Against Glop's Grain
Late last night/early this morn, myself and Silverbird threw together this post. It came out of a long conversation that, I'm now realizing, might now have made it into type the way we wanted it to.
The point of this post was not, as some of our feistier readers believe, to deride PER or say "shit is stupid!!!!" In fact, we didn't even pat ourselves on the back for putting the overrating effect in economic terms. Rather, this was intended to bring together two sides of the debate. Like, fine, PER can misrepresent things. But at the same time, it can call our attention to situations demanding attention. We sought to make these problems into something wonderful, not, as some have claimed, point at them and giggle.
"PER Inflation" has not, cannot, and does not mean that "these dudes suck." Instead, it's indicative of a player requiring more minutes, such that a higher caliber of opponent would correct this aberration. If not, and he stays that good, great. In this case, though, PER is qualitative indicator that someone's place in the rotation needs reconsidering. This, of course, is consistent with Ziller's light attack on coaching know-how. Similarly, someone wounded with "deflated" PER tells us either that, according to one viewpoint, he should be considered for demotion. Of course, it could also simply show that his team is barren, or that he has been chosen for a highly-specialized role, one that might not show up all that well in PER.
I don't really like stats, it's true, in part because when I watch a game, I don't see PER and +/-. Ironically, here I was trying to make a peace offering, to unite two warring factions, under the banner of qualitative cues. If I failed, may you all leave tampons in my ear.
posted by Bethlehem Shoals @ 9/10/2007 04:37:00 PM
22 comments
22 Comments:
That there is a disturbing-ass photo.
This comment has been removed by the author.
"I don't really like stats, it's true, in part because when I watch a game, I don't see PER and +/-."
This gets at the heart of everything I've written in the last few days. I've always thought that we should use stats to quantify or explain things that we think we see while watching a game. Ppg, rpg, apg, and even a new stat like eFG% or REB-r all do that. These seem like much more important factors in figuring out how to put together units, which is how basketball is ultimately played well, as a five-person team. (I do think you can see +/-, it's just more along the lines of units instead of individual players. I also think that's more useful given the unquantifiable variables you guys bring up in the post.)
Paul Millsap seems to be a popular example, so I'll use him as my case study. Watching a Jazz game, I know that Millsap is extremely effective in his limited minutes. It would be interesting to see him play more to see if he could keep it up. Does anyone disagree with that? PER simply quantifies that same observation.
There's no problem with quantifying that observation -- something as basic as ppg does the same thing with scorers, and it helps. But what does that number come to mean? Hollinger's general PER/role ranges are usually right, but do those tell me anything that we can't generally agree on without them? If PER quantifies that observation, then it should be useful as a comparison. So if Millsap is .5 productivos (name the units whatever you want) better than another player, what have I gained from that comparison? Maybe he's more productive, but so many independent variables (like role; team style, which is not the same as pace, etc.) play into that comparison that the .5 ultimately seems meaningless. And even if it weren't, how does that .5 explain his on-court performance? Yes, a large difference in PER can point to a large difference, but that shouldn't tell me anything that simple observation or the basic stats don't. So yeah, Paul Millsap and Amir Johnson should get the chance to play more minutes. Didn't we all know that already?
That's my problem with PER and stats like it. They don't tell me anything about how a game or season was played. Call me old-fashioned, but I still prefer my own perceptions of a game/season over everything else. If a stat shows me that I'm wrong, I'd at least like to know why I'm wrong. I guess PER can work to explain general trends in a player's career -- I'm just more interested in knowing why those trends occur.
vomiet: Russian vomit
What's annoying about all of this from the perspective of someone who is very involved in the APBRmetrics movement is that there are many out there who equate PER and APBRmetrics. But from what I've read of most of the leading minds in the field, PER is basically the "necessary evil." As Ziller has pointed out in numerous places, "holy grail" stats (stats that attempt to summarize every part of a player's production) are not encouraged at all in APBRmetrics - in fact that are often discouraged.
Ty, APBRmetrics holds the same exact view you do - use stats to explain why the things you see happen. For example, APBRmetrics rejects PPG in favor of finding out what goes in to PPG - usage and efficiency. PPG is determined by how often a player attempts to score and how well he converts his opportunities. This past year, Tracy McGrady scored with lots of opportunities and low efficiency - Dirk Nowitzki scored an almost identical PPG with less opportunities and higher efficiency. That kind of information is worth knowing.
You get it right at the end when you say: "I guess PER can work to explain general trends in a player's career." That's really why it's useful - to be able to do studies where you see the value of a draft pick, for example.
I just worry that everyone decrying PER turns into people decrying the entire stat movement, when 99% of it is useful. But the stats consumed by the general public are stats like PER, and then people see things like you've seen and begin to throw the baby out with the bathwater. I'm glad you have been very able to distinguish between PER and the general statistics movement. I just lament all those commenters I've read who lump it all together.
Thanks for the clarification, Ben. I should make it clear that when I say "PER and stats like it" I'm talking about the catch-all stats that you take issue with. I have issues with per-minute averages (mostly in terms of extrapolation -- if you're just comparing Dirk and McGrady I agree they're useful -- and team style), but they pale in comparison to the problems I have with the catch-alls. I have the same problem with QB rating, and that one's about as accepted as it gets.
I second MaxwellDemon, at first glance, that picture is FRIGHTENING. For a second, it really looked like they were pulling that kid's bloody lung out through his mouth. Sweet fancy Moses!
I come at this whole discussion from two disparate perspectives.
One, I grew up watching soccer, where fans and writers don't worry about any stat except goals scored, and that's how I tend to approach American sports, i.e. I don't care about Peyton Manning's yardage total, but I do care about his game management.
Then there's a part of me that abides by the Moneyball school of thought, which is all about debunking myths (or at least identifying market inefficiencies) through intelligent stat analysis. We can't quantify everything, but that doesn't mean we shouldn't try.
Ultimately, stats PER, +/- or otherwise, should help answer specific questions. What's always bothered me is the stat being the final answer, which I guess is what Ben alludes to with the "holy grail".
The stats are useful, though, because our eyes more or less lie: Reggie Miller's ugly-ass jumper went in all the time, for instance.
PER inflation reminds me of what (I think) 82games points out about Shareef Abdur-Rahim and why he'll never be on a good team: in order to be effective, Reef needs a lot of shots. But if your offense is structured so's to get him those shots, i.e. Shareef is the best player on your team, well, you're fucked.
Shoals, I think the reason people jump on the validity argument is because if you're looking to tab a guy for promotion or demotion, you'd better know that the stat you're using is really measuring how good he is.
I think the Win Shares people attempt that, and although the APBR people don't necessarily try to equate certain types of production with wins, they do stuff that's even more interesting (to me). For example, they try to measure how production goes down with increased usage, and subtle stuff like when there's an assisted bucket, how much credit does the passer get?
BDBfkaSD, as a person who is new to but enthralled by soccer, one of the things that interests me is the way the commentary flows and the abstract types of things they talk about. Since there are no stats, or even structured plays, you get stuff like, "they just didn't have an understanding that time." I love that the currency of a hard-fought sport can be something as liquid as "understanding".
but here's the thing: PER isn't like WOW, which seems intent on questioning a lot of assumptions about how basketball is played. or at least on installing its own hard and fast ideology. for the most part, PER syncs up with who we know to be "efficient," which most certainly is a subjective judgment you can make while watching games. i don't think there's any question that PER has some validity to it, if nothing else because its WTF moments aren't the norm.
while i've said that "i don't see stats when i watch games," i don't think this means that stats can't support what we see as it happens. if they didn't, then how could they ever realistically challenge it?
shanoff told me once that for him, PER became useful when it conflicted with reality. that's because, since the two were often in harmony, these instances made you sit up and try and find the source of the contrast. sometimes it has to do with a player's style, like iverson; we have already been over how risky scoring has its advantages. here, we're trying to figure out if we can broker a truce between these seemingly "overrated" PER-ites and "real" game.
Off topic, please take a look at this pre-draft Fesenko interview (via a link at Truehoop):
http://youtube.com/watch?v=N2Y95PomScs
Interesting which kind of voice and eloquence he attributes to the stereotypical low-post center. Is that why Duncan and Howard don't want to do center?
oh, and if people hate freedarko so much, take us off your blogrolls.
"that's because, since the two were often in harmony, these instances made you sit up and try and find the source of the contrast."
I do the same thing. But doesn't that become an instance of the game explaining the stat instead of the other way around? I guess it's good that it gets you focusing on specifics of a player's game, but, let's be honest, if you have the time to investigate PER/perception inconsistencies, you were probably doing that anyway. Again, I'm all for using a stat to add to a discussion of basketball; I just have trouble seeing what PER adds unless we're dealing with incredibly broad arguments like the draft value one that Ben talks about above.
I also agree with MaxwellDemon: the "Eye's for the Leaking" old guy with blue/green eyes was the scariest thing I've seen since BOB stepped over the couch in one step on Twin Peaks, and this lung-through-the-mouth picture is the scariest thing I've seen since then.
As a big mathguy/stat nerd, I wish everyone would stop linking to or giving that "Wages of Wins" guy any airtime at all. He just makes the rest of us stat nerds look bad.
Also, Jose Calderon is awesome; the stats don't lie about that.
Oh my God! The Patriots spied on the Jets! It's not like other NFL teams ever do this or anything. Only New England would ever stoop this low.
Seriously, though, this is maybe the biggest overreaction to a non-story ever.
lh--
i was only bringing up berri as an example of someone whose stat-world wants to lord over everyone else's basketball knowledge.
and in keeping with the "overrated by PER, underrated by game" thrust of our first post, jose calderon is most certainly awesome. now he should be given the chance to play like someone awesome should.
tk--
you know what, you're right. i didn't need PER to tell me those players are better than "we" think
Shoals,
I'm a little confused by your comment. Do you mean any of these: a) you agree with me, b) I should stop trying to speak for other people with "we", c) people who like the FD philosophy might see that, but not others. Thanks.
Shoals, for a second I thought you were comparing PER to World of Warcraft, and I didn't know where the conversation was going. Everything got a little uncomfortable at that point.
But back to your statement: I mean, given the time, anyone could build a stat that sort of jives with what we think of as good, but that doesn't mean it's describing anything real. And if it conflicts with one of my presupposed notions, I don't know what that's supposed to teach me.
I'd rather try to use a measure that distills the actual facts of what happens on the court into a more understandable picture of who does what. Maybe this just points to a different type of curiosity. Maybe I don't care so much about knowing who's good, or whether one player is better than the other. But one thing possession-based ratings can help tell you is why one guy is better than another.
tk--
no sarcasm. "we" as in "better than we think," but you and i already know this is the case. the we that nobody knows.
Ah, thanks for clearing that up. In retrospect, it seems obvious.
I'd say that many on the opposing side seem perfectly capable of developing those impressions that we think PER unnecessarily quantifies. In the end, I guess there's no problem with wanting that impression as a quantity as long as everyone recognizes the limitations, which the best commenters seem to do. I'll grant them their biases as long as they grant me mine.
Post a Comment
<< Home