APBRmetrics

gravityandlevity wrote:That's a good point: that the distribution of shot quality can depend on the shot clock time. I'd love to do better than assuming a static distribution (same mean and variance of potential shot opportunities at all shot clock times) if I could only come up with a reasonable way to figure it out.

I don't know if you saw my last post (last one of prev. page), but I think whether made shots were assisted or not would be one way to get an estimate. Also, shot location (if you have that).

DSMok1, sorry to overlook your earlier comment. It sounds like your suggestion is a good one, but I don't understand how one can discriminate between an assisted shot attempt and a non-assisted shot attempt. In other words, how can you tell which missed shots were assisted (or would have been assisted had they gone in)? Is that somehow available in play-by-play data?

When you said "assisted shot attempts have a higher success rate", how was that statement validated?

Thanks.

gravityandlevity wrote:DSMok1, sorry to overlook your earlier comment. It sounds like your suggestion is a good one, but I don't understand how one can discriminate between an assisted shot attempt and a non-assisted shot attempt. In other words, how can you tell which missed shots were assisted (or would have been assisted had they gone in)? Is that somehow available in play-by-play data?

When you said "assisted shot attempts have a higher success rate", how was that statement validated?

Thanks.

Some of the best research online has since been removed (writer got hired), but here was the seminal research: http://82games.com/assisted.htm

Further research indicated there are two distinct "skill curves" relating, at a high level, usage and FG% -- a flatter one for assisted FG% (if someone can send the assist, FG% is fairly constant for the player) and a steeper one for unassisted (marginal FG% continually drops--though for star creators the curve is less steep). This is related to your Ewing Theory post from a couple of years ago. That is the research which has since, I believe, been pulled offline.

Another post: http://www.48minutesofhell.com/spurs-st ... -efficienc

EDIT: found one site that has since been taken offline: http://web.archive.org/web/201102261948 ... lysis.com/ I don't know if one can get to the skill curve analysis.

I gots some things to say about Brian's work, what was first introduced to the board back in September - http://apbr.org/metrics/viewtopic.php?f ... 2188#p2188 - but curiously didn't evoke the same level of interest then. But first I've got to quibble with some of Guy's points and priors because no one else has. Apologies for any thread drift that ensues.

Guy wrote:

After thinking about the model some more since the time the paper was accepted for publication, I think I'm ready to say that I have very little confidence in the conclusion that NBA players "undershoot".

Kudos to Brian for being willing to rethink his conclusion. This is not always easy to do, especially once your work is public.* And I take him at his word that he did not hype his prior conclusion to journalists. But this paper is a great example of a serious problem with sports analysis by academics today: you will get a lot more attention if you find an ineffeciency in a sport. That's interesting, and allows writers to sound sophisticated and reference "Moneyball" (maybe even run a picture of Brad Pitt). If Brian's paper had concluded "NBA players shoot exactly when they should," I think media interest would have been much less, probably close to zero. And so researchers face a strong incentive that can bias their research, and certainly biases publication -- go find inefficiencies. (Similarly, when looking for signs of racial bias in sports, there is a strong incentive to find bias -- a study finding that "NBA Referees Not Racially Biased" doesn't get attention in the NYT.)

First, a relatively minor point. Is the implication here that the conclusions of Justin Wolfer's well-known research on this issue are suspect because of "incentives" for media attention?

Guy wrote:I'm not suggesting we should go back to the days when papers aren't made public until published in a refereed journal. We shouldn't have to wait 3 years to see research like this, and in any case journal referees likely wouldn't have spotted the problems mentioned here. (Lesson for young academics: use the Internet to get good feedback from subject-matter experts for free!) But I do think that journalists need to be more conscious of the bias in favor of this kind of finding. Get subject-matter experts to review this kind of paper BEFORE you write your article about it. The fact is, sports is in general a highly-efficient arena. Most findings of inefficiency will be wrong. In this case, it is highly unlikely (though possible) that teams systematically shoot too late. As long as there were some variance in how aggressive teams were, those who shot earlier would tend to win. That would encourage them to shoot still earlier, and other teams to emulate them. It's hard to imagine a league of professionals not finding the right equilibrium rather quickly.

* See: Berri, David.

Where do I begin to strenuously disagree with the idea that "sports is in general a highly-efficient arena"? There is a surely a truth to the point, but only when construed narrowly. Taking a larger perspective, the view seems essentially untenable. A coach comes up with a twist on well-known Xs and Os, and it is copied tomorrow. Synergy comes out, "everyone" uses it. In such areas, where things are within the realm of comfortable, conventional wisdom, teams compete like hell from the get go, with any initial advantages to first adapters being small and short lived.

But this is decidedly not the case where thinking outside the box is required for advantage to be taken. In these instances, where historically the stakes have been huge, the right equilibrium does not obtain "rather quickly", unless by this the meaning is decades. Some examples? Racism and xenophobia are two obvious categories where "traditional" thinking sacrificed the potential for very significant competitive advantage. More speculatively, what about the inability to anticipate the potential for unconventional, tweener players to revolutionize the game? Could there have been a Scottie Pippen before Scottie Pippen, for example? Too speculative? Then a very concrete example that I have brought up many times before, where the foregone gains are easily approximated: the glacial pace of adoption of the three point shot (at the expense of the long range two). This was yet another essentially free lunch that nobody decided to partake of in a competitively significant way.

But more to the point there is inadvertent historical irony in writing "As long as there were some variance in how aggressive teams were, those who shot earlier would tend to win. That would encourage them to shoot still earlier, and other teams to emulate them. It's hard to imagine a league of professionals not finding the right equilibrium rather quickly."

This has nothing to do with Brian's paper, but the mind instantly turns to the misguided offensive notions that originated during the Celtics-dominated '60s. The Cs ran a lot, and they won. A lot. Causation? Nope, but with this becoming the conventional wisdom it dragged down the offensive efficiency of the Association for generations. The Cs won because of defense and, to a first approximation, despite their offense. It took decades (what was a primary feature of the evolution of basketball in the '70s) for the right equilibrium in terms of game pace to be restored.

Not trying to be difficult, and apologies for the rantiness. This is, however, a very important point to get straight: not all competition is competitive.

Eh, maybe I'm wrong.

Carry on.

(For the third and I promise last time: I'm new, so please be kind in your responses)

Am I missing something? How can points per shot with 24 seconds (rounded up, so thus between 23.01 and 24 seconds on the clock) be at least 0.8? The only way I can see this occurring is on a put-back/tip in?

Assuming I'm correct there, why would you wish to include plays beginning with offensive rebounds when trying to determine the optimal time to shoot? Surely on a possession beginning directly under the offensive basket, the logical move would be to put the ball straight back up. Wouldn't this skew your results? I agree with an earlier post, that limiting the research into all possessions beginning with a made field goal would be more instructive. In this instance, the 'right move' which should determine the 'right time to shoot' would skew the entire set.

Again, if I'm way off base, I'd love to know why

3ch03s wrote:
Assuming I'm correct there, why would you wish to include plays beginning with offensive rebounds when trying to determine the optimal time to shoot? Surely on a possession beginning directly under the offensive basket, the logical move would be to put the ball straight back up. Wouldn't this skew your results?

It must be a good question, as it's been asked a few times already in the thread. The answer is yes. I would suggest quickly reading through the thread and you'll find a discussion of this.

Is the implication here that the conclusions of Justin Wolfer's well-known research on this issue are suspect because of "incentives" for media attention?

The implication is that all researchers know their study will get more attention if they find bias. What impact that has on any specific piece of research, I obviously can't say. It certainly creates a bias in terms of what the public learns about this kind of research.

It's been a couple of years since I read the Wolfers paper. What I recall is that the authors found very, very small racial disparities but found ways to make them sound very large (with the journalists perhaps further magnifying the findings). IIRC, the authors also drew conclusions -- e.g. that the bias disadvantaged minority players specifically -- that the research did not substantiate. But againg, it's been some time since I looked at this.

More germane here is the bias in favor of finding inefficiencies. Does anyone disagree that there would have been less media interest if this paper had concluded "it appears NBA players shoot at about the right time?"

Guy wrote:.. Does anyone disagree that there would have been less media interest if this paper had concluded "it appears NBA players shoot at about the right time?"

This is and has always been the case: "Man Bites Dog" makes headlines.
To report that most of conventional wisdom -- coaching decisions to fan consensus -- is generally right, you don't get any attention. That's fine, unless you're an attention hound.
Conversely, having radical interpretations based on filthy data*, you'll still get headlines. Some people apparently make a career of this. There must be some evolutionary niche for self-promotion.

* -- not an indictment of the article being discussed, just referencing a generic extreme.

Guy wrote:

Is the implication here that the conclusions of Justin Wolfer's well-known research on this issue are suspect because of "incentives" for media attention?

The implication is that all researchers know their study will get more attention if they find bias. What impact that has on any specific piece of research, I obviously can't say. It certainly creates a bias in terms of what the public learns about this kind of research.

It's been a couple of years since I read the Wolfers paper. What I recall is that the authors found very, very small racial disparities but found ways to make them sound very large (with the journalists perhaps further magnifying the findings). IIRC, the authors also drew conclusions -- e.g. that the bias disadvantaged minority players specifically -- that the research did not substantiate. But againg, it's been some time since I looked at this.

I think that this was a very poor example chosen to illustrate the point. I too haven't reread the paper since it came out, however, my recollections are these (besides the disgraceful behavior of David Stern). There were no questions about the econometrics being solid, and, furthermore, I seem to recall that the initial results were confirmed when expanded data sets were released later by the NBA. And perhaps more on point, it wasn't the case that the very, very small racial disparities were irrelevant. My recollection is that if you had a ref crew all of one race, officiating a game where there was a large racial disparity between teams that the expectation is that this would have a meaningful effect on the outcome of game. This is not nothing, and it wasn't oversold - though opinions could differ on this point.

Guy wrote:More germane here is the bias in favor of finding inefficiencies. Does anyone disagree that there would have been less media interest if this paper had concluded "it appears NBA players shoot at about the right time?"

Probably not, but hold the presses. My bet is that if and when the true story is told that significant inefficiencies will be revealed on this account, and that this will be another example of league of professionals not finding the right equilibrium rather quickly. I am just not sure that there will ever be adequate data to fully establish the claim.

Just so that everyone knows my own personal opinion, as the author of the paper in question, I'll state it explicitly:

I agree that the paper received an incommensurate amount of media attention, and that much of that was driven by the ability to write "NBA players are doing it wrong!" My goals for the paper were relatively modest, and I think my accomplishments were relatively modest (i.e., probably a step forward, but certainly not the last step). I tried to make that level of modesty explicit in the paper, but of course it's always easy for an author to over-sell his own results, so it's possible that the paper overstates some of its conclusions. If so, I apologize.

It seems that a number of you have that seem opinion about the paper, so I just wanted to say that you should feel free to say it without hurting my feelings.

3ch03s, I didn't include any possessions beginning with an offensive rebound in the data set. It's likely that a lot of those shots during the first second are "cherry picking" after a defensive rebound. It's also possible that some of them come from inaccuracies in the play by play data. I've noticed a lot of times when an event is logged at the wrong time by a few seconds, which would make a big difference for trying to figure out events early in the shot clock.

Brian, would it be possible to give us the shot distribution data broken down by team and even by season?

That's a bit of a tall order, but if there is interest then I can write some automated script this weekend to sort through the data in that way.

Are other people interested? Would you like the data also broken down by dead ball vs. live ball situations?

EvanZ wrote:

3ch03s wrote:
Assuming I'm correct there, why would you wish to include plays beginning with offensive rebounds when trying to determine the optimal time to shoot? Surely on a possession beginning directly under the offensive basket, the logical move would be to put the ball straight back up. Wouldn't this skew your results?

It must be a good question, as it's been asked a few times already in the thread. The answer is yes. I would suggest quickly reading through the thread and you'll find a discussion of this.

Thanks EvanZ,

I skim-read all posts since the graph was posted and didn't spot anything. No mind.

gravityandlevity wrote: 3ch03s, I didn't include any possessions beginning with an offensive rebound in the data set. It's likely that a lot of those shots during the first second are "cherry picking" after a defensive rebound. It's also possible that some of them come from inaccuracies in the play by play data. I've noticed a lot of times when an event is logged at the wrong time by a few seconds, which would make a big difference for trying to figure out events early in the shot clock.

Ah okay. Thanks for pointing that out

gravityandlevity wrote:That's a bit of a tall order, but if there is interest then I can write some automated script this weekend to sort through the data in that way.

Are other people interested? Would you like the data also broken down by dead ball vs. live ball situations?

I'm asking because I posted those plots on Golden State of Mind (someone started a thread on your paper). Someone replied wanting to see the data broken down by team, hypothesizing that the distribution might be different for different teams.

I am mainly interested in dead ball situations for now. It's interesting to think about shot distributions on offense, but also on defense. Each team likely has their own Shot # vs. time and PPS vs. time functions on each side of the ball. I think that would be fascinating to look at.

Another idea I had, but is probably way too much trouble, is to actually calculate an "adjusted shot clock time" for each player.

gravityandlevity wrote:That's a bit of a tall order, but if there is interest then I can write some automated script this weekend to sort through the data in that way.

Are other people interested? Would you like the data also broken down by dead ball vs. live ball situations?

If it is not too much trouble, I would like to see the data broken down.