APBRmetrics

My initial research into this was in this article which I'd posted here before: http://blog.cacvantage.com/2014/02/the- ... fense.html

Like that article, this research comes from Vantage Sports' data.

But recently, I've dug into it a bit more looking at the differences between contested Pull Up (or off the dribble) 3 point shots and contested Catch & Shoot 3 point shots. It turns out, there is a point where an open mid-range shot is preferable to shooting a contested three as can be seen in this graph:



The graph can be read as at x seconds, y = the PPS for all shots < x seconds. It turns out that off the dribble threes do become less efficient than an open mid-range shot when you get down to about 8 seconds and it's roughly equal from 13-8 seconds.

An important caveat though: as you get down to 1 second, the sample size is about 300 for contested off dribble 3 pointers (and hence probably why you see such a steep decline as opposed to the other types) and about 500+ shots at < 2 seconds. But at about < 8 seconds, the sample size is about 1600+ shots.

Also, to see the difference for looking at contested catch and shoot 3 pointers vs. open mid-range and contested close shots:



The same caveats from before apply except there is a larger sample size across each time interval. In total, there's about 3500+ contested off the dribble 3s and almost 11,000 contested catch and shoot 3s.

One last bit of info: about 62.5% of off dribble 3 pointers are contested vs. 52.3% for catch and shoot 3s.

The next part of this research would be to of course look at open mid-range shots off the catch and shoot. Perhaps this is more efficient than dribbling around and shooting 3s if contested.

Wonder what the level of difference is big enough to draw active NBA mgt / coaching attention to a shooting in time choice.

This is really good stuff. Thanks for posting.

I echo Crow's question. I would have thought coaches would already notice a big enough difference, but then I hear Randy Wittman defending his team's steady diet of 2pt jumpers by saying they're taking open looks, which (to him) means they're "good" shots. The charts here show that a contested three is generally a better shot than an open mid-range shot.

kjb wrote:This is really good stuff. Thanks for posting.

I echo Crow's question. I would have thought coaches would already notice a big enough difference, but then I hear Randy Wittman defending his team's steady diet of 2pt jumpers by saying they're taking open looks, which (to him) means they're "good" shots. The charts here show that a contested three is generally a better shot than an open mid-range shot.

One factor that works in Wittman's favor is that an open shot early (viewed under the lens as only PPS) ignores the turnover chance from time a to time b. It could be that a future contested 3 is a higher PPS (and it certainly is, as the charts above demonstrate) but the EV of the possession is likely lower, as getting to that point has probably cost the team non-zero turnovers.

Personally I think the beauty of a 7 seconds or less offense is not so much the looks it generated (transition layups or quick 3s) but rather just not committing future turnovers.

Is there evidence that a quicker pace yields fewer turnovers?
In this century, 445 team-seasons stratified by pace, some median numbers:That bottom line is just 45 slowest teams; the others are 100 a la http://bkref.com/tiny/OUFe8

Very interesting stuff. I apologize if you've already done this (and I'm sure this information is available somewhere else, independent of shot type studies), but it might be useful to simply have a line showing the total offensive efficiency by time left on the clock on the graphs (where 8 seconds would mean all possessions that last at least to 8 on the shot clock). This would show the opportunity cost of shooting a specific shot at a certain point in the shot clock, and the idea would be that one shouldn't take a specific shot if it resides below this line.

gzchen wrote:Very interesting stuff. I apologize if you've already done this (and I'm sure this information is available somewhere else, independent of shot type studies), but it might be useful to simply have a line showing the total offensive efficiency by time left on the clock on the graphs (where 8 seconds would mean all possessions that last at least to 8 on the shot clock). This would show the opportunity cost of shooting a specific shot at a certain point in the shot clock, and the idea would be that one shouldn't take a specific shot if it resides below this line.

That's a great idea. I'll add that in when I get the chance.

knarsu3 wrote:It turns out, there is a point where an open mid-range shot is preferable to shooting a contested three as can be seen in this graph:

In your dataset what's the ratio of open shots to contested shots for each shot type in the last 5 seconds? In other words, I wonder how often those open mid-range situations even happen with 5 or less seconds on the shot clock.

Bobbofitos wrote:It could be that a future contested 3 is a higher PPS (and it certainly is, as the charts above demonstrate) but the EV of the possession is likely lower, as getting to that point has probably cost the team non-zero turnovers.

But there's also a possibility of future open three, free throws or a layup so while the risk is higher the reward is way higher. Though overall I agree that those graphs would be even more awesome with turnovers somehow included because they would be the biggest equalizer for those best-case scenario shots.

Great stuff, thanks for posting. Love the new data you provide.

Do you have any interest in categorizing mid-range attempts separately as catch-and-shoot and pull up as well?

wilq wrote:

knarsu3 wrote:It turns out, there is a point where an open mid-range shot is preferable to shooting a contested three as can be seen in this graph:

In your dataset what's the ratio of open shots to contested shots for each shot type in the last 5 seconds? In other words, I wonder how often those open mid-range situations even happen with 5 or less seconds on the shot clock.

I'll try to look at this in my next article but if it can't make the cut (because there is a lot of information yet to be explored), then I'll certainly post the ratios here. But the very basic information can be seen in the article here: http://blog.cacvantage.com/2014/02/the- ... fense.html

Actually, I think the answer to your question should be in the article above.

Bobbofitos wrote:It could be that a future contested 3 is a higher PPS (and it certainly is, as the charts above demonstrate) but the EV of the possession is likely lower, as getting to that point has probably cost the team non-zero turnovers.

wilq wrote: But there's also a possibility of future open three, free throws or a layup so while the risk is higher the reward is way higher. Though overall I agree that those graphs would be even more awesome with turnovers somehow included because they would be the biggest equalizer for those best-case scenario shots.

I'd like to add this in eventually but it's a bit harder to acquire the data for.

fpliii wrote:Great stuff, thanks for posting. Love the new data you provide.

Do you have any interest in categorizing mid-range attempts separately as catch-and-shoot and pull up as well?

Yes, definitely. I'll have an article for Vantage's launch that will look at all of that- in addition to breaking down 3 point shots into Wing and Corner. But that is definitely an important question with an interesting answer.

knarsu3 wrote:

fpliii wrote:Great stuff, thanks for posting. Love the new data you provide.

Do you have any interest in categorizing mid-range attempts separately as catch-and-shoot and pull up as well?

Yes, definitely. I'll have an article for Vantage's launch that will look at all of that- in addition to breaking down 3 point shots into Wing and Corner. But that is definitely an important question with an interesting answer.

Very cool. Looking forward to it!