THE BOOK--Playing The Percentages In Baseball

So Marcel said, what .344 and .436 (I did not quite understand Tango’s intro) and they actually hit .336/.431?

Basically Marcel nailed them after considering that league offense was down across the board (I think).

Not even a whiff of a “breakout”, collectively? 

I am shocked and dismayed!

These “breakout” lists by anyone are pure BS of course.  I don’t mean “of course” as in, “we know that for a fact or even with a high degree of certainty,” and I don’t mean, “For anyone who will ever compile one, no matter who they are.”

I just mean, it is my opinion and in general.

And Joe and BP should know better.  If it is Buster Olney or Jason Stark, we give them a pass.  It is lists and statements like these that give otherwise good sites (BP) and people (Joe) a bad name and take their reputations for honest, good-quality work down a notch, again, at least in my opinion.

I’ll say this.  If you have a list of “breakout” candidates or any other “projections” that are significantly different from a basic projection system (like a Marcel), you better have some evidence that what you are saying and doing have some merit.  In my book at least.

Why should I care who Joe or anyone else (me, Tango, Bill James, etc.) thinks will “break out” unless and until they show me some evidence that whatever “methodology” they are using for an unconventional forecast (which is essentially what a “breakout” projection is) has some merit…

Just because Joe isn’t good at it doesn’t mean the rest of us aren’t.  I mean, I nailed my early season breakout guys last year - Corey Patterson, Jeff Keppinger… oh.

One this that BP does that bugs the crap out of me is how they hedge against PECOTA. On one side they claim it’s the most accurate forecasting system in the world and then in the capsule comments they hedge against it by saying things like “he should easily beat that projection”.

Full disclosure: I have not bought their annual for a few years now so maybe they don’t do that anymore, but back then they lost a lot of credibility with me. Of course, if Joe is predicting break-out candidates that were not picked by PECOTA then essential they are still doing it.

Matt

And Joe and BP should know better.  If it is Buster Olney or Jason Stark, we give them a pass.  It is lists and statements like these that give otherwise good sites (BP) and people (Joe) a bad name and take their reputations for honest, good-quality work down a notch, again, at least in my opinion.

Not to venture too far off course, but does anyone else get the sense BP is becoming more of a baseball reporting organization and less of an analytical one? I believe Joe’s (and by extension BP’s) method for “breakouts” still comes from somewhere within PECOTA, as they still publish those percentages in the player profiles. Of course, PECOTA is a Nate question, and Nate’s more focused on Obama right now. Not that BP would let him divulge much…

Sheehan also listed players that he excluded from the list of “breakout candidates” for semantic reasons: Billy Butler, Phil Hughes, Lincecum, McGowan, Hermida, Hart, Crawford, Felix Hernandez, Bonderman, Cano.

The Marcels for the hitters were 2528 PA, .301/.356/.479. In 2008 they had 2810 PA hitting .267/.313/.417.

The pitchers had a Marcel of 711 IP, 4.11 ERA, 3.94 FIP. In 2008 they threw 644 innings with a 3.58 ERA and 3.70 FIP.

Can’t complain about the one pitcher on his breakout list, Ervin Santana.

It is one thing to list a group of players who you “think” have a higher than average chance to exceed their projections.  If you do that and still stand by their “weighted average” projection, then you are in essence simply saying that these players have a higher volatility than the average player.  However, if you list a group of players who have a “good chance to have a breakout season,” I think that you are clearly saying that the collective performance of these players will significantly exceed their collective projection.

This is clearly a testable proposition, although obviously you would want/need a large enough sample size in order to “make a statement” about the results of that testing.  Then again, the burden is also clearly on the person making the statement to show evidence by virtue of the results that he is more than just a talking head like the guys on CNN who tell you which companies you should or should not buy or which way the market is likely to go (up or down).

And yes, for someone or some company to proclaim a projection for a player and then to say, “I/we think that so-and-so will or can easily exceed (or under-shoot) that projection” is a ridiculous statement, which makes no sense whatsoever.  Obviously any player CAN (and will) over or undershoot a projection.  We don’t need to be told that.  But, if you think that a projection is low or high, then you are not being honest in giving us that projection.

Now, if you are saying that you don’t think that the projection for a certain player is very reliable, for whatever reasons, that is a fair statement to make.  Obviously a player with little historical playing time will have a projection with very wide “error bars”. As well, if you expect a player to get little playing time, his results, as compared to his projection, will also fluctuate a lot.  There may be other reasons for a projection with less (or more) certainty than the average projection, and I don’t mind if you tell me that you think that a certain projection is more or less reliable than average (and I would certainly like to test your accuracy on those statements as well).  But, as Matt in #3 says, don’t tell me that you think or expect so-and-so to exceed their projection.  That makes no sense, unless you don’t believe your own projection of course.

I’ll add one thing to that, in favor of BP and other people or entities that do/say that:  Let’s say that BP is committed to posting Nate’s Pecota projections, however they come out.  If one or another BP writer does not “believe” or disagrees with Pecota, it is their right to tell us that.  Again, I’d certainly like to test their disagreements with Pecota.  Any idiot can “disagree” with a projection.  Whether they turn out to be “right” (in the long run of course - after lots of data) is another story.

The other thing along those lines is that a person themselves can spit out objective projections, not “tweak” them, and then simply say, “This is what the computer says using my projection methodology - I am going to present them exactly as they come out - but I will tell you which ones I ‘disagree’ with - for various reasons which I may or may not articulate.” When I “spit out” my projections, there are some that I don’t like.  Now, I don’t usually go so far as to say that I think they should be higher or lower - but that is only because I have no evidence that I have an talent for properly and correctly tweaking these computer projections.  Just because I might “think” I do, means nothing.

Here were his picks, and their percentile EQA (from Pecota) for hitters in 2008:

E Santana (breakout)
Kinsler (97)
McLouth (93)
Kubel (80)
Navarro (80)
Weeks (30)
Zimmerman (25)
M Cabrera (20)
Francoeur (5)
D Johnson (essentially OOB)

Clearly he would have done just as well by randomly picking ten 25-year-old players.

I guess I wonder what he was getting at.  Pecota gives a distribution of outcomes.  He’s essentially either claiming that the 50th

whoops...apparently percent signs do weird things to posts…

He’s essentially either claiming that the 50th percentile projection is too low; or that the distribution is wrong (high side too conservative?) If that’s in any way true and known to BP insiders, it seems like it’s a systematic flaw in Pecota, one that would need to get fixed.