Tango is doing the same kind of thing I will be doing, although my approach (actually Rob Wood’s) is more complicated, although probably needlessly so, since the results will be similar.

What is it about a pitcher’s w/l record over a career that can tell you more than his adjusted ERC, RA, ERA, OPS against, etc.?

I need to be more careful.  Maybe I should have said “extremely secondary, only interesting in extreme cases” or something.

From a pure value perspective, if there was a pitcher who for whatever reason did a good/poor job of “pitching to the score” or had an unusual distribution of runs allowed between his starts that would result in a higher W%, perhaps some of this residue would show up in his W-L record (of course, there’s a lot of other junk in there as well).  And to start my policy of being more careful, I’m not implying that I am convinced that there are skills that lead to the possibilities I just mention, only that if they did occur for a particular pitcher, and if you are only interested in value, then you would want to know about.  But guys, cmon, you know I’m not John Kruk or something

MGL, why should the test be “more than”?

I can probably tell you 90% as much, using only the pitcher’s W/L record, and his team’s W/L record.  And, I can claim it’s also park adjusted, without needing to do any park adjustments.

Can’t you determine at what number of decisions the pitcher’s W/L record becomes as good a statement of ability as his ERA, etc., using the regression to the mean formulas?

90%?  There has to be a minimum number of games.  At one year, w/l record tells you next to nothing.  And yes, there is an automatic park adjustment.

But there is so much junk in a pitcher’s w/l record, frankly I am not interested in it at all, in terms of a pitcher’s talent.

Of course, in terms of past value, you can say that w/l record is everything if you want.

As far as “pitching to the score” or what have you, until you determine that there is a significant skill among pitchers, I don’t want to hear about that either, again, in terms of talent (and future value).

When we are taking about past value, we can “debate” forever the stats we might want to use (WPA, w/l, etc.).  I don’t think that is what we are discussing here.  We are taling about some kind of context-neutral value which represents true talent and can be used to estimate future value, as far as I can tell.

MGL, you said this:

What is it about a pitcher’s w/l record over a career that can tell you more than his adjusted ERC, RA, ERA, OPS against, etc.?

So, I’m only talking about career, about the 90%.

Here’s Patriot’s next installment:
http://walksaber.blogspot.com/2006/09/evaluating-pitcher-w-pt-2.html

Patriot continues:
http://walksaber.blogspot.com/2006/10/evaluating-pitcher-winning-pt-3.html

The reason to do the approach as I, and he, describes it, is if you don’t want to bother with park factors. 

If you look at the career PF of all pitchers with at least 2000 IP (which is half of a long career), you get 396 pitchers, through 2005.  96% of pitchers have a PF between 95 and 105.  Among the notables is Fergie Jenkins on one side (105) and Warren Spahn on the other (94).  I’d bet the split between team off/def will be similarly close, career wise.

Even so, Bill James had a good discussion in his original Historical Abstract about not caring how Tony Pena hits on one park, as to how it affects Jim Rice in another park.  In effect, you treat your own team as the universe, and from that standpoint, the off and def is equal.  That is, even if you have a Cy Young staff, to the offense, that’s normal.

I’m unconvinced either way, other than I like the simplicity of this method.

That is a good point that I did not consider--off/def split should tend to even out over the course of a career. 

The Wood/Tango method is very simple and tracks the more elaborate Pythagorean version very well.  If you’re going to do a comparison of win/loss record to team W%, I think it’s definitely the way to go.  I think the standard methods are in use only because it seems obvious to compare directly to the team’s W% and few people have ever sat down and thought about the kind of assumptions that you are making by doing this.  The Wood approach is almost as simple (you essentially just take the halfway point between team W% and .500) and the assumption is much closer to reality for pitchers as a whole.