THE BOOK--Playing The Percentages In Baseball

Agreed. At the very least, the positioning of the defense would be different.

speaking of announcers, it was nice hearing today, with the UT TTech rematch game, the musberger call of the crabtree touchdown.  especially because the way he says it sounds like, CROB TRAY BREAKS FRAY!

At least when Michael Kay says this shit, he follows it by saying “that is, if you believe in the fallacy of the predetermined outcome.”

The one and only time I will praise Michael Kay. I feel a little dirty doing it.

I didn’t know that about Kay.  I watch lots of Yankee games and I have no problem with Kay whatsoever. I don’t know why he gets bashed so much.  I think it is mostly because of the ubiquitous Yankee hating.  Kay and Paul O’neill are quite entertaining in my opinion…

Well, I think this is a bit harsh.  They do not need to believe that things would have been *exactly* the same to make that statement.

In the parlance of philosophers of language and formal semanticists, they merely have to believe that it cost them a run in the closest possible worlds.  That is, in the set of situations relevantly similar to the actual one.  This is a pretty standard analysis of counter-factual reasoning - which we ought not rule out tout court.

That’s not to say the claim is true - differences in pitch sequencing, fielder positioning, psychological differences, etc. etc. are all at play.  At issue, though, is whether these details are sufficient to undermine the claim (namely, what situations qualify as the set of closest possible worlds).  If so, then it is an issue settled on a case by case basis, and not simply settled by an overly restrictive interpretation of the historian’s fallacy.

Those details are sufficient to undermine the claim, Dr. Wittgenstein.

The Pirates announcers do make a habit of saying, in this kind of situation “of course you can’t assume that everything would have happened the same”

Though I’m pleased to see the name in conversation, the views I used to make my point that counter-factual reasoning is formally respectable are hardly Wittgensteinian

I also was not particularly concerned to evaluate whether those details change the situation sufficiently to undermine that claim.  My point rather is that it is an empirical question that rests on the details of the situation.  Where was the ball hit?  Where would the fielders have been otherwise?  What pitch was thrown?  Would that pitch have been thrown otherwise?  My point is merely that I don’t think this is an instance of some general fallacy of reasoning, as I interpeted mgl to be suggesting (particularly in his forceful “anyone who is not mentally retarded who thinks...” claim).

MGL - I’m sorry that you have to watch Giants telecasts!

You’ll love this one that Kuiper and Fosse gave us on the A’s cast a few days ago:

“I can’t stand it when guys who didn’t play in the majors say they never got a chance.  If you’ve got the skills, you’ll get a chance to play in the bigs.”

jm, I am no philosopher, but, We have no idea what would have occurred had the runner not gotten thrown out (or whatever happened). None whatsoever.

The closest we can come to estimating the impact of the “gaffe” (the runner getting thrown out or whatever) is the standard lwts value of that out, given the outs and runners.

If I were an announcer, I would say something like, “Darn, that base running mistake just cost us .46 runs,” or something like that…

"If I were an announcer, I would say something like, “Darn, that base running mistake just cost us .46 runs,” or something like that.”

And there is one reason that MGL is not an announcer. They probably don’t need to say anything, but I’d rather hear the original comment than something like this. Precision has a place, and I’m not sure the commentator’s booth is the place to start.

We don’t know what would have happened, sure, that’s the right formulation of the historian’s fallacy.  Counter-factuals are just that, contrary to fact, and we don’t have any access to what would have occurred in those situations.

I think the pessimism about counter-factual reasoning, though, is a bit unjustified.  What we can do is make an inductive inference based on the other facts available to us.  Take another thread we had here recently - tango’s on peer review.  He had a nice post where he posed the question as: if we were starting over, would we adopt this model?  That calls for counter-factual reasoning, what other models (which we do not currently use!) might have done better than what we have? 

It is a tricky question of course, precisely because we don’t know.  So we do our best to look at what we know about other models, and make good inductive inferences about how they would play out if the facts were different (e.g., if we had an open review process which was widespread with institutional support).  The announcer here is assuming that if things had played out differently on the basepaths, the situation would still be similar enough that the same subsequent outcome is likely to have happened.  If the inference was made explicit, it might make reference to the ability of fielder’s to catch a certain type of batted ball, that the pitcher was throwing similar types of pitches when hitters were on base, etc. etc.

I think I actually agree with you about part of your point - namely that it is folly to suppose that we know what would have happened, and that the certainty with which announcers make these announcements is rather silly (especially the further the causal chain gets between the events).  My disagreement is rather with carving out some space for justifiable counter-factual reasoning.

I suppose you could say that if runner A would not have gotten thrown out in the Yankee game, the same thing would have occurred (that did occur) in the Braves game 1000 mile away.

But since every pitch by the pitcher and every swing (or not) by the batter is predicated on the the score, inning, base runners, and outs, the answer to the question, “What would have happened had X not occurred?” is quite simple.  “We have absolutely no idea.” The hit that did occur is merely one possibility, and in fact, one of the LEAST likely possibilities!

Now, the answer to the question, “What likely would have happened?” is also quite simple.  And the answer to that is, “The same thing that would be likely to happen in the future, which is the distribution of events which reflects the odd-ratio (log 5) matchup between the batter and pitcher (given the outs, inning, runner, etc.).”

Yes, but the hit is what happened. In this case, the caught stealing did cost them a run. In some cases, it wouldn’t. On average, it costs .46 runs. I don’t see anything wrong with the comment, frankly.

Maybe the announcer should just say “Boy, that triple would have been more satisfying if there had been a man on first.”

Same sentiment, with fewer handholds for the nitpickers…

Wow, David (Gassko), I fail to see your position even slightly.  I guess that makes for an interesting world - that two smart people can have different perspectives on things like that.

My question to you is, “Had the runner not been thrown out - i.e., if we could go back in time - what would have happened next?” I fail to see how there is any answer to that, any more than there is an answer to the question, “What is going to happen in the next PA?”

I think my point can made just as well in reply to the question put to David Gassko.

“Had the runner not been thrown out - what would have happened next?”

Is there *a* answer?  No, not really.  There is some set of possible worlds (situations, outcomes, whatever) that are consistent with that.  It is infinitely large.  Within that set are probabilistically likely events (sets of possible worlds that contain that outcome).  Counter-factual reasoning is inductive in the sense that one can only infer to an outcome that is highly probable.

One of our best sources of evidence here is the vast array of data collected in the statistical analysis of the game.  That is generalized data of course, but very useful.  Still, we are trying to predict a one-off event, which is far too complicated for this kind of analysis (too many variables to account for).  One source of evidence about what would have occurred is what actually occurred, operating on the assumption that a sufficient amount of those factors would remain fixed between the scenarios (two possible worlds with as much in common as possible).  What the announcers claim could be (charitably) interpreted as is the following: given that the world would remain the same with the exception of details x, y, z, it is highly likely that the hit would have still occurred and the run would have likely scored.

Your reply is that, well wait a minute, if you change x, y, z, there are a number of other details in the situation that need to be changed as well.  I’m comfortable with that, I think we tend to reason a bit too casually counter-factually, and the announcers are likely doing so.  Again, though, my point is that this is always an empirical issue - we have to look at each case and determine if the changes resulting from value changes for x, y, z are likely to causally affect the outcome.

(One could push this worry to a logical extreme, noting the interdependence of all events and suggest that no probabilistic reasoning is even possible.  I’m not sure how far that this goes, but given the ubiquity and importance of counter-factual reasoning in day to day life and in science, this would be a problematic conclusion.)

MGL: since you don’t like the “reconstruction” of the inning, then you must hate that they even have something like “ER”.

My question to you is, “Had the runner not been thrown out - i.e., if we could go back in time - what would have happened next?” I fail to see how there is any answer to that, any more than there is an answer to the question, “What is going to happen in the next PA?”

---

If you want to argue that the players would have approached the next PA differently, and so the triple might not necessarily have occurred, that’s fine by me. But I’m also fine assuming that this is not the case, as the announcer did. The point of his comment was to point out how costly a caught stealing can be, which is a good comment to make. In this case, reconstructing the inning was a particularly effective way of making that point, and I have no problem with it.

Right, the announcer was not trying to make a statement about conditional logic, or chaos theory, or anything of the sort.  He was just offering the observation that the CS hurt.  I don’t know why people would think he was trying to make a profound statement of any kind…

Actually, I have no problem with ER.  Actually, on the average, ER WOULD be exactly how many runs a pitcher would have given up had the error not occurred (of course that is in a world where no one makes errors).  When an announcer says that “such and such” cost a run or 2 runs (or whatever it is) they are always (greatly) overstating the average cost of the gaffe.

We’ve discussed ER before.  The ONLY problem I have with it is that it is (slightly) biased in favor of GB pitchers and pitcher with low K rates.  If that weren’t the case, then I would have no problem with it whatsoever.  It actually helps to smooth out the random fluctuation in a pitcher’s RA.  It is something like ERC or FIPS, other than the bias!  It is a theoretical RA and not an actual RA, to some extent.

MGL: for consistency’s sake then, would you want EH and EHR and ESO and EBB?

For example, all runs scored after an error that would have been the third out are unearned.  Therefore, why not also remove all hits and HR and BB and SO that occur after that error?  Why are we removing runs as earned, but not removing hits and walks as earned?

After an error that would have been the third out?  Sure, why not?  The difference is that “earned” runs are specifically defined as such.  If you want to create another category called “earned hits,” etc., then, sure.  Again, if not for the bias that favors the GB and contact pitchers, ER is just fine - it simply reduces some of the luck. 

As I said before, it is exactly like DIPS or FIP where you take the defense completely out of the equation.  You don’t seem to mind DIPS ERA or FIP, do you?  So why do you mind another metric which takes errors out of the equation?

I don’t like the way an inning is reconstructed for the non-2 out errors.  I also don’t like the inconsistency of not having EH, EHR, EBB, ESO.

I think the bias between GB/FB pitchers is big enough.

Also, since errors are subjective, they may also have bias.

Basically, I think it’s a poorly constructed framework, even if you want to argue that it may be better than ignoring errors altogether.

Ordinarily with a great hitter like Pujols in the on-deck circle and two outs, a pitcher would do anything to get the hitter and not have to face him with men on base.  But if minimizing your “EHR” is what you want, you’d actually want to have someone make an error (maybe you could throw a come-backer into right field), then you get to face Pujols at zero risk to your “EHR” metric, and you won’t have to face him “at-risk” again until his turn comes around again…

Cy Young Award, here you come!

Actually, this is probably an argument against the entire earned/unearned concept, rather than just the extension of the same…

It’s also an argument against letting statistics other than runs influence the play of the game (save rule, I’m looking at you).  Analysts in general, and most fans as well, treat unearned runs like they never happened, or at least that they had nothing to do with the pitcher, which is of course ridiculous…

Michael Kay gets bashed by Yankees fans, not Yankees haters (or maybe it’s both).  Familiarity breeds contempt.  I’ve been listening to him all season long for 12 seasons now.  There is worse out there, for sure, but after 12 seasons you’ve heard pretty much all the good stuff he’s got to say.  And you get to hear the bad stuff over and over.  He has a tendency to harp on an argument far too long, as well. 

He spent an entire inning discussing Carl Crawford getting thrown out at 3b by Golson to end a game.  The others in the booth (Ken Singleton and O’Neill, I think) were pointing out that the odds of Crawford being thrown out on that play were tiny, and thus perhaps it’s not some violation of the Holy Commandments of Baseball that he tried for 3rd in that situation.  Kay wasn’t adding anything to his initial “you can’t make the 3rd out at 3b!” He just kept repeating it.  Meanwhile, a game was in progress that he wasn’t calling.

I’m sick of the “fallacy of the predetermined outcome” line, b/c I’ve heard it a million times.

Kay-O’Neill is good in part because Paulie calls Kay out when he says stupid things sometimes.  It can be pretty funny.