Wednesday, July 27, 2011

Baseball Stats on TV - Fielding Percentage Is A Terrible Stat

I love advanced baseball statistics. I'll admit that at the outset. I recognize, however, that TV networks do not have some sort of duty to educate casual, traditional fans regarding the virtues of advanced statistics. As much as networks get blamed for perpetuating myths regarding the value of, say, pitcher wins, you can't blame them too much for providing their audience with the information it wants.

The only statistic provided by most networks regarding a baseball team's defense is fielding percentage.

Virtually anybody who follows baseball closely knows that fielding percentage is a terrible statistic. It measures nothing more than the percentage of plays in which a team (or player) makes an out rather than an error. The stat tells very little about a player's defensive ability, other than that the player doesn't make a ton of obvious mistakes. It doesn't tell you how often the player does the spectacular. It doesn't penalize sloth-like players who can't ever reach relatively easy fly balls--but don't bungle the catch when they actually make it in time. It doesn't punish players who take terrible defensive positions. It doesn't reflect the situations where players decide they'd rather not make a drive or crash into the wall in chasing after a ball. Whether a given play is a "hit" or an "error" is an arbitrary judgment that varies tremendously depending on the official scorer. Willy Mays turned a lot hits into outs. Manny Ramirez, on the other hand, raised a lot of batting averages even when he didn't commit errors.

Fielding percentage is quite obviously a terrible way to measure defensive performance. But, you ask, what are the alternatives? After all, most advanced stats (like plus/minus or UZR) are pretty controversial and depend on all sorts of measurements that casual fans don't see.

There is, however, a better stat that is easy to understand. Frankly, I'm very puzzled that it hasn't been adopted already. That is defensive efficiency. Defensive efficiency is just the percentage of the time that a fielder turns a ball in play into an out. When calculating defensive efficiency, we don't ask whether a defender should have made a play. We just ask whether he did. Now, on any given play, defensive efficiency might not tell you much. No defender is going to catch a screaming line drive to the gap. But, over time, defensive efficiency is a much better stat than fielding percentage because it covers more things that we want to know about how well a person defends his position. Players get credit for making players that they "shouldn't"; they also get punished for not making plays that don't involve obvious screw-ups but that most defenders would make.

Defensive efficiency is easy to understand and vastly superior to fielding percentage. So, why don't we use it more? Your guess is as good as mine.

Wednesday, July 6, 2011

The Casey Anthony Verdict

I did not follow the Casey Anthony trial intensely. But, I was massively annoyed by the overreaction to the verdict.

First, I'm generally curious when the public, at large, chooses to assume that the jury in any particular case gets a result massively wrong. Don't get me wrong, juries *are* sometimes wrong. Innocent people are convicted, and guilty people are acquitted. But, the jury generally has more information than Joe Public. Jurors are forced to endure trials, hear all the evidence, listen to the arguments, and receive instruction from the judge. Even ardent followers of trials generally hear less of the evidence. When a jury unanimously reaches a conclusion, people should take that verdict seriously. Except in unusual circumstances (like suppressed evidence), it is not reasonable to assume that a group of 12 individuals was stupid, biased, or otherwise irrational. It's extremely egotistic to assume that you, with less information, can arrive at a more sound conclusion than a jury.

Second, I had no objections to the verdict. In fact, it mirrored my own intuition: that Casey Anthony was probably guilty but that the prosecution hadn't proven its case.

Reasonable doubt is a very high standard of proof. The standard in civil cases is a "preponderance, i.e. "more likely than not" or 51%." Beyond that, there's "clear and convincing." I've always taken that to mean about 66%.

Judges are reluctant to quantify "beyond a reasonable doubt" to a particular percentage. It's far more than preponderance but less than "beyond a shadow of a doubt." The common definitions I've heard in court are "moral certainty" and "a certainty that you would rely on in your most important, personal affairs." Both definitions are, frankly, quite bad. The meaning of "moral certainty" is neither obvious nor intuitive. The level of proof that one would rely upon in their personal affairs is positively misleading. For example, I wouldn't trust someone who I thought even 2% likely to have abused a child to babysit my kids. On the other hand, I wouldn't bet my mortgage on even a 98% chance. Whatever. Personally, for me, beyond a reasonable doubt means around a 93% chance. Yes, that's arbitrary--but so are all the other definitions.

Popular culture goes into two different, and wrong, directions regarding reasonable doubt. First, people equate acquittals with innocence. I cringe when I hear people say that OJ was found "innocent" by a jury of his peers. He was not. He was acquitted under the "beyond a reasonable doubt" standard. In the subsequent civil case, he was convicted under the "preponderance" standard. So, in truth, juries collectively determined that OJ probably did it, but not a very high level of confidence.

The "not guilty" verdict doesn't mean Casey Anthony was innocent. That verdict doesn't even mean the jury thought she was innocent.

If you don't like it, moreover, you can do something about it! People behave as if "proof beyond a reasonable doubt" is a standard engraved in the stars. It's not. The "beyond a reasonable doubt" standard is a policy choice, reflecting the fact that our Founding Fathers would rather set a guilty person free than send an innocent person to prison. But, we don't have to agree with our founding fathers. We're free to craft our own standards. Granted, the "proof beyond a reasonable doubt" is mandated by the Constitution's due process clause according to the Supreme Court. But, guess what? We can change the Constitution.

Here's one proposal: create two different verdicts. If the jury thinks it's 51% likely that a person committed a crime, put them on parole. If they don't commit any crimes within 5 years, expunge the conviction from the person's record. If the jury thinks the prosecutor proved the case beyond a reasonable doubt, then check that box and send the person to prison.

Such a dual-choice system would impose different punishments based on the jury's level of confidence that the person committed a crime. Furthermore, if the jury is wrong, then the person can escape future, unwarranted punishment by avoiding other crimes. But, society has additional ability to monitor the accused during the period of parole. In other words: trust, but verify.

My guess is that prosecutors wouldn't actually like this option. If you gave many criminal juries that choice, my suspicion is that quite a few would choose the lesser option. The truth is, we're rarely totally certain about guilt or innocence. Eye witnesses, scientific studies show, are often wrong. Human beings are awful lie detectors, so there's not much reason to think juries always get it right in swearing matches, e.g., he said/she said cases.

Our system of justice isn't written in stone. We change change it. If you don't like the jury verdict in the Casey Anthony trial, think about whether your gripe is with the jury--or the system. If you don't like the system, think about changing it.