SSAC research paper posters posted

[steveshea]

Another 8 research papers are listed at SSAC website.

http://www.sloansportsconference.com/?page_id=462

Three of these are on basketball. One is on refs and the rate at which they call certain violations. Another is on quantifying shot quality in the NBA (they introduce a new EFG% which accounts for the difficulty of the 2-pointer). The last is "How to get an open shot." This last one looks interesting, but I couldn't find a link to the full paper.

[Crow]

2015 finalists to be named next week.

[steveshea]

2015 finalists to be named next week.

Or maybe not. Still no word. I'm looking forward to the announcement (and the conference).

[EvanZ]

Another is on quantifying shot quality in the NBA (they introduce a new EFG% which accounts for the difficulty of the 2-pointer).

Interesting way to put it. The 2-point shot isn't actually more difficult in itself, per se. Is it?

I mean if we control for assisted/unassisted, distance, defender distance, shot clock it seems to me the distinction between 2 and 3 points goes away (except for the final result!).

And if we do adjust for these things, how much more should we value players who seem to intentionally take "too difficult" shots that end up hurting their team (i.e. the Kobe's and Smoov's of the world)? It's not that these shots aren't necessary to some extent, of course. That's why in the past I tried to account for "marginal" shot selection based on expected usage, even (especially) for mid-range shots as I discussed here:

http://www.goldenstateofmind.com/2011/9 ... e-shooters

[Crow]

Evan's post is on point for part of the comment I recently made in the big man thread.

[steveshea]

Interesting way to put it. The 2-point shot isn't actually more difficult in itself, per se. Is it?

It's been a year since that post, but I believe I meant the difficulty of one 2-pointer as compared to another 2-pointer. I'll have to go back and check the paper, but I think they were trying to adjust a player's eFG% by the difficulty of the shot they take. eFG% gives the same credit for an open dunk as a contested mid-range jumper. I believe the paper was adjusting eFG% to give more credit to the player that made the contested mid-range jumper (for example).

Again, I'll have to go back and check, but I believe the paper focused on shot execution and not shot selection. So, it was not trying to judge players based on the shots they choose to take or not take.

[Crow]

http://www.sloansportsconference.com/?page_id=462

[Crow]

2 basketball titles in above link.

[Crow]

http://www.sloansportsconference.com/wp ... lation.pdf

Lineup simulation. Pretty brief write-up but the equations are given.

[mtamada]

KirkG has a good write-up at Grantland of the paper that he co-authored. I've only skimmed the article that they'll be presenting, but it looks like good work. An innovative and clever approach to figuring out who was guarding who (look at the location of the ball, the hoop, and each offensive player, take a weighted average of those locations, and see where the defenders were with regard to those weighted averages, and sum this information across time to see how much responsibility each defender had for each offensive player, cumulatively for the possession). That of course is a simplified description, they used something called a Hidden Markov model to model the changes in defensive responsibilities as time elapsed during the possession.

The Grantland article has good animated graphics to illustrate the results, with blue lines to indicate defensive responsibility.

I am concerned that the defensive matchup calculations which lead to those blue lines are too simplistic -- not from a mathematical point of view, but from a basketball view. That is, defensive assignments are more than a matter of being the closest to the guy you're supposed to be covering; sometimes the right place to be is not close to him because you have alertly realized that you need to switch. Or make the correct choices when defending the pick and roll. That is, the blue lines shown on the graphs might not be identifying the correct defender or the correct defensive responsibilities in a given play.

But that sort of "intelligent blue line" drawing will require even more complex modeling. In the meantime there's plenty of complexity and innovation in what KirkG and his co-authors have done in this paper. I think it represents additional good progress in using the detailed data about time and location that teams now have access to.

[NateTG]

...
I am concerned that the defensive matchup calculations which lead to those blue lines are too simplistic -- not from a mathematical point of view, but from a basketball view. That is, defensive assignments are more than a matter of being the closest to the guy you're supposed to be covering; sometimes the right place to be is not close to him because you have alertly realized that you need to switch. Or make the correct choices when defending the pick and roll. That is, the blue lines shown on the graphs might not be identifying the correct defender or the correct defensive responsibilities in a given play.
...

SporstVu data will never tell you what the 'defensive assignments' made by the coach in the huddle are. It can only tell you how the team matches up defensively on the floor. And - as you astutely point out - it's unclear how to assign credit (or blame) during denfensive transitions like the ones that happen in response to pick and roll plays. In fact, it's quite likely that that sort of thing is quite contextual depending on the questions you want to ask.

For the purposes of reasearch like this, the goal is usually to get a model that's "mostly right" and hope that sampling dilutes out the errors. The are a number of sensible approaches to inferring defensive responsibilities, and, really, any one that is repeatable and well-defined should be usable in research. Simple schemes for assignment have the advantage that they're obviously sensible and that they obviously have limitations so we might be less tempted to look too deep.

It is a bit of a reach to make implications on defensive capabilitiy before verifying that the player assignment scheme makes sense. It would really be nice to see AB sample testing or some other verification of the results in the paper. We could - in principle - run that kind of test ourselves, but the raw SportsVU data is not readily available.

[Crow]

Anyone here do or know who did the Move or Die Ball Movement paper? No link provided yet at Sloan.

[Crow]

In the film world, Sundance filled a need and then Slamdance and others came along and filled a need as Sundance evolved. Wonder if an alt-conference before, during or after Sloan could rise and fill perceived unmet needs. Perhaps just on basketball with lots of bloggers / independent analysts more willing to share new ideas and debate industry performance. Would be lot of work but could be fun & useful. Perhaps a basketball site with some staff and funds could take the lead. Scheduling could be aggressively oppositional to pretty compatible for at least Sloan's prime basketball stuff. Just a thought, if current satisfaction levels of attendees warrants.

[NateTG]

> ... Wonder if an alt-conference before, during or after Sloan could rise and fill perceived unmet needs. ...

I think NESSIS has better papers.
http://www.nessis.org/

[Crow]

Yes, there are a few other alternative conferences.

Does this one post papers or sell them as a journal? Didn't see any information that. Is it mainly academics or is there decent representation of practitioners / amateurs?