APBRmetrics

[HoopStudies]

One thing we all struggle with are the positions for various players. Some of us call Lebron a SG, others call him a SF. He plays PG at times, but I don't think we call him that.

I personally stay away from position designations too much. But there are two things I would like to take on regarding position designations.

First, if people have lists of what position they think guys play, submit them here so that we can all compare and maybe (but highly doubtfully) come to consensus on them.

Second, I'd like to raise an idea that came up back over on APBR_Analysis (the olden days), which was using a statistical approach to define position. EdK posted at

http://sports.groups.yahoo.com/group/APBR_analysis/message/4664

the full excerpt, which actually referred to another paper. Basically, the authors used statistics to look for patterns in players. They weren't necessarily looking for the classic positions, but looking for groupings of players. They used something called cluster analysis, which I'd never used before, though I had seen Principal Component Analysis, something that supposedly does something similar. Either way, I couldn't quite replicate the original authors' results. But the concept is interesting and perhaps a good project for someone. I have more info if people want to go this way.
_________________
Dean Oliver
Author, Basketball on Paper
http://www.basketballonpaper.com

[Dan Rosenbaum]

I avoided position designations for a long time as well, but finally gave in when my adjusted plus/minus results strongly suggested that the relationship between traditional statistics and adjusted plus/minus ratings differs by position. This was the reason for the strong results for versatitility in earlier versions of my ratings.

And when I use positions, I am tempted every third time to do something like this cluster analysis that would seem less arbitrary than using position designations. But whenever I go down that road, I start to see Vlade Divac popping up in groups with point guards. And while that may make sense in some respect, Vlade Divac is generally guarded by centers and generally guards centers. And ignoring that in doing a cluster analysis seems mistaken to me. Now I suppose there may be some hybrid between the two that would work, but at the end of the day this problem never completely goes away.

[kjb]

There are going to have to be exceptions made for atypical teams, such as the Wizards. The Wizards coaches do number their players 1-5, but they don't have the standard PG, SG, SF, PF, C meanings. On offense, there are guards, forwards and a center (or in Eddie Jordan's parlance -- bigs, littles and mediums).

The two guard slots are virtually interchangeable, as are the 2 forward spots. This is particularly evident in the backcourt when Arenas and Hughes are both healthy. They share ball handling almost 50-50, and their position designation (1 or 2) can change from possession to possession depending on what play is called and what side of the floor the guy is on. Defensively, those two often cross-match with Arenas guarding SGs and Hughes taking the PG. Plus, they play zones, etc., etc.

Of course, Eddie likes to add confusion to the situation by sometimes referring to Gilbert as a PG or Larry as a SG, even though such designations have almost no meaning within the system they're playing. And, he'll refer to Kwame as a "big" when Kwame plays mostly at forward when healthy.

[Nikos]

What exactly is their record since Hughes went down? How has their offense and defense faired since his injury?

[kjb]

[Kevin Pelton]

Almost precisely a year ago, we had a discussion at Hoopsworld about player positions where we essentially settled on looking at analysis of statistical skills, the thinking being that your skills define what positions you can play and, if you're interested in that sort of thing, a most appropriate position (statistically).

I like to call this analysis "the table," as in, "What does this guy bring to the table?" I define a strength in the system as one of the 10 categories where the player is at least a standard deviation amongst NBA average (amongst guys over a minutes threshold). You can do weakness the opposite way, but I don't use the weaknesses to look at positions.

So what I did is pull out this spreadsheet from midway through last year and look at how the strengths were distributed by position. (Positions are as assigned by Doug Steele):

[NickS]

[NickS]

Thinking a little more about KP's table, I think the question becomes, "are there any roles that every team wants to fill with 1 or 2 people, and which roles can be filled by committee?)

It's the conventional wisdom that every team would like to have one player who is at least 1 SD above average in ballhandling (however you measure that) and that it just isn't as good to get by with 3 or 4 players who are above average but not exceptional (with a notable exception, the various Phil Jackson coached championship teams did not have a "PG" as such). Similarly there's talk about the value of having a shot blocker, implying that there's a difference between a C/PF pair that gets 1.4/1 bpg and another pair in which the blocks are divided 2.1/.3.

On the other hand, people talk go back and forth about the value of having one "go-to" scorer or having "balanced" scoring. Similarly nobody cringes too much at the idea of "rebounding by committee" when it works.

I guess this would make me think that positions are over-rated. If you can think of an exception for every theory that would justify the idea of having specific positions it makes the theories seem suspect.

At the same time I think positiions do matter on the defensive end. As Dan points out Vlade Divac is going to have a hard time guarding a PG and Kirilenko may be leading the league in blocks but that doesn't make him a center.

I remember being struck by a comment in Basketball on Paper to the effect that the reason that there are so many bad 7-footers in the league is because people hope that they can use them to defend another 7-footer . The stiffs aren't playing because of their offense.

This has come up talking about Radmanovic's role on the Sonics. Calling him a SF or a PF doesn't matter so much as figuring out where he belongs on defense. His role on offense won't change, but sometimes he looks good guarding players in the post, sometimes on the perimeter, and sometimes neither.

But can you clearly define the difference between SG/SF/PF?

A related thought, I remember an ESPN about the Rockets struggles early in the season (before Sura took over the point) that said part of the problem was that other than T-Mac they lacked athleticism. That's exactly the sort of comment that, if true, seems difficult to quantify.

I know, for example, that I've never like Tony Parker from a stats perspective (poor A/To and FG% -- though he's been playing better) but I would have to admit that watching them play, his quickness is an important factor in the SA offense. Having, say, Brent Barry run the point would probably not be an improvement.

[Nikos]

Another thing that suprises me is that the Wizards have improved so much Kev, Jamison was a nice addition but just take a look at how much Arenas, and Hughes win % have jumped. How could this be? Have they really improved that much, or does Arenas improving and being healthy, and Hughes being healthy, and Haywood and others contributing, and Antawns addition really make the team that much better?

[Kevin Pelton]

[kjb]

[Ben]

[Dan Rosenbaum]

[gabefarkas]

here's my guess:

1 turnover percentage
2 blocks per 48 minutes
3 two-point %
4 assists per 48 minutes
5 steals per 48 minutes
6 offensive rebounds per 48 minutes
7 points per 48 minutes
8 defensive rebounds per 48 minutes
9 free throw attempts as a percentage of possessions
10 three-point %


are you going to post yours soon, Kev?

[Kevin Pelton]