NBA shot distribution compared to the past

[colts18]

I looked at the distribution of shots between the 2013 season and 1998 season (earliest season we have data for that isn't a shortened 3pt line year).

% of shots in paint:
98: 47.4%
13: 47.4%
Diff: +0.02%

% of shots in midrange:
98: 36.9%
13: 28.4%
Diff: -8.52%

% of shots 3 pointers:
98: 15.7% (19% of 3's were corner 3's)
13: 24.2% (28% of 3's were corner 3's)
Diff: +8.50%

paint FG%:
98: 53.3%
13: 53.7%
Diff: +0.4%

mid range FG%:
98: 38.9%
13: 39.3%
Diff: +0.36%

3 point FG%:
98: 34.8%
13: 36.2%
Diff: 1.4%

Teams took the exact same amount of shots in the paint in 2013 as they did in 1998. The difference is that 2013 teams took shots away from the midrange and turned those into 3 pointers. Plus teams are shooting a higher percent of their 3's from the corner compared to the past.

[Crow]

That is helpful to see. It would be even more useful to know what share of those mid-range shots came before crunch-time on the shot clock and what share was "contested". I assume the data for that is limited to last 3-5 years. I assume the trends would be for less early in clock mid-range shots and less contested ones but it would be interesting to know the year to year rate of change and if the rate of change is fairly constant, decelerating or still accelerating. There is probably a semi-hard frontier for how much mid-range shots can be diminished within current rules but are we about there or not? Houston's example would suggest that almost everyone could go further and whether it is wise to go as far as Houston is still a question especially in terms of the playoffs.

[nbo2]

There is probably a semi-hard frontier for how much mid-range shots can be diminished within current rules...

Houston's D-League team, the Rio Grande Valley Vipers, begs to differ.

[Crow]

The frontier I was thinking of was in the NBA, where I have been told teams can do a pretty good job denying or pressuring the 3 pt shot and reducing its effectiveness if they know that is what is coming most of the time. Maybe this is not so true or maybe it is particularly not true in the D-league? I haven't watched a lot of D-league play but was not that impressed by the defensive effort. Against the Vipers are the defenders making denying or challenging the 3 point their first priority? Are they making a good effort? The Vipers are hitting at such a high rate (according to that chart) I am hard pressed to believe that the defense is doing all they can or all an NBA team would do. But I'd appreciate any further description or analysis by those who have seen the Vipers play.

I had called for experiments like this a year or two ago. In the NBA pre-season or even in real NBA games. The D-league test is interesting but until it is done at the NBA level I am not sure how far the NBA frontier can be pushed.

The Vipers went from 27 3 pt attempts per game in 2012-13 to 46 a game now. But they have only done this for a bit over a month. Will their success diminish the 2nd, 3rd, 4th time they play teams? Only a couple teams have seen it more than once so far. They hit 34% overall for that 5 game stretch compared to 41% in first 5 games. The middle 4 games were at 35%. Shot above 38% in 5 of first 7 games but then only 2 of next 7. What will the future hold?


By the way what date / period of coverage is that shot chart from? Team site says they are only hitting 37% on 3s for the season.
http://www.nba.com/dleague/riograndevalley/statistics/


Only 1 other D-league is above 30 3pt attempts per game. Legends take the 5th most but make the highest FG%. OKC's D-league affiliate is near the bottom on 3 pt attempts and FG%.

Vipers pair their 3 pt attempt lead with a lead on offensive rebounds (you want both from this strategy). Rockets are #1 in 3 pt attempts in the NBA but are only 9th in offensive rebounds and OR%.

[schtevie]

I am compelled to add to this wonderful, year-end string as its specific content and the implicit comment it offers on the state of the statistical/analytics "revolution" in the NBA are favorite topics of mine.

The facts presented by colts18 clearly remind us of the free lunch that the three point shot is and has always been. Back up a few feet and each shot attempt is worth an extra 0.3 points! That is a lot of points, simply available in extraordinarily elastic supply.

Were only this point to have been clearly made way back then, surely some enterprising franchise on the cusp of a championship would have exploited this opportunity to accomplish great things. Well... apparently not.

But one might say that that was all the pre-modern past, whereas now with the availability of SportsVU there is a newfound ability by the most forward-looking franchises to multiply by 1.5 and finally pick the lowest-hanging fruit (what, again, has been sitting on the tree for about a half century). And Crow brings up contemporary Houston in this regard. Prior to addressing the lovely Viper diagram introduced by nbo2, it should be said that what Houston has been doing this past couple of years is not historically extraordinary.

Yes, Houston has been ranked first or second this year and last in terms of three point attempts (per 100 possessions) but the competition has been taking more too (averaging one additional attempt every two years, huzzah!) such that their current relative advantage on this account is historically uninteresting. If anything, the relative advantage of league leadership in this category has been decreasing slightly over time (last year being something of an anomaly) and this year in particular. The general point is pretty clear: no one, never, has exploited the potential of the three point shot to gain a "critical" competitive advantage (the case of Jim O'Brien's 2003 Celtics is an interesting exception, in a particular sense, and that probably didn't end up helping his career longevity).

And up until of a day ago, I was quite sure that exploitation of the three point shot would never, ever be an interesting competitive margin in the NBA, exploited by Houston or any other franchise, the dead hand of tradition and the myriad impediments to progress being what they are, but then...Vipers!

I had never paid any attention to the D-League, but here we have Houston's franchise, engaging in some meaningful experiments. Increased three point usage in 2011 and 2012, presaged the course taken by the mothership in 2013. And this big Viper jump in 2014? Maybe something very interesting is brewing for the latter half of this season or beginning next?

It will be interesting to see whether Daryl Morey has the courage of his analytic convictions (not to assert that these are his). Frontrunners and respectable franchises simply don't behave like the Vipers. They just don't because... NBA.

We shall see.

P.S. And as for the D-League experience not being relevant, I see nothing in the record (or conceptual arguments) suggesting that 30ish three-point attempts (per 100 possessions) is some sort of NBA upper bound.

[Crow]

The average rate of change for the league average for 3 pt attempts per game for 20 years is +4.7% per year. For last 10 years, +3.7%. For last 5 years, +3.3%. On average the rate of change is decelerating, using these guideposts. But the last 2 seasons the rate of change went way up to +7.6% per year. Will it start to decline again (and how fast?) or is the real revolution just getting started? Is there a point where the NBA rules committee intervenes? I'd guess that there is. I've wondered if at some point in the future there might be talk of 3 points for a dunk. From a fan satisfaction standpoint, I think there is an argument for that on its own, beyond the balancing with the 3 pt long shot argument. If the rate of change were to stay at 7.6%, the league average would get to 30 attempts per game in the 5th year after this one and Vipers' level in another half dozen years. I don't think the actual trend will be anywhere near that aggressive. If the actual long-term trend is +4%, it will take 9 years to get to 30 attempts and 7-8 more to get to 40. Since the distance change issue, the rate of change has exceeded 5% for two seasons only twice and it has not yet for 3 seasons.

[xkonk]

It seems like making a dunk worth 3 points would encourage more hard fouls and collisions near the rim. I don't know if that's something the league would want to do.

[Crow]

A lot of those would still be entertaining moments / sportscenter highlights.


In 2010-11 season 3.3% of team-games had 30+ 3pt FGAs by that team. Shortened 2011-12 season 2.6. Last season 5.6%. This season about 6.3%. Still a long way to go to make it the norm. Only 9 teams have done it at least once this season., 14 all last season. Portland leads with 10 such games this season. Houston will (only?) 7 of 34 or about 20%. Miami, San Antonio, Indiana and OKC all with none yet. Last season it was Miami 7, San Antonio 5 and Indiana & OKC with none. Last playoffs the frequency was 8%. Houston had 5 of the 14 times in 6 outings. They were W1-L4. Rest of teams with 30+ 3ptrs were W5-L4. Not that impressive. In 2011 playoffs teams achieving this frequency were 4-1. in 2010 it was about even. More data needed to judge the effectiveness of this strategy in the playoffs.

[xkonk]

A lot of those would still be entertaining moments / sportscenter highlights.

Probably true, but is the increase in Top 10 plays worth the increase in somber intros on LeBron's head laceration or Kyrie Irving's concussion problem? Alternatively, the league could introduce harsher penalties for defenders to discourage fouls on dunks, but at some point you have to wonder when they would just ban defense entirely.

[Crow]

Has not been a 50+ 3pt attempt game by a team yet in NBA though Rockets put up 46 in a game last spring. 50 attempts probably happens within next year.

[Crow]

More 3 pt game facts:

Took almost 4 years for first 30+ 3 pt attempt game. 1989-02-09 SAC - GSW. Don Nelson was there but the 31 attempts came from Jerry Reynolds' Kings who won. It then took almost 5 years for it to happen again with Rudy T.'s Rockets (in a loss). There have been 1086 such games all-time. The team record for games with 30+ 3pt attempts is 547-539. More than half the total belongs to 5 teams:

NYK
ORL
HOU
PHO
GSW

Utah has never done it.

Houston did it 37 times last regular season and 5 times in playoffs. They were 16-21 in the regular season and 1-4 in the playoffs.
The Knicks however were 27-11 overall last season in such games.
Team probably vary on whether they have such games more when they are hot, when they are just desperate or whether there are still trying to figure that out. And there is other relevant stuff going of course.

The average # of such games for the last 7 title winners was rare- a bit over 2 per team in the regular season and a bit under one time in the playoffs. The record for such team games overall in this timeframe is 23-26 and the 3pt FG% was about 36%. In the finals it is a record of 2-2 and a 3pt FG% of 38%. Has not yet been a title winner yet to make more than incidental use of this strategy, much less rely on it heavily. Not sure if their avoidance of the strategy is necessarily wiser than using it, just reporting what has happened to date.

For all finals games since 2007 the average 3 pt line is 7.3 makes on 19.8 attempts with slightly less than half the team games having 20+ attempts. Less than 20% had 25+ attempts with a W-L of 7-7 (6 wins by title winners). By contrast the team games with 15 or fewer 3 pt attempts had a W-L of 12-4 (8 wins by title winners). Golden State and Orlando each have 10 of these playoff games with Golden State going 4-6 and Orlando going 7-3. The only recent title winner with a W-L in such games in the timeperiod above .500 is Miami at a bare 2-1.

If 30+ 3 pt attempt games is the wave of the future strategy is not heavily based on past success.

How many teams knew these facts previously? How many considered them in deciding the prominence of the 3 pt attach in their overall strategy? I don't know but I'd guess few. Even if an analyst knew it, what about the GM and coach? How many will read this now and consider it? I don't know. The data may not have the "right answer" in it but it seems worth thinking about. I am supportive of thinking about a high frequency 3 pt attack (based on mid-range shot reduction) but I'd want to research it deeper and see more positive team results rather than just adopt it based on the logical consideration of simple comparative 2pt and 3 pt shot success / efficiency alone.


Lastest Vipers' data: an 8 for 38 outing.

[schtevie]

I'm all for research, deep research even, but it really isn't tenable to suggest that there is any question that significant, positive returns exist on the 3PA margin. Even decades after these should have been exhausted were the NBA a truly competitive league. But now we have Daryl and the Vipers, and hope is kept alive!

A few comments.

There is, of course, a significant and well-known bias in the "raw" data when looking for a relationship between team success and high volume three point offenses. 3PAs are the best available come from way behind strategy so we should expect a negative correlation between "excess" 3PAs and wins. And indeed we do.

Consider post-1998 game results for teams having Viper level shot selection: 40 or more 3PAs. We see 29 games and 19 losses. But scratch just beneath that surface, and you see that three pointers were very good shots indeed, even in losing efforts. The eFG% in the 10 wins was 0.66 (standard deviation of 0.11) and in the 19 losses 0.55 (sd 0.09). Compared to supposed displaced mid-range shots, only in one of these losses was the eFG% (.34) lower than what we expect average mid-range success rates to be.

More generally, if you take into account home-court advantage (conservatively 3 points per game and 19 of the 29 games were away games) and give equal weight to the per game score margins (so as to simply adjust for the shoot 3s when behind effect) for wins (16.3) and losses (-5.4) you get an average, estimated victory margin for taking 40+ 3PAs of 5.5. Thinking of this as reflecting perhaps 20 to 25 "excess" 3PAs compared to the average, you infer a return of about 0.20 to 0.25 points per "extra" 3PA. Makes sense, no?

Moving on, I note that Zack Lowe had a very nice Grantland article on the Viper phenomenon:

http://www.grantland.com/story/_/id/101 ... urting-nba

(what offers pre-absolution for the atrocious, first two paragraphs of his subsequent article on tanking solutions. ESPN, give it a rest.)

Herein, we see a few folks diminish their reputations (leaving aside Larry Brown being thrown under the bus by Billy King):

(1) Jeff Van Gundy for simply being a dope. (And he's one of the better coaches of all time!)

(2) Shane Battier for thinking that 40% 3P% is the relevant cutoff for optimality and having not a clue about the elasticity of supply of three point shooting ability.

(3) and the Commish (and minions) for apparently believing there's something more aesthetic about a 36% 2PA vs. a 36% 3PA.

And also featured are comments by Terry Stotts, a 3PA enthusiast, which I mention only so as to contrast with his extended remarks on the subject in a recent Henry Abbott interview on TrueHoop. These are interesting and worth reading, as they seem to make clear two points, first, that his views are informed by the history of the three point shot (though I would suggest his historical reckoning is a bit off) but second, and more importantly, despite his success and awareness that he has no particular interest in pushing the envelope.

Daryl, our hopes rest with you. Next year.

[Crow]

"it really isn't tenable to suggest that there is any question that significant, positive returns exist on the 3PA margin"

Of course I did not do that. Not sure if you meant to imply that on my part.



40+ 3 pt FG attempt games may look different in the historical record than 30+ attempt games.

From 2006=7 to present the record in 30+ attempt games is 390-358 or 52% win. Not that compelling.

Yes it is used to try to come back but it also often used to try to win earlier and did not lift win% meaningfully from break even.



And from 2006=7 to present the record in 40+ attempt games is 9-16 or 36% win.
http://bkref.com/tiny/u1xn2
Not good. Element of surprise gone?

40+ attempts in the playoffs? Tried twice (1W-1L), both overtime games and not done since 2007.

Most Houston (or anyone else) got off last playoffs was 37 so I am not sure how viable 40+ 3pt attempts is as a strategy in the playoffs. It takes will to try it but it also takes strategy and effort to make it happen and it is very rare to date. Of course I hope for more data points to help with considering the strategy.

In their 30+ attempt games they shot just under 50% eFG from 3 pt land but it was a meaningful to very large part of the reason they lost 4 games. It was a direct and meaningful part of the reason for just one of the wins (in that game they also shot tremendously from 2 pt land as well). In the other win it was great 2 pt shooting that mattered more than the 31% of the high volume 3s that went in.

[xkonk]

A better measure might be 3PA taken in the first half instead of the whole game. Do teams who shoot a lot of threes in the first half, reflecting their offensive preferences/abilities and leaving out most end-game strategic concerns, end up winning more often?

[schtevie]

I am supportive of thinking about a high frequency 3 pt attack (based on mid-range shot reduction) but I'd want to research it deeper and see more positive team results rather than just adopt it based on the logical consideration of simple comparative 2pt and 3 pt shot success / efficiency alone.

Crow, I don't want to misrepresent your views, but I take this comment as questioning the notion that the 3PA margin is a profitable one.

40+ 3 pt FG attempt games may look different in the historical record than 30+ attempt games. From 2006=7 to present the record in 30+ attempt games is 390-358 or 52% win. Not that compelling. Yes it is used to try to come back but it also often used to try to win earlier and simply fails to lift sufficiently.

My sense of the matter is that "excess" 3PAs show up in the historical record disproportionately in terms of comeback efforts. But no matter. Furthermore, I am prepared to believe that team results for games in the 30s range of 3PAs are somehow empirically different from those in the 40s. But, whatever this reality is, here too I would be very surprised to see a lack of "excess" return compared to mid-range 2s.

Where I want to quibble with your statement is your characterization that a 52% win percentage isn't compelling (again, recognizing that this is likely biased down by comeback or garbage time phenomena and also recognizing that this doesn't reflect the potential of a designed offense). This can alternatively be thought of as about an extra 2 wins per year or 1 point per 100 possession for an average team. And this is not nothing, especially given that it comes "free". Alternatively, for an elite team, such an uncompelling result could easily be thought as the difference between winning or losing a championship.