APBRmetrics

The discussion of the analysis of basketball through objective evidence, especially basketball statistics.
It is currently Tue Sep 02, 2014 11:31 am

All times are UTC




Post new topic Reply to topic  [ 1 post ] 
Author Message
 Post subject: Player typology (2005)
PostPosted: Wed Apr 20, 2011 4:02 am 
Offline

Joined: Thu Apr 14, 2011 11:10 pm
Posts: 2311
Author Message
jambalaya



Joined: 30 Jan 2005
Posts: 282


PostPosted: Mon Apr 04, 2005 5:35 pm Post subject: player typology Reply with quote
has anyone constructed a detailed player typology using segmentation, cluster or factor analysis, etc.?

i was thinking about the examples of claritas inc.'s groups for neighborhoods and households, the myers briggs personality types, the cluster analysis of voter types and thought it might be intersting to see or try to construct one for the NBA with some mix of position, age, stats, skill sets, style of play, salary/contract length and possibly other attributes.

Last edited by jambalaya on Mon Apr 04, 2005 6:31 pm; edited 6 times in total
Back to top
View user's profile Send private message
Ed Küpfer



Joined: 30 Dec 2004
Posts: 773
Location: Toronto

PostPosted: Mon Apr 04, 2005 5:42 pm Post subject: Reply with quote
It would be interesting, but considering the trouble we have getting accurate data on player weights, I don't think we'll ever be able to get personality measures.
_________________
ed
Back to top
View user's profile Send private message Send e-mail
jambalaya



Joined: 30 Jan 2005
Posts: 282


PostPosted: Mon Apr 04, 2005 5:47 pm Post subject: clarification Reply with quote
though my examples have a fairly heavy emphasis on personality, i wasnt focused on that with respect to the nba. i was focused on their game and the attributes that help define it.

the article about the blazer draft system seems to have a strong emphasis on personality. i assume it was to try to reassure blazers fans that the year old plan to try to draft just good character guys will continue and deepen that the system/mystery guy story slipped or was purposedly leaked.

on personality inclusion is a typology of those in the league perhaps you could include such things as role on team, teammate type, streaky vs. calm, clutch vs. fade, degree of professionalism, playoff experience, quality of pre-nba coaching, considered selfish by other players or respected as team-oriented, flexible, coachable vs. plays his own way regardless, been a star or the man in the past vs. always a role player, stays in offense vs. creative, basketball I.Q./ real-time thinker, etc.

my initial thought is 3-7 types per position, then subgroups based on quality and perhaps other criteria.

Last edited by jambalaya on Tue Apr 05, 2005 10:39 am; edited 9 times in total
Back to top
View user's profile Send private message
jambalaya



Joined: 30 Jan 2005
Posts: 282


PostPosted: Mon Apr 04, 2005 6:00 pm Post subject: initial brainstorm of types Reply with quote
to give a better idea of what i am reaching for, here is my initial list of major types:

Pure point

Drive and kick

Shooting point

Tweener

Backup point

Veteran stabilizer



Classic 2

Penetrator 2

Swingman

Strong defender 2

Spot up shooter

Playmaker 2



Shooting 3

Tall 3

Post 3

Defender 3

Point 3

Power forward 3



All round power forward

Rugged Rebounder

Post defender

Quick jumper

Shooting power forward



Post scorer

Center defender

Lane clogger

High post

Shooting 5

Shot blocker

Project


the subgroups might get much more specific and perhaps have colorful descriptive names. and you could have hybrids.

it could also go in other directions.

Last edited by jambalaya on Tue Apr 05, 2005 10:12 am; edited 1 time in total
Back to top
View user's profile Send private message
Ed Küpfer



Joined: 30 Dec 2004
Posts: 773
Location: Toronto

PostPosted: Mon Apr 04, 2005 6:23 pm Post subject: Re: initial brainstorm of types Reply with quote
jambalaya wrote:
to give a better idea of what i am reaching for, here is my initial list of major types:

Pure point
Drive and kick
Shooting point
Tweener
Backup point
Veteran stabilizer
[...]
the subgroups might get more much more specific and perhaps have colorful descriptive names.


Ah. What you're proposing has in fact been tried before. I can't find the cite right now, but someone used some complicated algorithm to determine player "roles" (bangers, walls, dishers, etc). They came up with 7 or 8 player types. I couldn't duplicate the results, but the idea of classifying players into roles is a good idea.

One way to do it would be to make an exaustive list of player attributes (passing, scoring, rebounding, etc), and score each player on those attributes. The player's classification would be a result of all those scores. This approach would work best if each player would begin with a "zero" score across the board. That is, that our prior knowledge of the player's position or passing ability or whatever would not inform the scores he gets for each attribtue -- so that, in theory, we could get a 7'2 "disher" or something.

This is doable to some extent. I'll see what I can hack up.
_________________
ed
Back to top
View user's profile Send private message Send e-mail
jambalaya



Joined: 30 Jan 2005
Posts: 282


PostPosted: Mon Apr 04, 2005 6:37 pm Post subject: followup Reply with quote
ok thanks for the reply. look forward to hearing more.

you're right it could be organized on factor analysis of major skills instead of being segmented right off the top by position i showed above. many possible approaches.

7-8 groups in that past study isnt a lot but if you find it i'd be interested in seeing it. i was thinking somewhere around 25 main groups might be ideal balance point for classifying a league of 400+ guys. but just getting started on thinking about the topic today.
Back to top
View user's profile Send private message
kjb



Joined: 03 Jan 2005
Posts: 855
Location: Washington, DC

PostPosted: Tue Apr 05, 2005 4:38 am Post subject: Reply with quote
I seem to recall Kevin Pelton doing something like this -- using stats to identify position profiles. I think he came up with 12 or 13 "types" but my memory could be faulty here. Kevin...

Smile

I definitely think there's a real value in identifying roles necessary for a team, and coming up with a way to evaluate how well guys fill those roles. Sorta like what Seattle seems to have done this season.
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
gabefarkas



Joined: 31 Dec 2004
Posts: 1292
Location: Durham, NC

PostPosted: Tue Apr 05, 2005 8:14 am Post subject: Reply with quote
I think superstar players would supercede some of the classifications. The first example that comes to mind is Shaq -- which C role would you call him? I guess Post Scorer, but does that really cover it?

In general, I like this idea. I used to be really into Myers-Briggs and Keirsey stuff, so I definitely see where you're coming from.
Back to top
View user's profile Send private message Send e-mail AIM Address
kjb



Joined: 03 Jan 2005
Posts: 855
Location: Washington, DC

PostPosted: Tue Apr 05, 2005 8:31 am Post subject: Reply with quote
The superstar player may be a category unto itself, perhaps with a few minor gradations.
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
Kevin Pelton
Site Admin


Joined: 30 Dec 2004
Posts: 969
Location: Seattle

PostPosted: Tue Apr 05, 2005 9:54 am Post subject: Reply with quote
WizardsKev wrote:
I seem to recall Kevin Pelton doing something like this -- using stats to identify position profiles. I think he came up with 12 or 13 "types" but my memory could be faulty here. Kevin...

I came up with statistical skills and identified which of them players had, but I never grouped them. It's subtly different, but frankly, I don't find the concept of groups particularly useful. There are certain skills teams need to have filled, but whether their shooting specialist also rebounds or passes or plays defense doesn't really seem to matter to me. *shurg*
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Ed Küpfer



Joined: 30 Dec 2004
Posts: 773
Location: Toronto

PostPosted: Tue Apr 05, 2005 1:02 pm Post subject: Reply with quote
admin wrote:
I came up with statistical skills and identified which of them players had, but I never grouped them. It's subtly different, but frankly, I don't find the concept of groups particularly useful. There are certain skills teams need to have filled, but whether their shooting specialist also rebounds or passes or plays defense doesn't really seem to matter to me. *shurg*

It matters to me. I want to do away with traditional positional categories. Shaquille O'Neal and Rafael Araujo and Dikembe Mutombo do not play the same position, regardless of what it says on the scoresheet. Or rather, they don't play the same role, which I think is more important.
_________________
ed
Back to top
View user's profile Send private message Send e-mail
jambalaya



Joined: 30 Jan 2005
Posts: 282


PostPosted: Tue Apr 05, 2005 1:41 pm Post subject: address and zip +4 template Reply with quote
last night i thought about a player typology like an address

address:

street
city state zip +4


nba player typology:

street address = salary
such as $400k minimum court, or $1.6m small exception lane, or $5m middle class circle or $12m star island

city= type of player for their position (using categories developed in my first post or a small set based on skill mix)


state= just the position (or it could be team i guess to indicate style of play))

zip= 5 offensive categories, 0-9 like ed suggested. maybe some scoring measure, a shooting efficency measures, FTs made, passing ability, turnovers

+4 digits= def rebounding, def. disruptions (blocks, steals), opp scoring or shooting percentage, and maybe tuck off. rebounding in here too.

that would be pretty comprehensive.

Last edited by jambalaya on Tue Apr 05, 2005 3:16 pm; edited 3 times in total
Back to top
View user's profile Send private message
Ed Küpfer



Joined: 30 Dec 2004
Posts: 773
Location: Toronto

PostPosted: Tue Apr 05, 2005 1:48 pm Post subject: Re: followup Reply with quote
jambalaya wrote:
you're right it could be organized on factor analysis of major skills instead of being segmented right off the top by position i showed above.


Maybe you can help me get started. Let's separate offense and defense to begin with, and go from there. Skills to fill:

Offense

* Passing
* Shooting
* Rebounding
* Free Throw Shooting
* Picks and Screens



Defense

* On the Ball (I don't know what to call this)
* Help
* Steals
* Rebounding



We can include subcategories:

Offense

* Passing
o Assists
o Turnovers
* Shooting
o 3-pointers
o Midrange
o inside
* Rebounding
* Free Throw Shooting
o FT%
o FTA
* Picks and Screens


Defense

* On the Ball
o Opp shooting%
o Opp FTA
* Help
o Opp shooting%
o Opp FTA
o Blocks
* Steals
o Individual Steals
o Opp TOs while player is on the floor
* Rebounding
o Individual Reb
o Team Reb while player is on the floor



Anything else?

Some of these would be hard, if not impossible, to score. What I'm after right now is an exaustive list of things to look for -- a place to start.

jambalaya wrote:
7-8 groups in that past study isnt a lot but if you find it i'd be interested in seeing it. i was thinking somewhere around 25 main groups might be ideal balance point for classifying a league of 400+ guys. but just getting started on thinking about the topic today.

As you can see above, a preliminary list of about 10 categories can be subdivided almost infinitely. What I envision is not cubbyholes to place individual players, but a player score for each category: for example, Chris Bosh would get a score like this on defense:

Code:

On the Ball 5
Opp shooting% 4
Opp FTA 7

Help 8
Opp shooting% 6
Opp FTA 9
Blocks 7

Steals 3
Individual Stl 2
Opp TOs 3

Rebounding 4
Individual Reb 3
Team Reb 4


With the scores being out of ten. All numbers made up on the spot as an example.
_________________
ed
Back to top
View user's profile Send private message Send e-mail
jambalaya



Joined: 30 Jan 2005
Posts: 282


PostPosted: Tue Apr 05, 2005 1:59 pm Post subject: categories Reply with quote
i think you have the main ones.

maybe fouls committed. minutes (and perhaps games) played.
maybe crunch FG%. probably not %own shot blocked.
i wondered about adding PER or off and def. retgs.
i wouldnt use +/- for this i dont think.

if you have picks, i think picks that lead to baskets are as valuable as assists. and i'd probably focus on just them rather than all picks and screens (perhaps that is what you meant for this?).

you said: "What I envision is not cubbyholes to place individual players, but a player score for each category"

i hear you, in my zip+4 analogy that part is similar. and of course it doesnt have to be just 9 but trying to constrain the number seems wise because otherwise it become the full database again, just with standardized scores. but that would be fine too i guess. it just wouldnt be a simplifying typology anymore. but folks could make various typologies from it as they wish.


salary could be group this way too with a little fitting using ranges ($1.6-$3 million, $3-5million, etc.) . size (maybe group 6"2-6'3 guys together and 6'-4"- 6'-5" together and so on). age, years of experience could be clustered 1 to 9 for simplicity of viewing with a little finesse.


i noticed usatoday released their under and over paid players based on performance recently. with ordinal league stat ranks.
Back to top
View user's profile Send private message
Kevin Pelton
Site Admin


Joined: 30 Dec 2004
Posts: 969
Location: Seattle

PostPosted: Tue Apr 05, 2005 9:12 pm Post subject: Reply with quote
Ed Küpfer wrote:
Shaquille O'Neal and Rafael Araujo and Dikembe Mutombo do not play the same position, regardless of what it says on the scoresheet. Or rather, they don't play the same role, which I think is more important.

They all defend opposing centers, which to me trumps anything else. Who you can guard is more important than how well you can guard; you need people who can defend each position and certain offensive skills, but it doesn't really matter which of the defenders has which skills.

But I should point out I'm against typology as a blueprint for team-building. Fundamentally, all you're doing is labeling similarity sets, and I'm a huge fan of using similarity scores.

Author Message
jambalaya



Joined: 30 Jan 2005
Posts: 282


PostPosted: Tue Apr 05, 2005 11:49 pm Post subject: on positions Reply with quote
[quote="admin
you need people who can defend each position and certain offensive skills, but it doesn't really matter which of the defenders has which skills.
[/quote]


i agree that position matters more to provide a balanced defense.

or your latter point i think there is room for debate.

you can move people anywhere you want on offense to use their skills however they are arrayed but i still think on balance over a game and over 82 games, who has those skills matters.

it matters that rashard lewis and vlade radmanovic are so effective from three point land from the 3 and 4 spots and that is less common. it mattered that gary payton could post up at pg. it matters than vlade and brad miller can pass so well from the 5. these example could be used as arguments that position doesnt matter in support of your argument. or it could be said the unusual skills are more effective for these guys because of their position. that is simply a difference of perspective and which side you are on of the debate.

but going against the natural pattern for a position can also have limitations. it doesnt always work, it cant always to be made to work as much as you might want to make it work. i still think position matters in part because offense turns into a liveball up for grabs and defense quickly. an above average offensive rebounder from the 2 spot isnt going to have as much impact as an interior guy. quentin richardson has to help cover the backcourt so he cant always crash the boards. i dont think a good 18-21 foot jumpshooting center like drobnjak is as valuable as a good shooting swingman if it costs you offensive rebounds and putbacks. sometimes you can compensate but that is more work and can have other impacts you have to chase down and tame.

it is not all or nothing, there is flexibility. team can twist the roles a bit here and there, between the 2 and the 3 as with mason and lewis, or between the guys on the front line or between a point three and the backcourt. when two guys have reverse skills to switch roles great, interesting, maybe an edge. but looking across 30 teams that arent that many big twists. it seems like position patterns are pretty strong and i dont think it is by accident or cowardice or lack of trying. height and training have a lot to do with it. coaches of the highest levels of competiton reached and reached for exceptional physical attributes- size, stength, mobility, speed and the skills to use these physical gifts to maximum advantage from specific spots. these specific roles and the player types that filled them create what we know as positions today.

i am not arguing for a rigid positon system by any means. i like guys who can play more than one position and have up or down the line additonal skills beyond their traditional postiton. i just dont agree with dismissing its relevance as fully as you have for offense.
Back to top
View user's profile Send private message
gabefarkas



Joined: 31 Dec 2004
Posts: 1313
Location: Durham, NC

PostPosted: Wed Apr 06, 2005 8:27 am Post subject: Reply with quote
admin wrote:
Ed Küpfer wrote:
Shaquille O'Neal and Rafael Araujo and Dikembe Mutombo do not play the same position, regardless of what it says on the scoresheet. Or rather, they don't play the same role, which I think is more important.

They all defend opposing centers, which to me trumps anything else. Who you can guard is more important than how well you can guard; you need people who can defend each position and certain offensive skills, but it doesn't really matter which of the defenders has which skills.

But I should point out I'm against typology as a blueprint for team-building. Fundamentally, all you're doing is labeling similarity sets, and I'm a huge fan of using similarity scores.


that's almost like a recursive argument. asking who you guard neccesarily begs the question "who guards you?" and then, it seems like the answer to one will give you the answer to the other.
Back to top
View user's profile Send private message Send e-mail AIM Address
jambalaya



Joined: 30 Jan 2005
Posts: 282


PostPosted: Wed Apr 13, 2005 1:05 pm Post subject: player typology organized by plotting graphs Reply with quote
having recently seen ed's graphs on offensive/defensive rebounds by size, i thought about how you could use graphs to group players by type and position.

i'd propose doing it somewhat differently by each position.

for point guards i would plot the data for assists against points.
those closest to the plot line would be considered typical modern point guards. those higher on both points and assists would be the star guards.
those higher on assist but lower than the line on points would be considered more traditional "pure" points. those higher on points but lower on assists would be the shooting points, the tweeners and the bowling ball scorers. those lower on both would presumably be the backups, veteran stablizers, defenders.

for shooting guards, i think i would plot outside points + assists against inside points (inlcuding most or all of FTs) +rebounds to help distinguish perimeter contributions from interior contributions. those on or near the plot line would be modern shooting guards. those higher on both would be the star all-around shooting guards and the high productivity wings. those higher on perimeter contributions would be more traditional perimeter 2s, those stronger than usual on the interior but lower on the perimeter score would be the rugged inside biased penetrating 2s. those lower than the line on both would generally be the weaker shooting guards.

for small forward, you could either do it the same way as shooting guard or perhaps a slight adjustment- outside points +2 * assists (to perhaps better balance the stratification of where they fall on the assist /perhaps higher rebound contributions mix) against inside points +rebounds. you would identify the star small forwards, the shooting 3s, the mini-power forwards, and the also rans. (or alternatively you could also look at just perimeter versus inside scoring, since scoring is often the dominant role for this position. or you could plot total points against assists +rebounds to investigate the scoring/teamwork contribution mix.)

for both power forwards and centers i think i would plot points against rebounds + 2 * blocks. a secondary sort of inside/outside points could separate the faceup shooting 4 & 5s from the post players.

these distributions are mainly aimed at helping recognize the group of similar players who could be called a type and though the graphs hint at key productivity elements it is admittedly not a comprehensive or sufficent measure of quality.


ed, if you ever feel like generating some more graphs in this vein, i, for one, would be quite interested. if not, i might try it later.

Last edited by jambalaya on Wed Apr 13, 2005 4:35 pm; edited 1 time in total
Back to top
View user's profile Send private message
Kevin Pelton
Site Admin


Joined: 30 Dec 2004
Posts: 979
Location: Seattle

PostPosted: Wed Apr 13, 2005 1:49 pm Post subject: Reply with quote
Here is an effort at that kind of mapping:
http://www.sonicscentral.com/perceptualmap.jpg
Back to top
View user's profile Send private message Send e-mail Visit poster's website
HoopStudies



Joined: 30 Dec 2004
Posts: 706
Location: Near Philadelphia, PA

PostPosted: Wed Apr 13, 2005 2:25 pm Post subject: Reply with quote
admin wrote:
Here is an effort at that kind of mapping:
http://www.sonicscentral.com/perceptualmap.jpg


Love the photos...

Given that we all have qualms about what assists actually represent, it seems that we should actually map something that isn't just based on assists, which I think this is.

As a suggestion, BobC has his pass ratings, shot ratings, turnover ratings, etc. Plot those, perhaps not as percentages of their touches, but as total touches compared to team touches or team possessions. That involves Bob making the data a little more accessible...
_________________
Dean Oliver
Author, Basketball on Paper
The postings are my own & don't necess represent positions, strategies or opinions of employers.
Back to top
View user's profile Send private message Visit poster's website
gabefarkas



Joined: 31 Dec 2004
Posts: 1313
Location: Durham, NC

PostPosted: Wed Apr 13, 2005 6:54 pm Post subject: Reply with quote
admin wrote:
Here is an effort at that kind of mapping:
http://www.sonicscentral.com/perceptualmap.jpg


why does Mike Bibby always look like MiniMe?
Back to top
View user's profile Send private message Send e-mail AIM Address
Kevin Pelton
Site Admin


Joined: 30 Dec 2004
Posts: 979
Location: Seattle

PostPosted: Thu Apr 14, 2005 1:13 am Post subject: Reply with quote
I can't answer Gabe's question, but, for the record, the axes represent my scoring rating (basically TS%*P48) and passing rating ((AST/MIN)^2*(AST/TO)).
Back to top
View user's profile Send private message Send e-mail Visit poster's website
mtamada



Joined: 28 Jan 2005
Posts: 377


PostPosted: Mon Apr 18, 2005 1:58 am Post subject: Re: followup Reply with quote
jambalaya wrote:

you're right it could be organized on factor analysis of major skills instead of being segmented right off the top by position i showed above. many possible approaches.


Many many years ago I did a principle components analysis (similar to factor analysis, but less flexible) of NBA statistics. I did not group the players by position, I looked at all NBA players. I looked at all of the standard counting statistics on both a per game basis and per-48 minute basis, but I looked at only a few ratio statistics (I think FG% and FT% were the only two that I used).

The first principle component, as one might guess, was basically just a measure of "who did stuff": some players got lots of points, assists, rebounds, FTAs, etc. and some got few. The ones at the top end of the scale tended to be superstars, the ones at the bottom tended to be scrub players.
The second principle component was clearly reflecting height and/or inside playing ability; the players with lots of rebounds and blocked shots were at one end of the scale, the ones with few at the other end. High FG% may've been here too (as well as in the first principle component) because inside players tend to have higher FG%s.

The third principle component may've been one that reflected assists and maybe other small-player skills, but I forget. It was around here that the principle components started getting hard to interpret (this is where factor analysis could have an advantage, in terms of being able to rotate the data to find interpretable factors, but I didn't do a factor analysis).

Cluster analysis could be useful too, although it's a bit dependent upon clusters actually exisiting. E.g. the "playmaking center" (I'm thinking of Bill Walton, Tom Boerwinkle, Alvan Adams, Sam Lacey, Cowens, Sikma, Unseld, etc. etc. and Bill Russell and Wilt to some extent) is pretty well extinct today except for Divac (for that matter, the "superstar center" is basically extinct today except for Shaq. And Duncan if one chooses to call him a center.)

So I don't know if a cluster analysis would be able to identify the Divac/playmaking center cluster, given that it has only one member these days, who probably wouldn't even be identifiable in the 2005 statistics given that Divac missed most of this season.

<A digression on a related point> The most impressive playmaking move by a center that I've seen this season, not counting Chris Kaman's remarkable behind-the-back dribble and behind-the-back-pass for the assist while leading a Clipper fast break: Lakers on defense, a ball gets loose, Divac gets his hand [not hands, just one hand] on it and without even really gaining possession immediately flings it downcourt for a perfect outlet pass to a streaking Luke Walton for the layin. </diregression>

That's about all I did. My main objective in the analysis was not to do an overall typology of players, but instead to see how well discriminant analysis could be used to identify all-star players vs. non-all-stars.

Morover, did I say I did this many years ago? To give you an idea how long ago: there was no 3-point line back then. And I did the analysis using SPSS -- but on computer punch cards.

But I got an A on the project, and from a big-shot statistics prof to boot. Wink


Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 1 post ] 

All times are UTC


Who is online

Users browsing this forum: Google [Bot], Yahoo [Bot] and 6 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum

Search for:
Jump to:  
Powered by phpBB® Forum Software © phpBB Group