That was awesome uh no. The combination of standard deviation and skew really capture what we're looking for. It'd be cool if kenpom could look at these stats nationally across all teams.
Printable View
That was awesome uh no. The combination of standard deviation and skew really capture what we're looking for. It'd be cool if kenpom could look at these stats nationally across all teams.
if anyone is interested:
for last years team
3 pt:
37% or 1.11 ppa
6.3 points stddev
-11.4 skew
2pt:
52% or 1.04ppa
8.6 points stddev
5.3 skew
So the only difference is that last year's teams offensive output on the whole was much more up and down...again, i'm not sure the predictive value here...because individual seasons and the tournament in general are such small sample sizes.
Didn't want to start a whole new thread on this but it seems that we are a different team when Andre isn't getting his shot. We drive it to the hole much much more. He seems to be our lynchpin and I think the word is out. When he's allowed to shoot we can blow teams out.
Yes, I agree to an extent. But in 10 min he had 3 turnovers and 0-3 from the field. I think the 3 TOs led to reduced minutes. He can do some things that will "allow" him to shoot more, and good defense + low turnovers are a couple of them.
Boy, looking at the box: we had 7 assists (on 20 made shots) and 12 turnovers. Hmm.
With Austin's speed, you would think he could be a little ball-thief, leading to easy buckets. He had no steals, and averages less than one per game. We average 6 steals a game but only had 2. We rank #190 in steals per game. I miss our ability to fast break on steals! Bring back Showtime to Durham!
This is very cool analysis. Thank you.
I bet we're unusual about our 3-point shooting having more variance than our 2-point shooting, which argues (as you point out) that taking a lot of 3-point shots makes sense for us. My guess is the skew is more typical, and probably gives life to the "live by the three, die by the three" mantra. Especially when our poor games have occurred in 28% of our games this season (8 of 29 so far we've shot worse than 30% from 3).
I'm really liking this stats debate over the value of threes or twos, but it seems like some of the numbers are focusing more on the individual shooting the ball than the overall impact on how the team does. I'd agree with Edouble that a guy shooting a two is more likely to get his own rebound for a second shot opportunity than a guy shooting a three, but isn't that kind of irrelevant? Wouldn't the more important stat be whether Duke as a team is more likely to get an offensive rebound on a three or a two? If Miles misses a shot from close and gets his own rebound that's great, but it's not any better or worse than Austin missing a three and Miles getting the offensive rebound. I don't have the numbers to back it up, but I would think that threes and long twos are more likely to lead to offensive rebounds that shots from inside 8-10 feet. But on the other hand, it seems like missed threes also lead to more fast breaks for opponents than missed twos because of the long rebounds. I'm not sure what stats would be important to figuring out the team impact of threes v. twos, but I think the points-per-shot numbers someone gave earlier in the thread is a great starting point. Another thing that might be helpful is points-per-possession on possession that begin with a two or a three point shot. Although it might get dicey with multiple shot possessions, it might be an indicator of how much we're looking to create certain shots or taking what the defense gives us.
I think that would be a project for someone not me :P to go through game logs and look at the results of our missed shots: did we get the rebound and a quick score (tip in/kick out) did we get the rebound and reset the offense, did they get the rebound and fast break, and did they get the rebound and reset. I think that while we may be able to theorize about one being better than the other, I think we'll find the correlation between the shot type and us getting the offensive rebound is insignificant.
Fine ill do it:
we offensive rebounded 7 of our 14 missed twos 50%
we offensive rebounded 5 of our 18 threes 27%
so in our sample size of 1, we rebound our missed twos at a significantly higher rate...almost twice as much in fact
now for an expected value analysis:
E of 3 pt shot: .25*3+.75*.27*Epos
E of 2 pt shot: .5*2+.5*.5*Epos
Epos: .46*E3pt+.54*E2pt
Solving this all we get
approx 1.4 pts per possesion (not counting fouls)
1.03 pts per possesion which starts with a 3 point shot
1.35 pts per possesion which starts with a 2 pointer
This will vary depending on our shooting for the day
We got fouled on 0 three point shots today...and i'd include the information about missed shots from 2 pt fouls, but ESPN doesn't indicate whether the foul shots were on a shooting foul or in the bonus, but in this case, they would only increase the lead that 2 point shots had over 3 point shots.
Going forward, I may try to do this on a game by game basis...counting fouls and offensive rebounds, what is our expected ppp filtered by the shot it starts with
I think the point is that a crucial component of any good offense is the ability to attack and make the opposition defend the rim. As long as you have the ability to do that, via slashing or post play, then it's fine to bomb away and ride the hot hand; the three-pointer is a great long-term option as far as upside and percentages go, but it's undebatable a higher-risk/higher-reward option than "safer" shots near the rim. I also think it's kind of arguing semantics to argue that that kind of offense is the same sort of "lottery"; if a you're getting to the rim and drawing contact, it's simply up to you to will the ball in the hoop or draw fouls; it becomes a game of power and intensity rather than skill and touch. Different scenario than launching the ball from 20+ feet away with a defender flying towards you and hoping you didn't kick your arm out 1/4" too wide on the release.
Anyway, if you're a less talented team looking to pick off a few superior squads, high-risk/high-reward is great. If you consider yourself a top dog and complete team, though, it's a bit risky to put all your eggs in that one basket; season-long percentages don't matter much in a single game, and there's a lot of game-to-game variance in 3-point shooting %s. This means you're much more vulnerable on your bad nights if your offense is mostly 3-point shooting... which, really, is the measure of most elite teams. Case in point, we've shot 28/83 from 3 in our 4 losses, good for 1PPS.
There's also the fact that shooting a live-action 22-footer isn't a pure percentage play like gambling with cards or something. There's a ton of human element that goes into it, and there's never any guarantee that the percentages will revert to a previous mean. Every shot is unique and not influenced by past results whatsoever.
However...
I think that this team has a pretty good balance. Rivers can get to the rim, as can Curry when he's healthy. Mason can get 4-8 footers pretty consistently (aside from from the past few games), and Kelly makes a couple good moves to the rim each game. But more interestingly, we've got 4 guys who can absolutely light it up from 3-point land, plus Thornton (who can get hot) and even Cook (who, I swear, can shoot...). This is a whole different ballgame than the Redick days, when we were pretty much screwed if he was off; this team can have two or even three of it's prolific shooters struggle on a given night and still have a guy or two with a hot hand from beyond.
It's sort of like playing the lottery, but buying a few hundred thousand tickets. :p
I am totally digging this analysis and accompanying discussion, but there's also a factor involved called THE OTHER TEAM'S DEFENSE. We're not just playing horse and randomly choosing whether to shoot twos or threes. And variations in shooting percentages are not only attributable to luck. The other team has a lot to do with it, and Duke in particular will adjust their offense according to what the defense gives.
I was surprised by the offensive rebounding numbers. I really thought threes were more likely to allow an offensive board
Its one game, could be an outlier for all i know. I think the "defense argument" is normalized by the fact that better three point defense should result in fewer three point shots. Thus you can say that your shots are of good quality, at least the three point ones. Now, we know this is not always true. I think the lack of correlation between three and two point shooting is important. It means that if we're getting shut down outside, it doesn't mean we're getting shut down inside. THe idea is that you can sort of assume that we aren't being forced into taking bad shots.
IF a team has good three point defense, we shoot more twos and the formula accounts for the decreased shot attempts...and we still only take the good ones. If A team is shutting us down in both cases, then shots will all be equally 'bad' and thus the 3pt number is still meaningful...since it indicates that we can be forced into bad shots allowing for abnormally low %s.
I hope that makes some sort of sense. I'm sometimes better with numbers than with explaining the numbers :)
Actually, given the propensity of 3 point shots to become offensive rebounds, it's oddly a severely incomplete analysis by Pomeroy. Back in the dark ages before stat29.com and kenpom.com (or anything.com), there was a local statistics prof that proved on the Coach K show that EVERY shot attempt should be a three pointer. We only do that sometimes.
As others have already pointed out, though, any statistical analysis fails to account for a motion offense philosophy of working the ball around, inside, outside, off the dribble, off the pass, etc. - all for the purpose of working for a "good" shot - regardless of distance from the basket.
I disagree with you with your conclusion: the game would still be fun to watch.
But I was bored at lunch, so decided to look at Dawkin's three point shooting this season. Dawkins has been described as streaky, and lethal "when he gets hot"
According to all the game trackers on ESPN, he shot 66/158 this season -- 41.8%.*
This season, after a miss he is 28/66 (42%). After a make he is 38/91 (42%).
After two misses in a row, if he shoots another in the same game, he is 10/25. After two makes: 11/24. After a miss and a make: 23/54.
After missing three in a row, if he shoots another in the same game, he is 3/10. After three makes... 3/10.
I'm sure if we took this out over all of his games with Duke we'd see a similar pattern: whether he has made his previous shots has no predictive value whatsoever on whether he'll make his next one.
* I think two attempts might be missing there, but I am not going to go through all the games to find it.
Extremely a propos.
Scoring efficiency by NBA court location:
Attachment 2450
Awesome graphic, throaty. It's interesting how much more efficient shots from in front of the basket are than from the wings. Also, corner threes are as efficient as dunks, but short two-point shots from the corners are awful. Long two-point jumpers are not good unless they're straight on.