Charting Duke v. Michigan NCAAT

[pfrduke]

Some interesting lineup combinations in this one. Count me as one vote against the Singler-at-center lineup. I thought we went away from Mason and Miles too much - the zone for Michigan was a lot easier to implement without any Duke size on the frontline. Not to take anything away from Ryan Kelly - I thought he played a very good game - but Mason and Miles have been very effective of late, and I think Mason's passing against the zone would have been a big help (his early turnovers in this game notwithstanding). And once we went small, we had a very difficult time getting stops - Michigan scored on 8 of their last 10 possessions. Over that same stretch, we had 9 possessions, scored 7 points (off of 2 offensive rebounds (one tip, one resulting in the Kyrie jumper) and 3 free throws), turned it over 3 times, and missed 3 3's. Long story short - I hope we don't have too many more games this year where Mason and Miles are both out for the last 9 1/2 minutes.

Individuals
Kyrie Irving 40-35 (+5)
Mason Plumlee 29-24 (+5)
Seth Curry 41-38 (+3)
Kyle Singler 63-61 (+2)
Miles Plumlee 26-24 (+2)
Nolan Smith 69-69 (0)
Ryan Kelly 51-52 (-1)
Andre Dawkins 46-52 (-6)

Per 40 Minutes
Mason Plumlee +11.1
Kyrie Irving +9.5
Miles Plumlee +5.3
Seth Curry +5.2
Kyle Singler +2.2
Nolan Smith +0.0
Ryan Kelly -1.7
Andre Dawkins -9.2

Lineups (Score, times used, margin)
Irving-Smith-Curry-Singler-Miles (7-0, x1, +7)
Irving-Smith-Singler-Mason-Kelly (11-7, x2, +4)
Smith-Curry-Dawkins-Singler-Kelly (9-6, x1, +3)
Irving-Smith-Curry-Dawkins-Kelly (10-8, x1, +2)
Irving-Dawkins-Singler-Mason-Kelly (4-2, x1, +2)
Smith-Curry-Dawkins-Singler-Miles (6-5, x2, +1)
Smith-Curry-Singler-Kelly-Miles (3-2, x1, +1)
Smith-Curry-Singler-Mason-Miles (6-6, x2, 0)
Irving-Smith-Dawkins-Singler-Mason (3-3, x1, 0)
Irving-Smith-Curry-Singler-Mason (0-0, x1, 0)
Smith-Dawkins-Singler-Mason-Kelly (5-6, x2, -1)
Irving-Smith-Dawkins-Singler-Kelly (5-6, x3, -1)
Smith-Curry-Dawkins-Kelly-Miles (0-2, x1, -2)
Irving-Smith-Curry-Singler-Kelly (0-4, x2, -4)
Smith-Dawkins-Singler-Kelly-Miles (4-9, x1, -5)
Irving-Smith-Curry-Dawkins-Singler (0-5, x1, -5)

[loldevilz]

This shows what I suspected. Mason and Irving were two of the best player today. Neither in my opinion got enough playing time.

[uh_no]

This shows what I suspected. Mason and Irving were two of the best player today. Neither in my opinion got enough playing time.

what exactly is your justification for kyrie? he was a beast from the FT line, but was an underwhelming 1-4 with 2 ast and 2 TO. He also played 21 minutes, which is plenty for someone who got cleared less than a week ago....

[loldevilz]

what exactly is your justification for kyrie? he was a beast from the FT line, but was an underwhelming 1-4 with 2 ast and 2 TO. He also played 21 minutes, which is plenty for someone who got cleared less than a week ago....

Did you not read the chart:

Per 40 Minutes
Mason Plumlee +11.1
Kyrie Irving +9.5

[MCFinARL]

what exactly is your justification for kyrie? he was a beast from the FT line, but was an underwhelming 1-4 with 2 ast and 2 TO. He also played 21 minutes, which is plenty for someone who got cleared less than a week ago....

Have to agree with you here--as well as with pfrduke's general point about the effectiveness of the small vs. big lineups today generally.

These stats raise a question for me, though, about what the plus/minus numbers do and don't show. I'm coming to this with absolutely no expertise, so I'm genuinely hoping someone can help me out here. My eyes told me that Andre Dawkins played more than usual today and had, overall, a very solid outing. And yet his plus minus numbers were the worst on the team for today's game. Obviously, it's a team game and some of how this works out is going to be a little random, especially on a single game basis. But how much? What conclusions, if any, can we fairly draw about Dawkins' game today (or anybody esle's) from these numbers, and what can't we?

[Saratoga2]

Have to agree with you here--as well as with pfrduke's general point about the effectiveness of the small vs. big lineups today generally.

These stats raise a question for me, though, about what the plus/minus numbers do and don't show. I'm coming to this with absolutely no expertise, so I'm genuinely hoping someone can help me out here. My eyes told me that Andre Dawkins played more than usual today and had, overall, a very solid outing. And yet his plus minus numbers were the worst on the team for today's game. Obviously, it's a team game and some of how this works out is going to be a little random, especially on a single game basis. But how much? What conclusions, if any, can we fairly draw about Dawkins' game today (or anybody esle's) from these numbers, and what can't we?

The eye test would indicate that Kyrie, Nolan, Kyle, Ryan and Andre all had good to very good games, whereas Mason and Seth did not. Miles was sort of in between. ince that seems to be the opposite of what the ) +/- points are saying, it is hard for me to justify drawing conclusions from the numbers. Seth had zero points, 5 fouls and had trouble guarding his man. How can that result in a positive?

[Bob Green]

What conclusions, if any, can we fairly draw about Dawkins' game today (or anybody esle's) from these numbers, and what can't we?

These +/- numbers are entertaining, and over the course of the season, can reveal things about the effectiveness of different line-up combinations, but looking at the numbers for a single game isn't productive. The bigger the sample size the better.

[DukeUsul]

They also won't tell you who should be on the floor during the stretch run. Someone who plays well against M2M for the first 30 minutes isn't necessarily the person who needs to be in when the opponent makes a drastic change to a different defense that needs to be handled much differently.

[Bluedog]

These +/- numbers are entertaining, and over the course of the season, can reveal things about the effectiveness of different line-up combinations, but looking at the numbers for a single game isn't productive. The bigger the sample size the better.

Agreed. Please see ken pom's analysis.

http://kenpom.com/blog/index.php/weblog/a_treatise_on_plus_minus/


In-game plus minus is useless.


In summary, plus-minus, while neat to look at, is a poor tool in college basketball analysis.

Personally, I thought on the offensive end we were more effective when the Plumlees were in the game. However, from a defensive standpoint, it might have made more sense to go with Kelly and Singler at the four - which is what Coach K did - but at perhaps some expense of our offensive efficiency if we weren't hitting outside shots (which we weren't).

[cspan37421]

We question the strategy of the intentional miss by Zoubek in last year's final ... question our lineup near the end vs. Michigan, but Coach K keeps on winning.

[ns7]

These +/- numbers are entertaining, and over the course of the season, can reveal things about the effectiveness of different line-up combinations, but looking at the numbers for a single game isn't productive. The bigger the sample size the better.

Even season long numbers in college basketball can be highly misleading if you use them out of context. Here is a kenpom.com article:
http://kenpom.com/blog/index.php/weblog/a_treatise_on_plus_minus/

However I really appreciate the effort pfr puts into these and they are valuable when combined with other stats/observations.

I do agree with the earlier points about Mason. I thought he would have helped in two ways against the zone:
1) passing out of the post when the zone collapsed on him
2) diagonal alley-oop style passes that he could convert easily to beat the zone

Not sure how Mason or Miles would have helped at the end on defense though, Michigan was deadly on converting jump shots so the Plumlee defensive rebounding would not have mattered here.

[MCFinARL]

Thanks to all who responded to my question--very helpful.

[COYS]

These +/- numbers are entertaining, and over the course of the season, can reveal things about the effectiveness of different line-up combinations, but looking at the numbers for a single game isn't productive. The bigger the sample size the better.

To go along with this point, Andre did have a pretty solid outing, but he was on the floor during two Michigan runs including a smaller one in the first half and the big one over the last five minutes that gave us all high blood-pressure. It is certainly unfair to blame him for the entirety of that run, but that sequence had a profound effect on his in game +/- stats.

[turnandburn55]

IIRC, Dawkins was on the floor for most of the last 10 minutes when Michigan's big run occured. I agree with the poster in the in-game thread who pointed out that he seemed to take too many forced threes (from the same position, no less) as the game wound down...

That being said, his game overall was solid. The long three earlier in the game brought a smile to my face...

[Kedsy]

Even season long numbers in college basketball can be highly misleading if you use them out of context. Here is a kenpom.com article:
http://kenpom.com/blog/index.php/weblog/a_treatise_on_plus_minus/

Here's the thing about the KenPom article regarding plus/minus: in his test, the 20 games should have all approached a zero scoring margin (i.e., a tie), but in 8 of the 20 games, the margin was 13 or more; in 5 of the 20 games the margin was 19 or more. Pomeroy suggests that these numbers mean plus/minus is a flawed stat, but if it shows that, doesn't his study also show that any scoring-based system is flawed? And isn't Pomeroy's whole rating system based on scoring?

Everyone points to this article as "proving" plus/minus is worthless, but I think the problem is either with the article or Pomeroy's test.

Having said that, I agree that one game plus/minus has too many uncontrolled and random variables to be particularly helpful.


I agree with the poster in the in-game thread who pointed out that he seemed to take too many forced threes (from the same position, no less) as the game wound down...

After Kyle drew his fourth foul at the 9:50 mark of the second half, Andre took exactly two three-pointers (one at the 4:53 mark and one at the 1:02 mark). Can two be described as "too many" when taken by a 43% three-point shooter? I suppose that could be debated, but not for very long.

[COYS]

After Kyle drew his fourth foul at the 9:50 mark of the second half, Andre took exactly two three-pointers (one at the 4:53 mark and one at the 1:02 mark). Can two be described as "too many" when taken by a 43% three-point shooter? I suppose that could be debated, but not for very long.

I agree with Kedsy, here. Andre's +/- suffered because he was on the floor during a terrible stretch of play by Duke, but I think little of the blame falls on Andre. Whether it was our game plan, communication errors, execution errors, lack of focus, or even just excellent play by the Wolverines (a well coached team desperate to extend their season), there were many more contributing factors than just Andre at work.

[turnandburn55]

After Kyle drew his fourth foul at the 9:50 mark of the second half, Andre took exactly two three-pointers (one at the 4:53 mark and one at the 1:02 mark). Can two be described as "too many" when taken by a 43% three-point shooter? I suppose that could be debated, but not for very long.

Sure, if you discount the one at the 9:54 mark.

And nobody's blaming Andre for the Michigan run... there's a number of factors in there, and as I've said repeatedly, the decision to stall-ball was the biggest. But if you look at what was effective at the time, it was us driving aggressively to the basket.... and those threes weren't coming off drive-and-kicks.

[Kedsy]

Sure, if you discount the one at the 9:54 mark.

Sorry, but you said "as the game wound down," so I'm not sure about the relevance of a shot attempt at the 9:54 mark.

To me it seemed obvious that the turning point was Kyle's fourth foul, not stall ball and certainly not Andre's shot selection.

[ns7]

Here's the thing about the KenPom article regarding plus/minus: in his test, the 20 games should have all approached a zero scoring margin (i.e., a tie), but in 8 of the 20 games, the margin was 13 or more; in 5 of the 20 games the margin was 19 or more. Pomeroy suggests that these numbers mean plus/minus is a flawed stat, but if it shows that, doesn't his study also show that any scoring-based system is flawed? And isn't Pomeroy's whole rating system based on scoring?


RMS error is ~11-12 for basketball according to most empirical studies, meaning that 32% of games between equal teams will be outside of +/- 11 points. (Stats people please correct me if I'm wrong here). The numbers he shows seem to be close (40% outside).

You may have a point that his season simulator could have some errors--I'm uncertain as to whether his variance for scoring accurately depicts a common D-1 team:


For both his team and his opponent, there’s a 3% chance of scoring one point on a possession, a 30% chance of scoring two, and a 15% chance of scoring three.

[Kedsy]

RMS error is ~11-12 for basketball according to most empirical studies, meaning that 32% of games between equal teams will be outside of +/- 11 points. (Stats people please correct me if I'm wrong here). The numbers he shows seem to be close (40% outside).

That's interesting and a little surprising, at least to me. But if that's the case, why would the randomness in his test condemn plus/minus as a meaningful stat and not, e.g., offensive efficiency?