Julien's Faster than Light Blog

Jump to top
Friday, January 23, 2004
it's nothing personal . . . i attack everyone.
alan schwarz wrote this article at a third grade level and i decided to make fun of it. no disrespect to third-graders. the title is "whip it good . . . statistically."

2003 obp, oakland: .327, 21st in the majors.

ooooooooh! studies!

thems must be science-ticians!

this is an almost useless article because it completely ignores the idea that different statistical measures can be combined for a complete picture. instead it compares apples and oranges to see which is "best".

actually, the accuracy of an arithmetic calculation depends on the person doing the calculating. the "computer", if you will.

ops has the problem that it counts batting average twice. batting average requires a huge sample size to accrue meaning, so huge that by the time you have it, the player's skills have changed.

so the predictive value of ops is nil. but it is a pretty good quick-and-dirty performance measure. having wrote that sentence, i must admit that i don't believe it. i'd rather see obp and slg separately. actually, for performance, show me avg/obp/slg and i'm reasonably content.

having written that sentence, i must admit that i didn't use the perfect passive particle properly. luckily, we no longer speak latin. or unluckily. latin is an elegant, practical language.

you only get one!

actually, give me one, and i'll take whip. . . . they're both horrible, because they include hits, but at least whip measures something. ops has randomness from so many places that you can't tell anything from it. it's a statistical tower of babylon. i am so literary. i'm writin fuckin literature over here.

wow! we should use this for other things, like politicians!

whip is most affected by control, defense, and luck. its value comes from the extent to which it approximates k/9 and k/bb.

or you could just look at what actually happened.

the pythagorean method is a great, useful tool invented by bill james that, like most james inventions, has been completely overused and blown out of proportion. it may be the single biggest problem in sabermetrics today.

i sum up its limitations: everybody knows that a great closer helps you win a lot of games. but to the pythagorean method, the ninth inning of a close game is no different from the fifth inning of a blowout.

we know there's a problem here but sabermetricians say there isn't. guess what, folks: that's an assumption. and they base all kinds of labrynthine calculations on it.

it's assumptions what kill science.

bill james, by the way, does not overuse his inventions. his current reliever valuation includes a "close game" factor.

oh. good. because squaring things is so hard.


uh . . . obp is important because getting on base is really good, because outs are really bad.

as long as it's helpful.

dimensions, sure, but total park effects, i don't know. coors field, for example, has more of an effect on obp than slg. that's about the most surprising counterexample possible.

the other thing he said is wrong, too. for an example of a team that scored a lot of runs playing station-to-station ball, look at the 1998 yankees.

most things people say are wrong. including this.

no, he needs strikeouts, too. but if you're left-handed and pitch in safeco, you don't need as much.

batters didn't catch up to him, he just stopped being hit lucky.

read: red.

bill james, for example, is a boston executive.

never heard of it.

this is funny because it's the same thing anti-sabermetricians say:

his job is to win games. if he succeeds at that, he's done his job.

my cat could catch her tail. then she would bite it. then she would stop.

"it has long been my observation---informal, not suported by research---that when a team has a fast outfield, they tend to have a good defensive efficiency record. . . . it is not illogical to argue that certain markers of defensive excellence may be tied to certain positions. if a team allows few stolen bases, we assume that they have good defensive catchers. if a team turns double plays, we credit this to the shortstops and second basemen. it is equally reasonable, when a team has a high der, to associate this more with one defensive postition than with another."

---bill james, win shares

there are 27 outs per team in a nine-inning game. you cannot change that.

better teams actually tend to make fewer plays per game, because they have better pitchers, and better pitchers get more strikeouts, which means less plays in the field.

conclusion: barry bonds is good.
Comments: Post a Comment

Powered by Blogger