Poor Man's Topological Data Analysis

[steveshea]

At the 2012 Sports Analytics Conference, Muthu Alagappan gave a presentation that redefined the positions on the court. The presentation can be found here:

http://www.sloansportsconference.com/?p=5431

At that conference, Muthu’s new list had 13 positions. He reduced that number to 10 for a TedX talk:

http://tedxtalks.ted.com/video/The-new- ... basketball

Alagappan uses topological data analysis (TDA) to redefine the positions and create some really cool images. TDA is complicated, still developing, and fascinating field that studies the shape of a set of data. If you don’t want to or don’t have the time to learn TDA, you can still create 2D visualizations of higher dimensional data.

See our most recent post for a couple examples:

www.BasketballAnalyticsBook.com

We started simple with this post. Our plan is to post progressively more complicated and more interesting images. We will do this by including more advanced stats on the players and by using more involved visualization techniques.

[Crow]

I share a interest in player topologies. There are some good old treads here with data from Ed Kupfer and David Sparks that might interest you. Browse or search if you want to read them.

[steveshea]

I share a interest in player topologies. There are some good old treads here with data from Ed Kupfer and David Sparks that might interest you. Browse or search if you want to read them.

Thanks, I will look for them.

[Crow]

Here are some of the old ones I was thinking about:

viewtopic.php?f=2&t=163&hilit=typology
viewtopic.php?f=2&t=140
viewtopic.php?f=2&t=106
viewtopic.php?f=2&t=116
viewtopic.php?f=2&t=112

I think DSMok1 did some form of clustering here and / or at his site as well.

[fpliii]

Here are some of the old ones I was thinking about:

viewtopic.php?f=2&t=163&hilit=typology
viewtopic.php?f=2&t=140
viewtopic.php?f=2&t=106
viewtopic.php?f=2&t=116
viewtopic.php?f=2&t=112

I think DSMok1 did some form of clustering here and / or at his site as well.

Bit of a bump, and some interesting links. Thanks for compiling the threads.

Anybody have any new thoughts on player clustering/role classification?

[Crow]

Muthu Alagappan 10 main types seem decent. One next step option would be to look at how best to categorize within these types. one simple way would be to look at the 4 quadrants defined by offensive and defensive RPM.

For point guards, one could also look at overlap of his typology with Seth Partnow's recent research at nylon calculus.

[fpliii]

Muthu Alagappan 10 main types seem decent. One next step option would be to look at how best to categorize within these types. one simple way would be to look at the 4 quadrants defined by offensive and defensive RPM.

For point guards, one could also look at overlap of his typology with Seth Partnow's recent research at nylon calculus.

Wasn't aware of the blog, great read. Thanks for the suggestion.

Assuming this is the article you were speaking of:

http://nyloncalculus.com/2014/10/16/put ... uards-box/

There's some very interesting stuff there. Would be very interested in follow-up articles for other positions.

[knarsu3]

I'll have a cluster analysis of Vantage's defensive stats out tomorrow, which will redefine positions defensively.

[fpliii]

I'll have a cluster analysis of Vantage's defensive stats out tomorrow, which will redefine positions defensively.

Awesome. Looking forward to your research, as always.