Tuesday, August 24, 2004

Bill James on projections

Richard Lederer has been reviewing the history of the Bill James abstracts on his weekend blog. I played thousands of games of Strat-O-Matic in the 1980s - even joined large organized leagues with a draft and all that - but somehow missed the Bill James thing.

Lately there's a tendency in a lot of the baseball discussion that takes place over the internet to obsess with projections. A player like Jack Wilson exceeds all expectations and some people have more faith in a projection system than they have in their own two eyes. Because he exceeds projections, his improvement cannot be believed by some. The impulse should be "Back to the drawing board, we were really wrong in our analysis of that guy's past."

Another player has a dynamite year at Double A but because he is one of the older players in the league, the meaning of that performance is severely discounted since the age does support an optimistic projection of next year's stats. The truth is that performance is performance, and just because a player is a certain age, that doesn't mean that he didn't have all those eye-gouging RBI numbers. The age may temper our expectations since we've rarely seen guys advance far from that age at that station, but you know what, just because it hasn't happened often before doesn't mean it won't happen often in the future. With the greater emphasis on college players and the new reluctance of teams to delay starting a player's arbitration clock, there's no doubt that we will see more 24-year-olds having great seasons at AA that we used to see, say, ten or twenty years ago.

Here's a smart thing that Bill James wrote. I lift it from Rich's blog:

The goal of science—and sabermetrics is a science—is not to predict what will happen but to understand what does happen; predictability attains significance only as a test of knowledge. However well we might speculate about the future, it is an article of faith that that future, once accomplished, will resemble the past far more closely than it resembles any of our speculations about it.

The meaning of stats lies not in their ability to predict the future but in their ability to describe the past. So the primary goal of statistical anaylsis of baseball should not be the accurate prediction of the future stats but the fuller and wiser comprehension of the previous stats. Why? Only one of these goals is within our reach, and the other will never be achieved until the first is met.

No comments:

Post a Comment