PDA

View Full Version : Are 3's more variable than 2s?



niveklaen
03-07-2014, 09:15 AM
Several people have asked throughout the year whether relying on the 3 really created greater variance - ie is 3pt shooting realy more hot and cold than 2 pt shooting?

Its been 20+ years since I took stat at Duke, but below is my stab using a simple excel file (if I fat fingered an entry I apologize)
and I used the ESPN numbers which are apparently inaccurate for rebounds, so that may be another source of error...(I can email the file to anyone who wants to tinker with it...)

anyway, raw data on top, some stat stuff on bottom - we average more PPS on 3s than 2s on the season. the standard deviation on 3s higher than on 2s (.24 to .15) we have more bad nights from 2 than 3 (8 to 3) but the same number of really bad nights (2 to 2) and our worst night from 2 is worse than our worst night from 3.

our shooting from 3 is better than our shooting from 2 in both our losses and our games against ranked teams.

My opinion, while 3s have higher variance than 2s, the average return on 3's is so much higher that 2's still result in way more bad nights
(<.9PPS) than 3s so even with the higher variance of 3s, we are almost 3 times more likely to suffer a bad game from 2 than from 3 - keep bombing away!

On a side note, our recent offensive slump has been all about 3s, our 2s have been fine (in last 6 games, 5 to 1 bad from 3 (inlcuding our 2 worst from 3) compared to 3 to 3 from 2.

I would worry that defenses may have figured out how to suppress our 3's, but I read somewhere that usually comes in the form of limiting 3p attempts and our attampts have not gone down, just our shooting %.


I wish this was wysiwyg - I cant get it to format :(

GAME 2PM 2PA 2PPS 3PM 3PA 3PPS
DAV 25 33 1.515 13 21 1.857
KU 23 40 1.150 7 18 1.167 L,R
FAU 22 39 1.128 12 23 1.565
UNCA 21 31 1.355 10 23 1.304
ECU 22 43 1.023 5 14 1.071
UVM 18 30 1.200 10 27 1.111
ALA 16 36 0.889 6 15 1.200
ARIZ 21 44 0.955 4 14 0.857 L,R
MICH 20 33 1.212 8 23 1.043 R
WEBB 19 34 1.118 9 22 1.227
UCLA 20 32 1.250 11 32 1.031
EMU 14 31 0.903 10 29 1.034
ELON 22 47 0.936 10 19 1.579
ND 12 30 0.800 12 28 1.286 L
GT 15 31 0.968 9 20 1.350
CLEM 12 34 0.706 8 25 0.960 L
UVA 12 27 0.889 10 22 1.364
NCST 22 42 1.048 11 24 1.375
MIA 16 31 1.032 8 25 0.960
FSU 10 39 0.513 8 20 1.200
PITT 15 33 0.909 13 25 1.560 R
SYR 16 36 0.889 15 36 1.250 L,R
WAKE 15 27 1.111 12 26 1.385
BC 21 36 1.167 11 21 1.571
MD 13 30 0.867 5 24 0.625
GT 14 35 0.800 10 18 1.667
UNC 22 41 1.073 5 22 0.682 L,R
SYR 16 27 1.185 7 21 1.000 R
VT 14 29 0.966 10 33 0.909
WAKE 21 36 1.167 6 27 0.667
UNC
AVERAGE 1.024 1.195
STD DEV 0.1566 0.2410
MIN 0.513 0.625
MAX 1.515 1.857
#<.9 8 3
#<.8 2 2
#2PPS>3PPS 10
AVE IN LOSS 0.963 0.981
AVE V RANK 1.053 1.080

sagegrouse
03-07-2014, 09:43 AM
I wish this was wysiwyg - I cant get it to format :(


GAME 2PM 2PA 2PPS 3PM 3PA 3PPS
DAV 25 33 1.515 13 21 1.857
KU 23 40 1.150 7 18 1.167 L,R
FAU 22 39 1.128 12 23 1.565
UNCA 21 31 1.355 10 23 1.304
ECU 22 43 1.023 5 14 1.071
UVM 18 30 1.200 10 27 1.111
ALA 16 36 0.889 6 15 1.200
ARIZ 21 44 0.955 4 14 0.857 L,R
MICH 20 33 1.212 8 23 1.043 R
WEBB 19 34 1.118 9 22 1.227
UCLA 20 32 1.250 11 32 1.031
EMU 14 31 0.903 10 29 1.034
ELON 22 47 0.936 10 19 1.579
ND 12 30 0.800 12 28 1.286 L
GT 15 31 0.968 9 20 1.350
CLEM 12 34 0.706 8 25 0.960 L
UVA 12 27 0.889 10 22 1.364
NCST 22 42 1.048 11 24 1.375
MIA 16 31 1.032 8 25 0.960
FSU 10 39 0.513 8 20 1.200
PITT 15 33 0.909 13 25 1.560 R
SYR 16 36 0.889 15 36 1.250 L,R
WAKE 15 27 1.111 12 26 1.385
BC 21 36 1.167 11 21 1.571
MD 13 30 0.867 5 24 0.625
GT 14 35 0.800 10 18 1.667
UNC 22 41 1.073 5 22 0.682 L,R
SYR 16 27 1.185 7 21 1.000 R
VT 14 29 0.966 10 33 0.909
WAKE 21 36 1.167 6 27 0.667
UNC
AVERAGE 1.024 1.195
STD DEV 0.1566 0.2410
MIN 0.513 0.625
MAX 1.515 1.857
#<.9 8 3
#<.8 2 2
#2PPS>3PPS 10
AVE IN LOSS 0.963 0.981
AVE V RANK 1.053 1.080

This may be a bit better.

FWIW our worst games for 3-pt. shooting of the year were UNC, Maryland, and Wake -- no surprise that we lost two of those games. Arizona was our next worse.

niveklaen
03-07-2014, 09:45 AM
Sage - thanks!

nocilla
03-07-2014, 09:49 AM
What do the R's at the end stand for? I gather that the L's mean losses.

Edit; I figured it out, Ranked opponent.

Turk
03-07-2014, 10:35 AM
The variability comes from who is taking the shots, and where they are taking them (no duh).

Using NBA data, Kirk Goldsberry is doing some fun stuff and creating great shot charts.

http://grantland.com/the-triangle/courtvision-the-best-shooters-so-far-2/

http://courtvisionanalytics.com/

The variability of 2 pt shooting would be directly related to how many shots Duke gets in the paint. Midrange shots are evil, unless from the foul line area. Again, no duh. Unfortunately, I don't think that kind of shot chart data is available to civilians.

As for 3's, again, I would track individual players. Who is jacking more with less success, and from where?
And I would try to find a way to measure open threes (when feet are set, square, and the look is clean) vs. rushed threes (off balance, or challenged).

Some guys I am OK with taking rushed threes because they're that good of a shooter (Andre, TT, maybe Rodney). Other guys I think it's a bad shot (Jabari, Quinn, Sheed). I have no stats, that's just my proprietary "eye test / ulcer ratio" metric.

uh_no
03-07-2014, 12:56 PM
i ran some analysis a couple years ago, and the result were:

if you normalize for amount of shots taken, the variability of points made from three point shots is actually LESS than the variability of shots from 2, but, with that said, the distribution was skew left....which meant if we were to have an abnormal shooting night, it was far more likely to be a poor night than a good one

and those numbers obviously will change from team to team year to year, but it mostly fits with what you would expect....we'll generally shoot somewhere close to our season average most nights, but some nights we won't hit the side of the barn....because when we have those nights, it's often that the offense isn't working as efficiently as possible and thus we end up taking poor shots

i also did a further analysis on expected points from possessions starting with a 3 point shot or a 2 point shot (accounting for fouls and offensive rebound % as well) and our expected points per possession was higher for possessions starting with a 3 than with a 2

my conclusion was that you should always focus on taking good shots, and there if you are taking good shots, then you'll more often avoid the dreaded left side of the distribution (horrible nights), and if you are taking good shots, there is no reason to focus on not taking threes, as they are more valuable in the long run

CDu
03-07-2014, 01:10 PM
my conclusion was that you should always focus on taking good shots, and there if you are taking good shots, then you'll more often avoid the dreaded left side of the distribution (horrible nights), and if you are taking good shots, there is no reason to focus on not taking threes, as they are more valuable in the long run

Yes. A made 3 is roughly 50% more valuable than a made 2 :).

I agree as well with the idea that the most important thing is that it is a good shot (i.e., as open as possible and taken by a guy who has a good chance of hitting from there). For example, a wide-open 3 by Plumlee would be a terrible shot. On the other end of the spectrum, there's a reason we don't see Thornton take many shots within 18 feet of the basket.

If you have multiple good 3pt shooters, then working your offense to create good looks from 3 is an excellent strategy. Especially if you lack a good post scorer. If you have one or two guys who you can dump the ball to on the blocks and can hit 60% of their shots from there (and draw fouls), then it's a good strategy to go for more 2s in the paint.

As you said, it all comes down to the makeup of the team.

Listen to Quants
03-07-2014, 02:33 PM
For what it's worth, the standard deviation of a proportion (like shooting percentage) *assuming-each-shot-independent* is:

square root ((p x (1-p))/n) p=proportion of makes (1-p)=proportion of misses n=number of shots

Ends up very similar for a proportion of .33 (kind of ordinary 3 point rate) and .5 (kind of ordinary 2 point rate). Only gets weird when proportions get out to the extremes like .90

The variability of the number of points is higher for 3s since the makes are multiplied by 3 (insight! :) ) versus 2 to get the points. This in contrast to the variability or the shooting percentage which is similar. Think I have this right, but it has been far more than 20 years since my last stat course.

greybeard
03-07-2014, 10:35 PM
Gentlemen, the "when" of each defines the odds and comprises the game called basketball.

No premier needed. Step in 3s rock. To get them, pass penetration is best and will only work maximally if there is a bona fides threat to score at the rim, which still is a MUST for a variety of reasons. See Bobby Knight about having your big take it to the other guy's and getting the other guy's in foul trouble and getting easy "buckets."

Inside out play is the only left to the game other than high screen. All I ever read about here is about how Duke can't keep anybody out of the paint. Now it is unimportant?

You do not threaten to score at the rim and do so meaningfully, say Goodnight Gracie, which is why they have Parker parked down there battling two guys that are bigger than him as often as possible.

Anybody have numbers showing relationship between effective/ineffective inside play and effective/ineffective 3-ball.

K must have all different ways to juggle creating threats at the rim to generate 3-scores, or not needing to. How that computes defines how well Duke controls the game, the tempo of it, the confidence level of it, the essence of it. Like anything else, it's the how that numbers are produced that are the story, that define the game and what makes one team feel, look, perform, better. Just the way it goes.

Numbers happen, they create nothing. The game shapes how they are produced, which is why it is fun to play and watch.

uh_no
03-07-2014, 10:37 PM
Gentlemen, the "when" of each defines the odds and comprises the game called basketball.

No premier needed. Step in 3s rock. To get them, pass penetration is best and will only work maximally if there is a bona fides threat to score at the rim, which still is a MUST for a variety of reasons. See Bobby Knight about having your big take it to the other guy's and getting the other guy's in foul trouble and getting easy "buckets."

Inside out play is the only left to the game other than high screen. All I ever read about here is about how Duke can't keep anybody out of the paint. Now it is unimportant?

You do not threaten to score at the rim and do so meaningfully, say Goodnight Gracie, which is why they have Parker parked down there battling two guys that are bigger than him as often as possible.

Anybody have numbers showing relationship between effective/ineffective inside play and effective/ineffective 3-ball.

K must have all different ways to juggle creating threats at the rim to generate 3-scores, or not needing to. How that computes defines how well Duke controls the game, the tempo of it, the confidence level of it, the essence of it. Like anything else, it's the how that numbers are produced that are the story, that define the game and what makes one team feel, look, perform, better. Just the way it goes.

Numbers happen, they create nothing. The game shapes how they are produced, which is why it is fun to play and watch.

the closest i ever came was demonstrating there is no statistically significant correlation between 2pt% and 3pt%...which might be enough to counter the point.....it may be in he long term there is a small correlation, but i didn't see it in my season long sample size.

greybeard
03-08-2014, 07:05 AM
the closest i ever came was demonstrating there is no statistically significant correlation between 2pt% and 3pt%...which might be enough to counter the point.....it may be in he long term there is a small correlation, but i didn't see it in my season long sample size.

Would your analysis get to the strategy that I was trying to get at. Metaphorically, Frazer comes out and tears into Ali's body, then the left hook to the jaw becomes more available, he gets to time it better, orient his body a little better, because the arm has dropped.

In bb, you get it inside to Parker, he makes strong move after strong move to the rim. Even if some are stopped before they happen, and the first 3areblocked or sufficiently disrupted to miss, the arms are dropped to defend the body. A spurt of 3s drops, the inside defenders who have been working like crazy make a few errant passes 5 minutes later that give rise to early offense 3 that is open, and so on.

It seems quite clear that the game requires attacks at the rim, challenges against the rim defenders who are inside scorers at the other end, to take away their breath, their legs, and to get what was designed to be the "goal" of the game which was and remains to score the ball from as close to the goal as a person can get which now means looking down.

That was the nature of my question, whether the sophisticated flow sheets would show the impact of such punching combinations. Hard to imagine since not every inside attempt takes as much out of the defense, make or miss, as others, not every one comes off clever exterior/interior movement/play such as to assert dominion and break down confidence of opponent, not every attack causes the defense to reshape, makes outside defenders to make plays by dropping in, to step out of the ordinary, which might be insightfully executed one moment but, having broken with defensive strategy, the "shape" of the defense if we are talking Futball, such that the next time mistakes that leave openings that can be exploited.

The numbers, it seems to me, point to the efficacy of working an offense to get 3s of a quality that Duke does, but the quality and numbers are a byproduct of the whole, intertwined with all other aspects, which produces ebbs and flows, a great look (even a hand in the face can be one) from 3, create and run through a run, etc.

Your analysis is terrific stuff, it puts teeth around exactly why the free world knows that Duke will beat you with 3s. To create a game approach, offense noit separate from defense at the moment not separate from desired tempo, not separated from . . . is what makes K K.

slower
03-08-2014, 07:23 AM
Am I the only one who recalls hearing (in a recent broadcast) that "the committee" is considering a 4-point shot?

What a horrible idea.

sagegrouse
03-08-2014, 08:19 AM
Am I the only one who recalls hearing (in a recent broadcast) that "the committee" is considering a 4-point shot?

What a horrible idea.

I heard it also -- but it was the NBA, and I believe it was a speculative idea of the new commish, Dukie Adam Silver.

slower
03-08-2014, 08:33 AM
I heard it also -- but it was the NBA, and I believe it was a speculative idea of the new commish, Dukie Adam Silver.

Thanks, Sage.