Observation vs. Statistics

Nothing very relevant or topical about this, but I had some thoughts about sabermetrics and its critics. A common criticism goes something along the lines of, "You can cite all the numbers you want, but I saw so-and-so play, and you can't tell me they weren't great."

Now, leaving to one side the fact that you can very easily be deceived by what you see, I'm very intrigued by this concept that seeing a player tells you more than their stats do. I think this is a curious concept in a game like baseball, which is much less about what you do on any given day than about how much you do over a long period of time. Case in point: take a player like Chad Moeller. On April 27, 2004, Chad Moeller hit for the cycle. If you saw him on that day, you'd think he's a great player. But Chad Moeller has a lifetime OPS+ of 61. He's terrible. The only way you'd know that is by watching him over a large number of games, and seeing how he tends to do.

And what exactly happens when you watch someone over a large number of games? What exactly do you see that lets you know if they're a good player or not? You certainly don't see them get on base every time they step up to the plate; the odds are overwhelming you see a player make an out more than half the time. You do, however, see them get on base sometimes, drive in runs sometimes, get big hits sometimes, etc. More importantly, you see how often they do this. Even the worst player will hit a home run sometimes, and even the very best player will strikeout sometimes, but what separates the best from the worst is how often they do each thing. So if you rate a player well, you remember seeing them succeed often. Well, what's better? Saying, "Yeah, I saw this guy, he always came up with the big hit," or going back and counting...and seeing just how often the guy came through? I mean, any time you're trying to figure out how often something happens, which is better: trying to remember how often it felt like happening, or going back and counting how many times it actually did happen? Clearly, counting, that is, figuring out what the actual facts are, rather than just what they seemed like, is the way to go.

Of course, even the most vehement saber-critic recognizes this, which is why they still use numbers like batting average and RBI. More often than not the argument is not about whether to use numbers, but about which numbers to use. It is quite clear we need to use numbers, contra any "I saw him play" argument.

This brings to mind another thought which I've had on a few occasions. I feel like even many non-sabermetrically inclined people recognize that batting average and RBI and so forth don't tell you everything, and that's why they're so quick to say, "The stats don't tell you everything; you have to watch them play." They're used to stats that can be very misleading, and so they insist the same about all stats, even though sabermetric stats are designed to avoid the pitfalls of more conventional stats. It's almost like a Catch-22.