THE BOOK--Playing The Percentages In Baseball

People would rather be certain than right in the short run, because being certain, even when you’re wrong, makes you less anxious.  Confirmation bias assumes that we live in a mental world where people are either all good or all bad (and we have the evidence to prove it!) It’s a brave man that can live in an uncertain world.

Very interesting.  Is that why negative political ads are so effective?  Instead of us having to think about things about the various uncertainties in everyone’s position, having a black/white “he’s associated with a criminal” ad makes things easier for us to decide upon?

Absolutely.  Think about what would happen if we held a real, detailed, honest discussion of the actual policy proposals of the two candidates were to take place, even assuming a country of voters who a) got it and b) cared.  They’d both be forced to admit all the uncertainty that was involved in both of their proposals.  About the best that they could say are things like “Well, there’s a 35% chance that we’d have an economic recovery, which is better than my opponent’s 33% chance.”

This is why candidates rarely actually run positive ads other than the “picture of candidate looking off into the distance with the American flag in the background” one.

Negative ads do two things for a candidate.  If I made an ad that said that B-Rock (or Johnny Mac, doesn’t really matter) likes to torture orphaned puppies, and you already don’t like B-Rock, well there’s one more piece of evidence on the pile.  If you do like B-Rock, you will largely ignore the attack, or you’ll begin spinning ("that bill wasn’t about torturing puppies, it was about creating some much needed character and endurance in our canine population") or take it as evidence that the other guy is a conniving political opportunist who will “say anything to get elected” while passing over the fact that your guy is also a conniving political opportunist who will say anything to get elected.  And he will probably hit back with an ad next week about his how his opponent once supported replacing the National Anthem with a duet between Hannah Montana and Kevin Federline.

The more important function of negative ads though is to prey on that all-or-nothing thinking in a different way.  Let’s say I’m truly on the fence, and I like a little bit from B-Rock and a little bit from Johnny Mac.  One of them comes out with a negative ad.  Well, I certainly can’t vote for that guy!  He favors torturing puppies.  It might not make me vote for the attacker, but as long as I’m not voting for the attackee, then their mission has been accomplished.  If there’s something wrong with that candidate (whether or not it’s actually true), I can’t vote for him… the rest of his policies will probably be just as awful.  Negative ads drive down voter turnout.

That was a great post. 

Unfortunately, you are the first person to actually use Fed-ex’s name in this blog, ever.  Confirmation bias must therefore force me to disavow any appreciation I may have had for your thoughts.

So Hannah Montana is fair game?

"If you do like B-Rock, you will largely ignore the attack, or you’ll begin spinning ("that bill wasn’t about torturing puppies, it was about creating some much needed character and endurance in our canine population") or take it as evidence that the other guy is a conniving political opportunist who will “say anything to get elected” while passing over the fact that your guy is also a conniving political opportunist who will say anything to get elected.”

That’s some fine backward thinking by Pizza Cutter’s hypothetical person. And sadly, this is how most people think (about one thing or another anyway).

How does that old line go? “We don’t know what we love; we love what we know.” Those two clauses don’t actually relate very well now that I think about it....

"The only way out of it is to rely on data, like UZR.  But, if the data goes against your biases too much, you will simply ignore that data, thereby reaffirming the initial bias you wanted to avoid!”

It’s true. I had an online discussion at ESPN regarding the merits of Shane Victorino’s (and eventually, Aaron Rowand’s) defense. Phillies’ fans were aghast at the Fielding Bible +/- because no, their eyes told them Shane was a superb centerfielder, but his +/- at the time was -4. He’s definitely improved from earlier in the season (I think he’s at +6 now) but some people just don’t believe what their eyes tell them otherwise, even if a fairly objective system tells them so.

When I showed one commenter the THT RZR’s for centerfielders, he pretty much called it garbage because Aaron Rowand was listed as the highest RZR. He wouldn’t accept a metric, not because of its methodology or anything like that, but because of what the results said. I say all this because I’m not so sure about RZR because Rowand is still ranked up there, but then that’s an example of my bias. Fortunately, I don’t only rely on RZR for my defensive metrics.

And therein lies the problem with getting people to accept Sabermetrics.  People want to believe that Player X is amazing.  We say that he’s not.  We must be wrong.

UZR and its ilk are not objective.  They are numeric.  That’s not the same thing.

UZR (etc.) are objective measures, I think. They are an objective recording of plays per chances, for a specific definition of chances.

Now, we can (and should, and do) take time to consider what biases the definition of chances may or may not, how well it measures actual defensive skill, and so on. But it’s not a subjective opinion but an objective statement of certain facts. The question is what the meaning of those facts are.

Going back to the original post, doesn’t Jeter have one of the lowest agreement levels of any player in the Fans Scouting Report? For me, that’s evidence that the fan ratings are influenced by the wider discussion.

Yes, Jeter always has one of the lowest agreement levels.  Good point.

Colin: it is an objective statement of facts and opinions.  How hard a ball is hit, where it was hit, what it’s trajectory was, are all opinions.  Those opinions may or may not have high uncertainty levels.  MGL also applies park factors with subjective regression in certain cases.

As well, UZR has gone through different versions, and in a few cases you get wildly different results.

I’m not saying that that’s not good.  He’s got a keen baseball sense.  But, let’s not call UZR objective.  It is highly intelligent, and uses somewhat subjective decision-making on data that has alot of subjectiveness to it.

I also imagine that in the case where everyone is sitting around the table deciding what to do about Bob the AA 2nd Baseman, there are incredible social pressures to reach a consensus. So, not only do you likely have the “Confirmation Bias” at work, but you also have a very strong normative influence.

Someone should replicate Asch (http://en.wikipedia.org/wiki/Asch_conformity_experiments) with baseball statistics.

I think that in the spirit of the commonly accepted definitions of objective and subjective as they relate to defensive evaluation, I am very comfortable calling UZR or any other metric for that matter “objective.” While there are certainly subjective components to UZR as there are to many metrics (e.g., errors in anything that includes hits and/or errors), I don’t think too many people here or otherwise would
disagree with the dichotomy of objective and subjective primarily referring to using numbers and data to rate a player defensively, and using observation and perhaps the opinions or statements of others to formulate their own “opinion” on a player’s defensive value or skill.

I looked up several definitions of “subjective” and “objective” and without splitting hairs, there is little doubt that baseball metrics, including UZR (and fielding percentage and virtually everything else) fall into one category (objective) and something like, “I’ve seen that player play a thousand times, and he looks like a great defender to me,” falls into the other category (subjective).

For example:

Based on (or related to) attitudes , beliefs , or opinions , instead of on verifiable evidence or phenomenon.

I would think that verifiable evidence or phenomena can include some subjective elements, especially if the subjectivity is not related to the question at hand.  For example, while the classification of how hard a ball was hit, or a park factor computation, in no way, shape, or form, is influenced by the players themselves, such that if there is a bias for or against a certain player or team, it is likely purely by accident (I suppose you could say that I might subconsciously have a poor park factor in LF at Fenway because I do not like Manny’s defense, but that would be a stretch, I think).

And BTW, is anyone else not the least bit surprised that a few scouts and front office people sitting around a table with beer and cigars making decisions about who gets cut and who doesn’t is not going to be very productive, scientific, and accurate, whether it be due to “confirmation bias” or just plain lack of competence?

I would hope that rather than looking at this week’s preseasons stats that they are instead judging a true talent level, by whatever means, and then determining at what level a player is ready to compete.

I would say subjective is based mostly on opinion, and not on numbers. UZR is modeled on some assumptions, such as assigning split zones, and what types of plays to include and what not to include, but I would certainly classify it as objective.

Dan/13.  An excellent idea.  It ties into the idea of “groupthink.” That’s the pressure on a room full of people to stifle debate and come to a consensus, however pre-mature.  I think that baseball teams (and other people who make decisions in groups) would benefit from hiring a designated “contrarian.” If I said that Mother Theresa was a kind human being, it would be the contrarian’s job to argue that she was mean and evil.

I recommend the book “The Big Sort” for several reasons, but in this particular case, there’s a chapter in which the author (I’m forgetting his name right now… it’s way too early in the morning) talks about how this sort of thing works in politics.  Worth the read on several levels, especially if you are interested in this topic.

Is OPS+ considered objective?  That is, the haphazard combination of OBP and SLG, adjusted by league and park.... is that objective?

And if two people calculate the same metric different ways, are both of those metrics objective even though they are “same but different”?

Is it an objective assessment to say that Carlos Gomez is one of the fastest players in baseball, by simply looking at him run?

You can split hairs on things that are objective and subjective, and it seems to me that the reason people do so is to give greater weight to an objective assessment than a subjective assessment, when in fact it could very well be, and often is, that the subjective assessment should carry more weight.

Whether a certain subjective or objective assessment is better than the other or what the weights should be if you combine them is another story, but I think that it is pretty obvious which things are which.

Heck, I am a firm believer in subjective
assessments for pitchers.  I think I can guarantee that if you give me two or 10 pitchers with the same projection, if I watch each one for 3 games each, I can differentiate among them (and change their projections) quite easily.

I mean it does not take a genius to see that Lincecum (or Joba) is the real thing after watching him for a few games when he first came up, as compared to say, Clayton Kershaw (although I am not suggesting that they had the same projections based on their minor league stats - I don’t know).