PDA

View Full Version : Duke's Defense Charted (Courtesy of Duke Hoop Blog)



loran16
02-16-2012, 02:07 PM
Tommy D over at Duke Hoop Blog has been charting Duke's Defensive Performance this season ever since the Western Michigan game. You've probably see him post over here (I'm assuming "Tommy" on this board is the same guy, since he posts about the columns here) with some additional facts on our D.

What do we mean by charting? Well Tommy's been using a guideline from David Hess to go over basically EVERY play in each Duke game and to thus record who was responsible (directly) for the results of each possession on the defensive end. So Tommy's recorded how often does each Duke player get a "stop" (0 points) whenever his opponent tries to beat him for a shot, how often does each Duke player force a turnover - including charges and other forces not recorded in the box score -, how often each defender is attacked by opposing teams, and an overall defensive rating for each player.

This is really really good stuff, and now more importantly, Tommy posted (well yesterday) the cumulative results of this charting up from W.Mich. to the Maryland game. http://www.dukehoopblog.com/2012/02/15/project-defensive-scoresheet-almost-cumulative-stats/#comment-4029 It's really in depth stuff so you should all read it.

That said, some thoughts:
1. Andre's D is really really a problem. He doesn't force turnovers, opponents are shooting over 60% against him, and is last on the team by far in getting stops. More importantly, opponents seem to notice this - opponents attack him on a greater percentage of possessions than any other guard when Dre is in the game. The other defenders aren't seemingly doing a great defensive job, but Dre has been just miserable.

2. TT actually is pretty solid on Defense and is far better than Quinn Cook so far....but he's not notice-ably better than either Seth or Austin, contrary to what one would think. If we're on D holding on to a late lead, you'd rather he be in than Dre for sure (it's not even close). But he's not stopping our defensive woes (And he needs to stop fouling, but that's not covered here).

3. On that matter, Curry and Rivers are actually decent defenders.

You all should read this - considering how terrible our D has been this year (well terrible compared to typical Duke Defenses), it's a good analysis of what's gone wrong.

Obviously

Kedsy
02-16-2012, 03:48 PM
Hmm. I don't understand Andre's stop percentage in the article. According to Tommy's own post (http://www.dukebasketballreport.com/forums/showthread.php?27581-Our-Defense-vs.-Maryland-%28Cameron%29-and-a-few-extra-numbers-too), Andre's cumulative stop percentage is just under 50%. Now, that's not great, but it's way better than 32%.

Josh's and Quinn's stop percentages in the article are also way lower than in Tommy's post here, and Mason's and Michael's are much higher. Everyone else is close to the same.

Tommy, if you're reading this, please explain the discrepancy. If Andre's stop percentage is 50% instead of 32%, I imagine his overall DRtg would be closer to the rest of the team.

tommy
02-16-2012, 05:34 PM
Hmm. I don't understand Andre's stop percentage in the article. According to Tommy's own post (http://www.dukebasketballreport.com/forums/showthread.php?27581-Our-Defense-vs.-Maryland-%28Cameron%29-and-a-few-extra-numbers-too), Andre's cumulative stop percentage is just under 50%. Now, that's not great, but it's way better than 32%.

Josh's and Quinn's stop percentages in the article are also way lower than in Tommy's post here, and Mason's and Michael's are much higher. Everyone else is close to the same.

Tommy, if you're reading this, please explain the discrepancy. If Andre's stop percentage is 50% instead of 32%, I imagine his overall DRtg would be closer to the rest of the team.

Here's the difference: On the charting in which the Dean Oliver/David Hess chart and formulas are used, like I've posted over at dukehoopblog, that "stop percentage" refers to the percentage of plays on which the player in question is involved that we get a stop on. In this system, stop% is calculated as stops/Dposs, meaning stops divided by defensive possessions that he's involved in.

So how are stops and Dposs calculated? The "stop" number is complex, because it includes a defensive rebounding component, and involves a sliding weight between forced field goal misses and defensive rebounds. This is because forcing a missed shot does not equate to a stop unless we get the rebound. There are percentage weights attached to these values that originated with Dean Oliver.

DPoss is defined as the total defensive possessions that were credited to (or blamed on) a player. It's esentially stops + scoring possessions. Scoring possessions are field goals allowed plus a FT-related factor.

The stop % numbers I've been posting on this site are much simpler -- they refer to the percentage of possessions on which a given player is on the floor on which we get stops vs. those on which we give up hoops, regardless of the player's direct, quantifiable involvment in or impact on the play. The genesis of this was my admittedly crude attempt earlier in the season to try to measure the Tyler Thornton effect that many on these boards believed was operating -- that even though he didn't pass the eyeball test in many respects, was getting beaten often, etc., that due to his "intangibles" we were a better defensive team with him on the floor than without him.

Does that help?

Kedsy
02-16-2012, 06:16 PM
Here's the difference: On the charting in which the Dean Oliver/David Hess chart and formulas are used, like I've posted over at dukehoopblog, that "stop percentage" refers to the percentage of plays on which the player in question is involved that we get a stop on. In this system, stop% is calculated as stops/Dposs, meaning stops divided by defensive possessions that he's involved in.

So how are stops and Dposs calculated? The "stop" number is complex, because it includes a defensive rebounding component, and involves a sliding weight between forced field goal misses and defensive rebounds. This is because forcing a missed shot does not equate to a stop unless we get the rebound. There are percentage weights attached to these values that originated with Dean Oliver.

DPoss is defined as the total defensive possessions that were credited to (or blamed on) a player. It's esentially stops + scoring possessions. Scoring possessions are field goals allowed plus a FT-related factor.

The stop % numbers I've been posting on this site are much simpler -- they refer to the percentage of possessions on which a given player is on the floor on which we get stops vs. those on which we give up hoops, regardless of the player's direct, quantifiable involvment in or impact on the play. The genesis of this was my admittedly crude attempt earlier in the season to try to measure the Tyler Thornton effect that many on these boards believed was operating -- that even though he didn't pass the eyeball test in many respects, was getting beaten often, etc., that due to his "intangibles" we were a better defensive team with him on the floor than without him.

Does that help?

Yes, it helps. Thanks. Let me ask this, though: if the defensive player "causes" a missed shot, and the rebound goes long and the other team gets the offensive board, that doesn't count as a stop? Even though the failure had nothing to do with the defensive player? If I understand it properly, that would seem to be a flaw in the model.

tommy
02-16-2012, 06:27 PM
Yes, it helps. Thanks. Let me ask this, though: if the defensive player "causes" a missed shot, and the rebound goes long and the other team gets the offensive board, that doesn't count as a stop? Even though the failure had nothing to do with the defensive player? If I understand it properly, that would seem to be a flaw in the model.

No, it would not translate into a full stop. Yes, it would count as a forced miss, and forced misses are a component of stop% (and of overall DRating) but a forced miss does not equal a stop. Precisely because a forced miss needs to be coupled with a defensive rebound for a team, in reality, to get a stop. So in the situation you raise, the player will get credit for the forced miss, and his stop% will improve as a consequence of it, but it wouldn't on its own constitute a full stop. That's where the "sliding weight" involving defensive rebounding comes in, in calculating stop% -- how much the forced miss counts, and how much the defensive rebound counts, in attributing "credit" for a stop.

Sorry. I know, not very clear.

Kedsy
02-16-2012, 06:48 PM
No, it would not translate into a full stop. Yes, it would count as a forced miss, and forced misses are a component of stop% (and of overall DRating) but a forced miss does not equal a stop. Precisely because a forced miss needs to be coupled with a defensive rebound for a team, in reality, to get a stop. So in the situation you raise, the player will get credit for the forced miss, and his stop% will improve as a consequence of it, but it wouldn't on its own constitute a full stop. That's where the "sliding weight" involving defensive rebounding comes in, in calculating stop% -- how much the forced miss counts, and how much the defensive rebound counts, in attributing "credit" for a stop.

Sorry. I know, not very clear.

It's as clear as you can make it. But it seems like a slippery slope. Coach tells you to play up in a shooter's face, make him drive toward the big man in the middle. You execute it perfectly but the guy makes a circus shot against the big man, and it counts against you. You force him to take a tough mid-range shot, it clanks off the rim and his teammate grabs it and it counts (at least partially) against you. You let him take a wide open look and he misses it, someone else chases down the rebound, and it counts in your favor. Defense really is difficult to quantify.

tommy
02-16-2012, 07:15 PM
It's as clear as you can make it. But it seems like a slippery slope. Coach tells you to play up in a shooter's face, make him drive toward the big man in the middle. You execute it perfectly but the guy makes a circus shot against the big man, and it counts against you. You force him to take a tough mid-range shot, it clanks off the rim and his teammate grabs it and it counts (at least partially) against you. You let him take a wide open look and he misses it, someone else chases down the rebound, and it counts in your favor. Defense really is difficult to quantify.

First of all, you're right. It's very difficult to quantify in many respects, in meaningful ways. But at least this type of effort comes closer (or seems to) than just looking at steals and blocked shots, which for years and years was all that was done as to individuals. The other major point on all of this -- and anyone involved in it has to just admit it straight up, as it's just part of the deal -- is that there is a not insignificant degree of subjectivity in the charting. I might evaluate a given play differently than you in terms of who did what, who caused what action, who should've been where, etc. We also don't know what the coach was telling them to do. But like I said, you watch enough ball, play enough, know enough over the years, and take the time while charting to really study the play, you can get a pretty good idea on a given play of what did happen, what didn't, who did or didn't do what they should've, and go from there.

Not that you really asked me to, but to address the situations you raised:


Coach tells you to play up in a shooter's face, make him drive toward the big man in the middle. You execute it perfectly but the guy makes a circus shot against the big man, and it counts against you.

Not necessarily. On something like that, I'd have to make a judgment like this: Did the original defender "lose" his man in any real sense of the word, indicating poor play of some sort? Or did he pretty much stay with him and continue to be a factor on the drive? Did the big man fail to step up and help appropriately, or fail to even put his hands up, making the shot easier? Did both guys do their job but it was just a good offensive play? Depending on how I evaluate those sorts of questions, I might attribute the "made FG against" to the original defender, or to the big, or, as I often do, give each of them a half a made FG against.



You force him to take a tough mid-range shot, it clanks off the rim and his teammate grabs it and it counts (at least partially) against you.

No it wouldn't. You'd still get the credit for the forced field goal miss, which will help your stop% and your overall DRating.



You let him take a wide open look and he misses it, someone else chases down the rebound, and it counts in your favor.

Again, if you really forced the miss, or at least were a factor of some sort, then like the above example, you'd get the credit for the forced field goal miss, which would help your stop% and overall DRating. On this example though, if it truly was an uncontested shot, where you were not close enough to even get a hand up or bother him or be a factor at all, I wouldn't give credit for a forced miss there. I'd credit hte forced miss to "team."