APBRmetrics

article here: http://blog.cacvantage.com/2014/02/the- ... fense.html

20-24 second interval could be biased by put back offensive rebounds as well

Thoughts? Suggestions?

A lot of good work here. I like all the stepwise detail and then summary but depending on your prime audience there may be times to boil it down a bit more. The final chart on the distribution of shots is important. I don't recall if I have seen that before.

D'Antoni deserves some credit for being early with the early offense bias. I'd say ahead of the analysis but I am not 100% sure on that.

have you looked at offensive efficiency and other detail based on how many seconds were used in the previous opponent possession? I am pretty D'Antoni didn't do this and basically refused to consider that.

Crow wrote:A lot of good work here. I like all the stepwise detail and then summary but depending on your prime audience there may be times to boil it down a bit more. The final chart on the distribution of shots is important. I don't recall if I have seen that before.

D'Antoni deserves some credit for being early with the early offense bias. I'd say ahead of the analysis but I am not 100% sure on that.

Thanks for the feedback, appreciate it. I'm not exactly sure what you mean by boil it down a bit more (maybe I should've included some regressions for example outside of including it on the last graph) but yes, I also could've gone into much more detail, right down to each specific shot location. A question I kind of wrestled with was how much detail to go into for an introductory article- for example, do we need to see the distribution of shots for Right Corner 3s? Or was it ok to lump it into a general Threes category for now. It's something I could go back to and look at the distribution for each specific shot location in a future article. In fact, I intend to do that at some point. But for now and with some of the other projects I'd like to do, I left that aspect of detail out of it. Another issue was there would've been significant sample size issues going into that level of detail.

You may have seen the distribution of shots chart before for anyone else who may have done a detailed shot clock study. That specifically wasn't really unique to Vantage stats but I doubt you've seen the shot defense broken down by the shot clock. Maybe someone has done it with SportVu but if they have, I haven't seen it which isn't to say it's not out there.

As for D'Antoni, he does deserve credit. It'd be interesting to see if it's something he thought of "without the stats" that just happened to be supported by the numbers or if he'd actually seen a study on it and incorporated it in.

Crow wrote:have you looked at offensive efficiency and other detail based on how many seconds were used in the previous opponent possession? I am pretty D'Antoni didn't do this and basically refused to consider that.

I haven't and at the moment, I don't have the data to do it. But that would be an interesting study and Vantage does have the data to do it.

I would agree, I think D'Antoni was just concerned with playing fast. I think what I've found is first and foremost, you do want to get an open shot, which is more important than shooting early for the most part. At the same time, I think you want to avoid the last 10 seconds of the shot clock. My acronym would be GOSE- Get Open, Shoot Early. Obviously easier said then done but I do think if you watch a lot of teams, they slowly dribble the ball up the court and just stand around a lot of the time instead of getting into their sets and plays immediately. All that extra lag time should be eliminated and they should just start running their play as soon as they cross half-court. No need to dribble at the top of the key for 5 seconds just to kill time. I think this is something thats obviously more prevalent in college bball but even in the NBA, there are some teams that do this.

Also, I doubt D'Antoni was concerned with getting the other team to take long possessions. He seemed more interested in a fast pace- with both teams playing that way. I think the research suggests on offense- shoot early but on defense, force the other team to shoot late in the clock. So this doesn't necessarily mean both teams playing a fast pace, it means one team playing that way. All of this is probably pretty obvious but something I doubt D'Antoni preached.

One suggestion: I would like to see breakeven points for open midrange shots, contested 3's, etc. Let's say a league average player has an open midrange shot at 15 seconds. He shoots it, it goes in around 0.90 PPP (a guess). What's the NBA average for PPP after 15 seconds. do that for each second interval to see when its ok to shoot a league average midrange shot or contested 3

... see when its ok to shoot a league average midrange shot or contested 3

Is there really such a thing as a league average shot or shooter?
A guy who is better than avg at 17' is not concerned about what the avg shooter would do. And if a player is 'feeling it', that matters more than what he'd do if he's not feeling it.

Of course professionals should be pretty consistent about the jobs they do. And still, the more you play, the more you experience the ebb and flow of the hot hand. Nothing worse than watching a guy who 'normally' makes a shot, yet on this night he keeps missing it, out of obligation it seems.
Nothing better than watching someone light it up unconsciously.

colts18 wrote:One suggestion: I would like to see breakeven points for open midrange shots, contested 3's, etc. Let's say a league average player has an open midrange shot at 15 seconds. He shoots it, it goes in around 0.90 PPP (a guess). What's the NBA average for PPP after 15 seconds. do that for each second interval to see when its ok to shoot a league average midrange shot or contested 3

That's a good suggestion and something I'd have no problem incorporating in. Although as Mike G said, players are never really league average. Still, it's something worth looking at. I was thinking of doing this for some specific players as well. I'm not sure that I'll have the time to publish an article for specific players (and really, it's just a rehash of what I've done for the league) but I may post those graphs in this thread and on my twitter.

For all shots:


That graph is pretty much what colts was talking about earlier (though haven't done open midrange vs. contested 3s/close yet). Read it as: at 10 seconds, the average PPS from 1-10 seconds is (whatever the point on the graph is)

I could change it so it reads at 10 seconds, avg PPS from 1-9 secs is but that'd cut off a bit of the graph (or I can just expand the axis...). Not sure which is better but either way, they function the same. In the first case, if you want to know what the avg value would be after 10 seconds, you'd look at 9 secs.

Edit: I might add the other one as well. Here's that one:


Same interpretation except as I said, for 10 secs, PPS=avg PPS from 0-9 secs now. Think the 1st one is easier to read and interpret.

Here's the breakeven for open mid-range vs. contested 3s and contested close shots


The efficiency for contested threes really dives as you get later in the shot clock, to the point where it's better to be shooting an open mid-range shot at the end of the shot clock.

Do you have a stat of points per possession by shot clock? Let's say there is 15 seconds left on the clock, what is the average PPP for a possession that reaches at least 15 seconds on the clock? That is important. Without that, we can't determine any breakeven points. We need to know what an open mid range shot is worth at 15 seconds then compare that to the league average for rest of possession starting at 15 seconds.

There are some that argue there's a ceiling on the amount of three's you can take because you can't force them or manufacture them (unless you're Stephen Curry) like a Kobe Bryant midrange shot, which he can get off in most situations. That a team like Houston will run into a wall because good defenses know how to take away three's and rim attempts, and that you need good midrange shooters as a counter to those defenses.

There's also the wild thought some people have that even forcing three's when they're contested is a better strategy than ever taking a midrange shot.

So that leads me to believe there's a distinction between a contested three and a forced contested three that's important in projecting the ceiling or potential of an all-three's strategy. Those are really cool results to see the efficiency of contested three's drops when the shot clock forces an attempt (there's a bias in earlier shot clock attempts, I think, because guys like Curry or Lillard are more likely to take those shots, rather than the "average" three-point shooter.)

By the way, I just read this:

Stotts takes it one step further, suggesting that dismissing midrange shots is a fatal strategy. “It’s not easy getting to the basket, and it’s not easy getting open shots in this league,” he says. “Half of the guys in the league, I would not want to leave open at 15 feet. An open midrange shot is a quality shot in this league. Teams basically make twice as many shots if they’re uncontested. If they’re contested, they make 30 percent; if they’re uncontested, they make 60 percent.”

What are his source? What do you think of that?

http://grantland.com/features/portland- ... -aldridge/

I'm guessing it's rather a description of the extremes. Seldom is a player truly wide open at 15', unless he's an incompetent shooter. League-wide, such shots might be made at 60%, with many players doing better than that.

And not that often does someone have to take a seriously contested shot at that distance. At mid-range, you can usually give it up to a teammate further out or closer in. At some level of 'contestedness', due in part to the clock, players shoot <30%.

Players and teams have always and will always take what the defense allows. Has there been a successful team that swarms the arc and protects the rim, while literally giving opponents the open mid-range shot?

colts18 wrote:Do you have a stat of points per possession by shot clock? Let's say there is 15 seconds left on the clock, what is the average PPP for a possession that reaches at least 15 seconds on the clock? That is important. Without that, we can't determine any breakeven points. We need to know what an open mid range shot is worth at 15 seconds then compare that to the league average for rest of possession starting at 15 seconds.

Do you mean the PPP at exactly 15 seconds or the PPP leading up to 15 seconds (from 24 to 15 seconds)? I could do either. I think you must mean the latter though (since I did the former in my article). I could look at that.

One thing about the graph above, is I could also look at the change from say 15 to 14 seconds. So from 0-15 secs and the next point is from 0-14 secs, the change could be important in determining when to shoot. A steeper change means you lose much more value in shooting a second later.

One issue is that I don't have turnovers data at present. That could be a big factor. In speaking to Dean Oliver on twitter, his models using eFG% found the same thing but when including turnovers and off rebounds, the line became more flat till about 3-5 seconds.

AcrossTheCourt wrote:There are some that argue there's a ceiling on the amount of three's you can take because you can't force them or manufacture them (unless you're Stephen Curry) like a Kobe Bryant midrange shot, which he can get off in most situations. That a team like Houston will run into a wall because good defenses know how to take away three's and rim attempts, and that you need good midrange shooters as a counter to those defenses.

There's also the wild thought some people have that even forcing three's when they're contested is a better strategy than ever taking a midrange shot.

So that leads me to believe there's a distinction between a contested three and a forced contested three that's important in projecting the ceiling or potential of an all-three's strategy. Those are really cool results to see the efficiency of contested three's drops when the shot clock forces an attempt (there's a bias in earlier shot clock attempts, I think, because guys like Curry or Lillard are more likely to take those shots, rather than the "average" three-point shooter.)

By the way, I just read this:

Stotts takes it one step further, suggesting that dismissing midrange shots is a fatal strategy. “It’s not easy getting to the basket, and it’s not easy getting open shots in this league,” he says. “Half of the guys in the league, I would not want to leave open at 15 feet. An open midrange shot is a quality shot in this league. Teams basically make twice as many shots if they’re uncontested. If they’re contested, they make 30 percent; if they’re uncontested, they make 60 percent.”

What are his source? What do you think of that?

http://grantland.com/features/portland- ... -aldridge/

I think for each player it's different. While the numbers in the article point to the league choosing contested threes over open mid-range shots, those numbers are going to be different for each player. Some players should be perhaps shooting more threes while others should be shooting more mid-range shots. It's actually an exercise I've been meaning to do, so maybe I can write something up. But I would like to adjust for catch and shoot vs. off dribble, as well as unassisted vs. assisted so I may hold off on it or revisit when I get the data.

But I have constructed a pair of graphs for both Lebron and Durant that look at this dilemna. There are going to be sample size issues but this graph measures y as PPS for the average time from 0 to x seconds. Both players started out with a maximum n of over 100 for all three shot types, so the sample size issues should be less damaging. Obviously, when you get down to 1 or 2 seconds, the points on the graph aren't very useful but at that point, hopefully the trend line has picked up those issues (perhaps weighting the points less at the end would make more sense). Anyways here they are:





Not sure that these would work for many players because of the sample size issues I mentioned above. For example, tried it out with Curry but he just didn't have enough mid-range or close shots for it to work.