Duke, 3P% Defense, and Luck

[SCMatt33]

So after enough mid and low tier ACC teams seeming to use some kind of black magic to become significantly better at shooting 3's against Duke, I decided to see just how much that was actually the case vs. "woe is me" paranoia. It turns out that unranked ACC opponents really are using black magic, or are more likely just getting "lucky," while ranked ACC opponents are terribly "unlucky" from downtown against Duke. Here's some basic numbers of Duke's 3P% defense (or opponents 3P%) on the year:

Season: 148-454: 32.6%
ACC Play: 92-249: 36.9%

Now on the surface, those don't seem like entirely unreasonable numbers. After all, ACC play is by nature going to be tougher than non-conference play, but 4% is a fairly big swing. In fact, over the KenPom era, I can only find 2 other years in which that swing is bigger than about 2%, 2007 and 2014. That sounds pretty bad, but also keep in mind that in 2012, the 3P% defense was about 2% lower in ACC play, so those years are more likely a coincidence rather than a "sky is falling, we're losing in the first round" prediction. In fact, over the last few years KenPom has looked at the subject of 3 point defense in quite some detail and found that independent of other factors, namely luck, and good 2P% defense forcing tougher 3's, a team has fairly little control over opponents 3P%, with even the best teams out there only affecting it by about 3% (http://kenpom.com/blog/index.php/weblog/entry/one_last_post_on_3p_defense) from what their opponents shoot on the year.

For the really crazy stuff, you have to break Duke's numbers down even further:

Non Conference: 56-205: 27.3%
Ranked ACC: 25-90: 27.8%
Unraned ACC: 67-159: 42.1%

Obviously, one of those things is not like the others. You can pretty easily rattle off the games that caused this spike, State, Miami, GT, FSU, Cuse, VT. Duke's defense has looked pretty bad in those games, but those guys all shot 3 beyond any reasonable expectations. On the other side, you think of Louisville, UNC, and UVA. All of those teams shot well below their season numbers from 3 (all shot under 24% from 3). I certainly don't think that Duke played a much higher level of defense in those games, but there was a certain amount of luck on which games teams knocked down their 3's. From that standpoint, Duke has been pretty lucky this year. For the most part, the opponents who shot the ball ridiculously well were the ones that we could afford to give up a few extra points to and still win the game.

All of this is to say that Duke's perimiter defense probably isn't as bad as it looked yesterday against VT, but it's also not as good as it looked against Clemson or Notre Dame. Overall, Duke's troubles defending the paint, will probably leave some open 3's for teams that want them, but Duke's tremendous inside scoring is also going to give us a lot of open looks, and with the talent Duke has, I like our chances against most teams when both have similar shooting nights.

[Kedsy]

I certainly don't think that Duke played a much higher level of defense in those games, but there was a certain amount of luck on which games teams knocked down their 3's.

Cool stats, thanks. The differences between the three numbers on your chart are pretty crazy.

I agree with your overall premise, and I agree there is some luck involved. But I disagree with the above quoted statement. I think Duke absolutely played a much higher level of defense against Louisville, Virginia, and UNC they we did against State, Miami, GT, FSU, Syracuse, and VT. It's conceivable, given our youth, that we put out a different amount of defensive effort in big games than we do in games we expect to win, and that once our opponent's confidence is high, we find it difficult to stuff the genie back into the bottle. If so, it might bode well for our upcoming tournament games, at least after the first round.

[Seattle Hoo]

Virginia's shooting percentage was within a much more extensive trend. Other than the UNC game, before last night, Virginia had been horrendous from the arc for quite a long stretch of games encompassing the Duke game. Also, three-point shooting is the primary offensive weakness of both Louisville and UNC.

What it all probably says is that Duke's three-point defense is pretty bad, but when the team is inspired, it can play decent three-point defense. But given sample sizes, I don't think holding those three teams below their season average in any one game says a whole lot about the defense.

[AIRFORCEDUKIE]

So all we gotta do is get out of the first weekend and play the really good teams that survive and the numbers tell us we will play good D against them. If we stay a one seed that sure would help a lot.

[Kedsy]

But given sample sizes, I don't think holding those three teams below their season average in any one game says a whole lot about the defense.

I agree. But after watching the games, I do think Duke played better defense against Louisville, Virginia, and UNC than we did, for example, last night.


What it all probably says is that Duke's three-point defense is pretty bad, but when the team is inspired, it can play decent three-point defense.

I disagree with this. As the OP's data shows, in non-conference games, we held opponents to 27% in 200+ attempts. I don't think that can be solely attributed to luck.

[SCMatt33]

Virginia's shooting percentage was within a much more extensive trend. Other than the UNC game, before last night, Virginia had been horrendous from the arc for quite a long stretch of games encompassing the Duke game. Also, three-point shooting is the primary offensive weakness of both Louisville and UNC.

Those teams are certainly not lights out from 3, but those teams definitely shot well below their averages against Duke. Louisville has shot 31.7% from 3 in ACC play. They went 4-25 (16%) against Duke. UVA has Shot 33.5% in ACC play. They went 3-13 (23.1%) against Duke. UNC has shot a surprisingly decent 35.6% from 3 in ACC play. They went 2-10 against Duke. So even for their meager averages, they had bad games against Duke. I'll definitely agree that Duke's D bothered UNC given how much they dogged Paige, but Louisville just had a plain bad game. Duke's zone certainly made them take more 3's than they otherwise would have, but an average game for Louisville would have resulted in an extra 3 or 4 makes. I don't think Duke had much to do with just how awful they were. By the same token, I don't think that bad defense is what suddenly made FSU a 50% shooting team, they just had their best game of the year against Duke. It's definitely not all luck, and there is a bit of "turning the switch" with Duke, but there's definitely an undeniable luck factor involved.

[gus]

This and the "Quinn Cook - elite shooter?" thread make think of a simple question with a hard answer. I'll put it here rather than in a new thread:

What is a good three point shooting percentage (for an individual / for a team)?

"the higher the better" isn't right, and I point to the discussion about JJ to support that. Redick could easily have shot above 50% for his Duke career. There's no doubt in my mind that had he waited for shots and been fed the ball in good shooting position, he would have made more three point shots than he missed. But Duke would have won far fewer games.

A simplistic answer is that 33% from deep is equivalent to 50% from inside the arc, but I'm not sure that's right either. It's obviously far more complicated than that. I don't think a 33% shooting team or individual stretches a defense, for example. A 40% shooter does. And maybe a 50% shooter is being too patient.

Answering this would be really hard, and I don't have the time or patience to even attempt to do an analysis. So instead, I'm just throwing it out there for other people to chew on.

[devilsadvocate85]

Is there some reason we have decided not to include Notre Dame as conference games?

[The Gordog]

This and the "Quinn Cook - elite shooter?" thread make think of a simple question with a hard answer. I'll put it here rather than in a new thread:

What is a good three point shooting percentage (for an individual / for a team)?

"the higher the better" isn't right, and I point to the discussion about JJ to support that. Redick could easily have shot above 50% for his Duke career. There's no doubt in my mind that had he waited for shots and been fed the ball in good shooting position, he would have made more three point shots than he missed. But Duke would have won far fewer games.

A simplistic answer is that 33% from deep is equivalent to 50% from inside the arc, but I'm not sure that's right either. It's obviously far more complicated than that. I don't think a 33% shooting team or individual stretches a defense, for example. A 40% shooter does. And maybe a 50% shooter is being too patient.

Answering this would be really hard, and I don't have the time or patience to even attempt to do an analysis. So instead, I'm just throwing it out there for other people to chew on.

Long shots lead to long rebounds which lead to fast breaks. I think 37% or better is "good" 3 point percentage, but it all depends on what you are comparing to. Is your best option in the paint is a 50% shooter or 68%? How good is your opponent on the break?

[SCMatt33]

Is there some reason we have decided not to include Notre Dame as conference games?

They're games are included in my numbers for sure. The ND games are the only reason that the "ranked ACC" numbers aren't under 20%. They shot well when they beat us, but not Completely out of touch with their season numbers, about +5% or 6%, with the opposite number in the game Duke won. I guess they would really represent the "average luck" type of game. They shot pretty normal across two games, with Duke coming close in their house, and crushing them in ours. The luck seemed to match the level of defensive play in those games.

[Bluedog]

Long shots lead to long rebounds which lead to fast breaks. I think 37% or better is "good" 3 point percentage, but it all depends on what you are comparing to. Is your best option in the paint is a 50% shooter or 68%? How good is your opponent on the break?

True, but long shots also lead to a higher OR% typically since it's more of luck who ends up with the ball. So, you perhaps get more second chance opportunities on offense when shooting 3s.

[BluDvlsN1]

Duke Basketball (@dukebasketball)
2/26/15, 8:23 AM
"I’ve had to really tone down the practices. As a result, you get slippage"—K on Duke's defensive struggles Wednesday
chron.it/1DsgZyb

[captmojo]

Sometimes you're the windshield. Sometimes you're the bug.

Sometimes an opponent breaks out of his/her shell of mediocrity and the ball ends up in places (through the net) that completely surprises the shooter, no matter how well they are defended. Stuff does happen, no matter how well you prepare against it.

Sometimes, the other guy delivers a shot that's so bad, that even if you are in the best percentage chance for a rebound with your opposing number behind you, the ball comes off the rim so unpredictably long that it goes over his/her head and even they can't reach the carom. We saw this against unc. Also it happened during the Clemson game. They may as well been trying to shoot with a football.

[bob blue devil]

Fwiw, according to ken Pom, 3p defense is nearly the same as free throw defense. Good defense discourages the shot, but does not reduce the percent made.

http://kenpom.com/blog/index.php/weblog/entry/3_point_defense_should_not_be_defined_by_opponents _3p

[Seattle Hoo]

Fwiw, according to ken Pom, 3p defense is nearly the same as free throw defense. Good defense discourages the shot, but does not reduce the percent made.

http://kenpom.com/blog/index.php/weblog/entry/3_point_defense_should_not_be_defined_by_opponents _3p

As tbyers demonstrated earlier, I don't get advanced statistics enough to know what the heck he's talking about. So making opponents take contested threes doesn't effect their percentage over letting them take those open, in-rhythm shots? I also don't get what the pre-12/4 and post-12/4 distinction means. Sorry, I'm a words nerd, not a numbers nerd.

[BlueDster]

As tbyers demonstrated earlier, I don't get advanced statistics enough to know what the heck he's talking about. So making opponents take contested threes doesn't effect their percentage over letting them take those open, in-rhythm shots? I also don't get what the pre-12/4 and post-12/4 distinction means. Sorry, I'm a words nerd, not a numbers nerd.

I would think that this is the case because usually a well defended 3 point shot is not taken, rather than forced, except for in time pressure situations. If you're defending the three really well as a team, you're likely to stop another team from shooting as many threes rather than the other team taking the same number of threes they always take but more of them are contested.

[bob blue devil]

As tbyers demonstrated earlier, I don't get advanced statistics enough to know what the heck he's talking about. So making opponents take contested threes doesn't effect their percentage over letting them take those open, in-rhythm shots? I also don't get what the pre-12/4 and post-12/4 distinction means. Sorry, I'm a words nerd, not a numbers nerd.

Sorry to hear about your disability.
I can't say with confidence why. I speculate that there aren't teams regularly giving up uncontested 3s (if it happens, it becomes a defensive priority pretty quickly) and that shooters have a reasonable internal gauge of when a defense has reduced their ability to hit a shot vs the norm and exercise discretion in a fairly consistent way. But that's just a guess.

[SCMatt33]

As tbyers demonstrated earlier, I don't get advanced statistics enough to know what the heck he's talking about. So making opponents take contested threes doesn't effect their percentage over letting them take those open, in-rhythm shots? I also don't get what the pre-12/4 and post-12/4 distinction means. Sorry, I'm a words nerd, not a numbers nerd.

So the link posted by Bob was KenPom's initial analysis on the subject. His initial method was to show that the best and worst 3 point defensive teams during the first part of the year (before December 4) showed very little difference the rest of the year. After December forth, the teams who were in the top 20 were less than 2% better than the teams who had been in the bottom 20 over the rest of the season. Now over the course of the next year, after hearing a lot of complaints from people (namely Syracuse fans who insisted that the zone forced bad 3's), he revisited the data a couple of times and was able to draw out some more definitive conclusions and changes his stance a bit. His final stance as stated in the post that I originally linked:


Anyway, to hopefully conclude this series, here are my opinions in bullet points.

- The offense is largely in control of the quality of 3-point shots it takes.
- These decisions are affected by the quality of the opposing 2P% defense
- 3P% is also influenced by effective challenging of shots.
- All of that can add up to about a 3% swing from average.
- So 3P% defense is not totally random
- But a defense has considerably more direct impact on 2P% than 3P%.

So he's essentially saying that you can affect 3 pointers a little by good challenging, but that bigger components are the luck of the shots, and importantly in the case of Duke, good 2P% defense dictating the types of 3's that teams settle for. In Duke's case, poor 2P% defense allows teams to not settle so often, and they more often only take good 3's because they know to pull the ball back out and wait for a better shot if Duke closes out.

[UrinalCake]

I was hoping to see a comparison of how teams shot the three against us versus how they shot it the rest of the season. But then of course you have to consider that maybe our three point defense just isn't that good.

The game that really stands out to me as being "unlucky" was Georgia Tech at home. We left them open because we knew they were horrible at shooting the three, and they went on to hit something like 8-11 against us.

[Listen to Quants]

<snip good stuff>

In fact, over the last few years KenPom has looked at the subject of 3 point defense in quite some detail and found that independent of other factors, namely luck, and good 2P% defense forcing tougher 3's, a team has fairly little control over opponents 3P%, with even the best teams out there only affecting it by about 3% (http://kenpom.com/blog/index.php/weblog/entry/one_last_post_on_3p_defense) from what their opponents shoot on the year.

<snip more good stuff>



Sorry to pick on a lesser point from an excellent post, but I am not convinced that KenPom is right in the above idea. The problem is if 3P defense is a small, and 3P offense a strong, contributor to 3P% (KenPom thesis) then team 3P defense %, across all teams should show far less variance than 3P offense % (since a team faces a wide variety of 3P offenses which would trend toward the mean leaving only the small variance from 3P defense; each particular teams 3P offensive ability, however, is there in all games and only the, relatively unimportant, opponent's 3P defense averages toward the mean across many games.)

My quick look at the spread of 3P% defense and offense this year found them about equal.

[Seattle Hoo]

I'm having a hard time grasping what KP is saying here, and the validity of his 12/4 dividing line. I looked at the ACC statistics, and using conference-only stats (all of them after 12/4), Virginia has a 3FG% defense of .291, while Pitt is last at .390, which looks like more than a 2% difference. Virginia has a FG% defense of .361 with Pitt having .467, so the 3FG% and FG% defenses pretty much track.

Moreover, I thought it was said that good 3FG defense would deter shots rather than affecting them, so is it counter-intuitive that Virginia's opponents have attempted 35% of their shots from the arc while Pitt's have only attempted 30% of theirs from the arc? I guess that could happen because Virginia's 2-point defense is SO good that opponents have no choice but to shoot contested threes.

[SCMatt33]

Sorry to pick on a lesser point from an excellent post, but I am not convinced that KenPom is right in the above idea. The problem is if 3P defense is a small, and 3P offense a strong, contributor to 3P% (KenPom thesis) then team 3P defense %, across all teams should show far less variance than 3P offense % (since a team faces a wide variety of 3P offenses which would trend toward the mean leaving only the small variance from 3P defense; each particular teams 3P offensive ability, however, is there in all games and only the, relatively unimportant, opponent's 3P defense averages toward the mean across many games.)

My quick look at the spread of 3P% defense and offense this year found them about equal.

No need to apologize. I wouldn't post things if I didn't want a good conversation about them. I thinks there's been a bit of confusion about exactly what 3 point defense is referring to in this case. In my original post, I merely used KenPoms analysis about luck to help highlight how Duke has seemed to face some bad luck running into season best shooting night from some lesser teams, and some poor shooting nights from better teams. These feed the notion that Duke is playing to the level of its competition. While I generally agree with that notion, some of this luck based stuff has exaggerated this effect to a significant extent.

KenPom's analysis, while relevant to the conversation, is about a slightly different phenomenon. His research relates to how much a team's ability to defend 3 point shots (i.e. challenging shots effectively) has on an opponents 3 point percentage. His original assumption was that because the offense largely controls the shots that they take, a team's ability to challenge has a much greater affect on the number of 3 pointers attempted (3PA/FGA) than it does on the actual 3 point percentage.

After a few iterations of his analysis, he found that not to be entirely true, but that successfully defending the 3 point line has a small, but significant effect. He also shows that based on very high variances from one part of a season to another that seems to exhibit randomness across all of college basketball as compared to other stats, 3 point defense has a higher element of luck. While the overall variance between the best and worst teams is about the same, the degree to which luck determines who is where within that variance is higher.

He also later introduces a third element into the equation, showing that strong 2P% defense has a fairly sting correlation with 3P% defense. That is to say that 3 point defense does not exist in a vacuum. Teams with strong 2 point defense will often be able to force teams to settle for much tougher 3's. The converse is also true for teams with less than stellar 2 point defense. This is case with Duke. Duke's bad defense on drives has led to a lot of open 3 pointers. Luck has certainly played into it as well. Neither GT or Louisville has great three point shooting, but both got open looks against Duke. There's definitely some luck involved with Lousiville shooting them at below their season average, and GT shooting it at well above their season average.

[Listen to Quants]

No need to apologize. I wouldn't post things if I didn't want a good conversation about them. I thinks there's been a bit of confusion about exactly what 3 point defense is referring to in this case. In my original post, I merely used KenPoms analysis about luck to help highlight how Duke has seemed to face some bad luck running into season best shooting night from some lesser teams, and some poor shooting nights from better teams. These feed the notion that Duke is playing to the level of its competition. While I generally agree with that notion, some of this luck based stuff has exaggerated this effect to a significant extent.

KenPom's analysis, while relevant to the conversation, is about a slightly different phenomenon. His research relates to how much a team's ability to defend 3 point shots (i.e. challenging shots effectively) has on an opponents 3 point percentage. His original assumption was that because the offense largely controls the shots that they take, a team's ability to challenge has a much greater affect on the number of 3 pointers attempted (3PA/FGA) than it does on the actual 3 point percentage.

After a few iterations of his analysis, he found that not to be entirely true, but that successfully defending the 3 point line has a small, but significant effect. He also shows that based on very high variances from one part of a season to another that seems to exhibit randomness across all of college basketball as compared to other stats, 3 point defense has a higher element of luck. While the overall variance between the best and worst teams is about the same, the degree to which luck determines who is where within that variance is higher.

He also later introduces a third element into the equation, showing that strong 2P% defense has a fairly sting correlation with 3P% defense. That is to say that 3 point defense does not exist in a vacuum. Teams with strong 2 point defense will often be able to force teams to settle for much tougher 3's. The converse is also true for teams with less than stellar 2 point defense. This is case with Duke. Duke's bad defense on drives has led to a lot of open 3 pointers. Luck has certainly played into it as well. Neither GT or Louisville has great three point shooting, but both got open looks against Duke. There's definitely some luck involved with Lousiville shooting them at below their season average, and GT shooting it at well above their season average.

Bolded part is why I didn't totally want to drift onto that tangent, but I did because I just don't think anyone should take that specific KenPon analysis as definitive. It predicts, as I said, a unequal distribution of Def. and Off. 3P%s. Those fail to occur.

[Ima Facultiwyfe]

Whatever became of the notion of allowing the team with the lead to take the ball out of bounds OR take the foul shot during the final minute. Does that idea still have legs?
Love, Ima

[captmojo]

"I believe we had an agreement there would be no math." ;)

As one who abhors statistical mathematics, I'm just going to say that I believe in the power of good fortune.
Luck it is...and was.

[SCMatt33]

Now that some of the dust has started to settle on the season being over, I thought it was a good time to wrap up this thought. At time I posted it, Duke was at what would turn out to be its defensive low point of the year immediately after the Va Tech game. At the time, Duke was ranked in the 80's in adjusted defensive efficiency, and was giving up 32.6% shooting from 3 on the year, and nearly 37% shooting from 3 in ACC play. From that point on, however, Duke only gave up 28.5% shooting from 3 for the rest of the year, and only a single team (@UNC) shooting over 35% in any game.

Breaking down the numbers a bit shows that luck likely played at least some role in the turnaround, though in this case, it is more of a regression to the mean given how bad Duke's luck had been earlier in this category. For the season, Duke's opponents shot 27.8% of their field goal attempts from 3 point range. In the final stretch of the season (vs. Cuse through NCAAT), they shot 28.5% of their field goal attempts from 3. So for the most part, Duke did nothing to run teams off of the 3 point line more than before. Teams took just as many 3's on the whole, they just made less of them. So was it forcing tougher 3's that caused the turnaround, or simply good luck that teams made less of them. Personally, I think it would be a bit naive to say anything other than both. I've seen a guy like Luke Winn break down screen shots from games to determine how many were "open looks" before, but with 633 attempts from opponents on the season, I don't have the time nor the inclination to perform such a task.

Anecdotally, I felt that Duke did a better job of contesting 3's overall. The parade from the 3 point line that FSU and Va Tech had was ridiculous, and we certainly made a concerted effort to close out and rotate better. I also though that Jah provided much better help in the back at the end of the year. The match up nightmare against Kaminsky notwithstanding, he did a good job of challenging shots when guards were beat on a drive. The guards had a lot more confidence to push out on the perimeter and try to force turnovers and really did a better job of making teams uncomfortable because of it. Having a harder time getting two pointer off sometime forced teams to really settle for bad threes. I think we really saw that manifest itself in the second half of the championship game, where the guards suddenly have trouble driving and Amile did a great job of containing Kaminsky in the post.

All of that being said, it's hard to ignore some of the luck. San Diego State, Utah, and Gonzaga all missed some very make-able three's in those contests, and while San Diego State wasn't a great shooting team to begin with, it didn't stop them from going off against St John's, nor did it stop GTech (a very poor shooting team) from having a huge game against Duke earlier in the year. With Utah and Gonzaga, you might be able to say that NRG had a bit to do with it, but there were still some missed shots that they might normally get.

This isn't to say that "Duke just got lucky" or anything. Quite the opposite, I think these kind of games go to show just how unlucky Duke had been earlier in the year. I don't think Duke's 3 point defense was ever quite as awful as they looked against Va Tech, nor do I think that it was as tremendous as it looked against Gonzaga, but I do think that things worked out to where the final numbers look just about right. Also, the rest of the defense provided a nice assist.

Of coarse the saddest thing is now having to talk about this team in the past tense. At least we'll have that nice, big banner to remember them by;)

[uh_no]

Whatever became of the notion of allowing the team with the lead to take the ball out of bounds OR take the foul shot during the final minute. Does that idea still have legs?
Love, Ima

it never had any legs because the other team would just foul again and the first team would just take it out of bounds again (meanwhile risking a turnover), so they would wait until a good shooter got fouled and then go to the line. so effectively you're just letting the team with the lead choose who they want to shoot...which is a reasonable thing, but why go through the trouble of making them inbound and the other team foul over and over?

[captmojo]

... having to talk about this team in the past tense. At least we'll have that nice, big banner to remember them by;)

This most important part, should never be overlooked.:)

[MarkD83]

Of coarse the saddest thing is now having to talk about this team in the past tense. At least we'll have that nice, big banner to remember them by;)

Keep in mind that every year Duke team's are different and every past team is talked about in the past tense.