Saturday, September 08, 2007

People Hate Statistics

And I'll probably never figure out why. Tom Ziller shouldn't have to defend PER at Ballhype, but he does anyway. Baseball has been numerically sliced and diced to the point that refuting SABRmetric evidence as nonsense will earn you a significant internet pounding. Why are APBRmetrics constantly open to ridicule? The beauty of PER is that it defines average production, and allows one to compare players with a shorthand number. It has a utility similar to baseball's OPS+.

Owning an Odom jersey doesn't make you right, and goofy statements like "a player's height should be considered when looking at rebounds" aren't going to lead to rational conclusions. You know what we should look at when considering a player's ability to rebound? REB-r.

The trick is using statistics the right way, not questioning and criticizing a statistic because it doesn't do everything you think it should.


Blogger BenGo07 said...

Presumably, if you could actually refute SABRmetrics as nonsense, you'd be in a much better, respectable position, objectively speaking. But I agree, people who dismiss SABRmetrics as nonsense are assholes.

6:38 PM  
Anonymous Jason said...

luolcats > statistics !!!

(I'm pretty sure the height-rebounding claim wasn't serious--see the video to which it was linked.)

6:52 PM  
Blogger Ty Keenan said...

I didn't write the post you dismiss here--my cowriter did, but his job made him go in today--but I want to clear up a misconception about our point. New stats like rebounding rate and eFG% are good (we don't only like the old stats, that would be stupid). They're good because they don't divorce the stats from the context in which these numbers occur and tell something tangible about on-court production. All that's possible while still recognizing the limitations. I like OPS+ for the same reasons.

Our issue with PER is that it throws everything together. I don't think many people know what the number signifies (what is the real difference between a PER of 22 and one of 21?) other than that having a PER above 15 is good (and I actually like that it's always normalized around the same number). What does a difference of 1 PER point signify in terms of on-court performance? Every time I look at PER, I have to look at the stats that go into it to make sense of it. If I knew the limitations of the stat itself, that wouldn't be necessary, and if you can explain them I'd be happy to listen. As of now, all I know is that it seems wrong a lot of the time.

7:58 PM  
Anonymous Matt said...

"all I know is that it seems wrong a lot of the time."

I don't doubt that's all you know, and dismissing it as 'wrong' or 'right' shows the issue you're having with interpreting and understanding the figure.

Your opinion seems conflicted because you show willingness to learn while simultaneously not bothering to read up more on PER. If you can't find it online pick up one of Hollinger's old season prospectus (R.I.P.).

You're right, looking at the component numbers is more thorough. Like HSCS said, PER is shorthand for all those factors together. And it's not perfect, and Hollinger (and reasonable people) know and say as much.

8:14 PM  
Blogger Ty Keenan said...

Matt: Thanks for the response. I've read the explanation of how they compute PER and the explanation of the general PER classifications at alleyoop. It was a poor choice of words by me to say "wrong"--I should have said "off" or some other less severe word.

The issue for me is that when you get to comparing fairly similar players (which is how I would use a catch-all stat, to compare players who're deemed basic equals by every other definition), the PER is so close that I don't know what the difference really is. Maybe I should trust the stat more, but it's difficult to do that when I don't know its exact limitations. So I invariably turn to the stats that go into it, which I can understand and criticize. Can you do that with PER? (That's not only meant to be a challenge to your POV--I legitimately want to know if anyone can do that. I'd like to learn how.)

The only specific limitation I can think of with PER is that it combines a number of different stats. PER's definitely not useless--as a writer, it's great if you want to show that one player's generally good in a few words. It's the best basketball stat of its kind, for sure.

You're right, reasonable people realize that no stat is perfect, but there's something to be said for the ways in which each isn't perfect. PER's limitations seem to be products of how its constructed as opposed to things that can't be quantified or explained by another stat.

10:04 PM  
Blogger Carter Blanchard said...

Woah there. First off, I don't own an Odom jersey. Van Exel is all ever I'll ever rock.

Second, I don't hate statistics. I really don't. I know it was a long post, but way down at the bottom I even admitted as much ("When trying to evaluate a player, 82games or Basketball Reference is almost always my first stop"). I wasn't trying to blindly ridicule APBRmetrics. I just think PER tends to obscure too much of the story.

As you said, the problem's not with the stat but how it's used. If you want a really quick shorthand to use for quick comparisons, by all means, PER's probably the best option out there. But if you want to analyze a player's performance, or make an argument about someone's relative worth (which is how I often see it used), I'd much rather see an entire stat line (including cool shit like REB-r and eFG%). That's all I'm saying.

12:45 AM  
Blogger Hot Shit College Student said...

Carter, I read your entire post, even the bottom. The headline is something I type whenever someone complains about statistics. Hence, "people." It isn't directly related to the content in the post.

What I still don't get, is why you're hellbent on dismissing PER because you've found that it's often misused. Why not have a beef with the people who are wrong? Throwing everyone who uses PER on the flames was beyond not cool.

10:03 AM  
Blogger Carter Blanchard said...

Where do you get that I'm completely dismissing PER and all who use it? I've admitted that I used it in the past and will continue to do so in the future. I'm not sure how you equate, "thinks PER has some drawbacks worth exploring" to "thinks the Italian mothers of those who use PER should be [expletived] in their [expletives for lady parts]."

3:05 AM  
Blogger Hot Shit College Student said...

Yeesh. Quit being a big know-it-all.

You keep repeating the same lame points about PER's drawbacks that have been explored before, and in depth, by ya' know, people who know what they're talking about. This is hardly a new topic worth exploring.

It is insulting that you're claiming people who use PER misuse PER, when you clearly don't have a grasp on the thing in the first place. Odom has been declining, and his unspectacular +/- rating from last season is seemingly worse than a slightly above average PER. Get over it. Above average isn't a bad thing either. At least Dave Berri has his own dubious metric to disprove PER with.

8:45 AM  
Anonymous Anonymous said...

awesome how Carter and Ty are totally milking this. who heard of their website before? maybe I'll go question OPS and get a million hits.

10:19 AM  
Anonymous Matt, your friendly Bulls Blogger said...

That is an interesting point. Like the original premise of this post has said, we should be at the point where a 'what's the deal with PER?' post is simply dismissed, instead of writing essays to point out the seemingly obvious.

Just wait everyone, Hollinger's PER rankings are technically out of the bag, so time for a lot of 'how can this guy be ranked higher than that guy! LOL!' posts. (LOL in this case standing for 'Lamar Odom Love')

10:45 AM  
Blogger Hot Shit College Student said...

Anony - It is getting borderline disingenuous, but I doubt that much traffic is getting drummed up for a discussion (ooh, Miss Gossip doesn't trust PER!) on rate based, and pace adjusted stats. I know I keep hitting refresh, but something more exciting will come along eventually. Isn't First Take on soon?

10:52 AM  
Blogger Doctor Dribbles said...

HSCS, you I can understand. (FreeDarko, not always so much). Nice, quick explanation and dismissal.

2:32 PM  
Anonymous Anonymous said...

Instead of complaining about PER, why not use a better statistic like Dave Berri at the Wages of Wins, Wins Produced per 48. Lamar Odom was rated a .210 in 05-06, which means he was twice as good as the .100 average, and .164 this year, in a year in which he struggled with injuries. Boykins was rated at .071, or below average. Does this make sense?

I don't think the issue here is that per minute stats don't work, or that the idea of a composite stat is foolhardy. The problem is with PER and specifically with the way it overvalues scoring. High scoring players like Boykins end up looking as good as versatile players like Odom.

I am in the middle of a discussion about PER over at APBR right now. Specifically, about the fact that Dennis Rodman has a below average PER for his career, and that in his best year, 91-92, his PER was only a 17, which is what Eddy Curry posted last year.

Anyway, try the Wages of Wins and see if it works better for you.

7:40 PM  
Anonymous Anonymous said...

Berri's PAWS/min is pretty good for evaluating bigs (though it looks shot defense as many metrics do) but I don't think it is a good tool for perimeter players.

4:52 PM  
Anonymous Anonymous said...

should be "lacks" not looks in above post

4:53 PM  
Anonymous Anonymous said...

Yes exactly, in some moments I can bruit about that I agree with you, but you may be making allowance for other options.
to the article there is quiet a without question as you did in the go over like a lead balloon a fall in love with issue of this solicitation professional 7.004 ?
I noticed the phrase you have in the offing not used. Or you functioning the black methods of development of the resource. I take a week and do necheg

2:41 PM  
Anonymous Anonymous said...

WtpVth [url=]Moncler Giubbotti[/url] FquTth

XybCzk [url=]Nike Free 3.0[/url] GakLhb

ChdXtz [url=]outlet Moncler[/url] ItsFls

KiwJnb [url=]Canada Goose Parka[/url] QkwCkx

VwbHuf [url=]Canada Goose Parka[/url] BonEio

QzxZcs [url=]Canada Goose Jakker[/url] WgwJtm

RwcAky [url=]Canada Goose Jakker[/url] EanRec

8:18 PM  

Post a Comment

<< Home