PDA

View Full Version : Charting Duke vs Maryland (I)



loran16
01-10-2011, 12:27 AM
Fascinating game for the +/- numbers. The top two should not surprise anyone. And of course, Nolan and Kyle played all 40 minutes, so their +/- was of course equal to the margin of victory.

Tyler Thornton's +/- by the way is lower than it should be....Coach K at the end of the game was switching out Thornton and Curry repeatedly...Curry would go in for offense, while Thornton was in for defense. With Thornton not in for O, he lost 3 points of +/-, though Seth didn't gain much.

This was a bad game for all of the bigs, so the Plum-haters should find vindication in these numbers. Mason of course fouled out, while Miles didn't contribute positively at all. Meanwhile, Both Kelly and Dawkins (okay not a big) were harmed by not being in the lineup during the 8-0 run, resulting in them having negative +/-s.

(Once again, in an odd trend, the big lineup for Duke was only used once, which included both TT and Curry. Quite odd).

+/-
Seth Curry +13
Tyler Thornton +8
Nolan Smith +7
Kyle Singler +7
Mason Plumlee +3
Miles Plumlee +0
Andre Dawkins -1
Ryan Kelly -2

(NO +/- per 40 right now, due to a lack of time, sorry guys.)

LINEUPS
Thornton-Curry-Smith-Singler-Miles +8 (14-6) (1x)
Thornton-Curry-Smith-Dawkins-Singler +4 (4-0) (2x)
Smith-Curry-Dawkins-Singler-Mason +3 (9-6) (1x)
Smith-Dawkins-Singler-Kelly-Mason +1 (10-9) (5x)
Smith-Curry-Singler-Miles-Mason +0 (5-5) (1x)
Thornton-Curry-Smith-Singler-Mason +0 (0-0) (1x)
Thornton-Smith-Dawkins-Singler-Mason -1 (3-4) (1x)
Smith-Curry-Dawkins-Singler-Miles -2 (4-6) (5x)
Thornton-Smith-Dawkins-Singler-Miles -3 (5-8) (6x)
Smith-Dawkins-Singler-Kelly-Miles -3 (17-20) (4x)

proelitedota
01-10-2011, 12:45 AM
Thornton-Curry-Smith-Dawkins-Singler +4 (4-0) (2x)

Starting line-up please. :cool:

jv001
01-10-2011, 07:56 AM
Thornton-Curry-Smith-Dawkins-Singler +4 (4-0) (2x)

Starting line-up please. :cool:

This will not happen anytime soon. 4 guards and SF. Where will the rebounds and post defense come from? However Coach K now has an idea of what Thornton can do in a tight game. Good to see Curry play probably his best game ytd. Go Duke!

nocilla
01-10-2011, 10:36 AM
Thornton-Curry-Smith-Dawkins-Singler +4 (4-0) (2x)

Starting line-up please. :cool:

This line-up was used twice in the final minute when Maryland was expected to foul right away. So it was merely Coach K putting in ball handlers and free throw shooters. Right after the free throws were hit, Miles came back in the game. This is not a line-up we will see at any other point in the game, as loran pointed out, we would be destroyed on the defensive boards.

Indoor66
01-10-2011, 11:00 AM
This line-up was used twice in the final minute when Maryland was expected to foul right away. So it was merely Coach K putting in ball handlers and free throw shooters. Right after the free throws were hit, Miles came back in the game. This is not a line-up we will see at any other point in the game, as loran pointed out, we would be destroyed on the defensive boards.

It is also an example of why charting has very limited usefulness on a one game basis. IMO it is only meaningful with many games of charting - like maybe 20+ games. Short of that it is very misleading.

nocilla
01-10-2011, 11:09 AM
This line-up was used twice in the final minute when Maryland was expected to foul right away. So it was merely Coach K putting in ball handlers and free throw shooters. Right after the free throws were hit, Miles came back in the game. This is not a line-up we will see at any other point in the game, as loran pointed out, we would be destroyed on the defensive boards.

Oops, that should be as jv001 pointed out.

loran16
01-10-2011, 11:19 AM
This line-up was used twice in the final minute when Maryland was expected to foul right away. So it was merely Coach K putting in ball handlers and free throw shooters. Right after the free throws were hit, Miles came back in the game. This is not a line-up we will see at any other point in the game, as loran pointed out, we would be destroyed on the defensive boards.

Errr to be fair, that lineup was used once and got fouled (made 1 free throw), but before that was used in the final minute and resulted in Singler's Dagger 3.

But yes, that lineup will never see crunch time at all.

tbyers11
01-11-2011, 05:18 PM
This would probably be better served in the cumulative +/- thread tacked at the top of the board. Since pfrduke is still on his honeymoon and it is not being updated and I doubt that many will look at it, I put it in this most recent +/- thread.

In his blog today, Pomeroy looks at plus/minus (http://kenpom.com/blog/index.php/weblog/a_treatise_on_plus_minus/) from the college basketball perspective and basically declares it of little value. Not just from a one game sample as many of us (including Indoor66 three posts up) have noted but from a multiple (20 game) sample. He crunches some numbers and finds that the variance is so large that it is quite misleading. He states that adjusted plus/minus used in the NBA might be relevant on a season long basis but that the adjustments that the NBA uses don't apply as well to the college game.

Very interestingly, he links this article from The Audacity of Hoops (http://audacityofhoops.blogspot.com/2010/12/how-much-will-kyrie-irvings-absence.html) that discusses the extremely strange Kyrie Irving +/- effect that we all noticed in several early season games. Irving, while playing very well to someone watching the game, had poor +/- stats in the Kansas State, Michigan State and Butler games.

Basically, I think that plus/minus is interesting to look at and I am grateful to Jumbo, pfrduke and now loran16 for doing it. However, I definitely agree with Pomeroy that we shouldn't read too much into it.

loran16
01-12-2011, 12:12 AM
This would probably be better served in the cumulative +/- thread tacked at the top of the board. Since pfrduke is still on his honeymoon and it is not being updated and I doubt that many will look at it, I put it in this most recent +/- thread.

In his blog today, Pomeroy looks at plus/minus (http://kenpom.com/blog/index.php/weblog/a_treatise_on_plus_minus/) from the college basketball perspective and basically declares it of little value. Not just from a one game sample as many of us (including Indoor66 three posts up) have noted but from a multiple (20 game) sample. He crunches some numbers and finds that the variance is so large that it is quite misleading. He states that adjusted plus/minus used in the NBA might be relevant on a season long basis but that the adjustments that the NBA uses don't apply as well to the college game.

Very interestingly, he links this article from The Audacity of Hoops (http://audacityofhoops.blogspot.com/2010/12/how-much-will-kyrie-irvings-absence.html) that discusses the extremely strange Kyrie Irving +/- effect that we all noticed in several early season games. Irving, while playing very well to someone watching the game, had poor +/- stats in the Kansas State, Michigan State and Butler games.

Basically, I think that plus/minus is interesting to look at and I am grateful to Jumbo, pfrduke and now loran16 for doing it. However, I definitely agree with Pomeroy that we shouldn't read too much into it.

I agree, it doesn't really seem to have much predictive value at times. However, when looking at the lines, you can truly see what lines were effective in a specific game which can allow you to validate previous judgments.

Basically, it's a fun statistic. A good bit flawed, for sure, and not something to be taken as the be-all end-all. But it's fun to see. For example, it does capture in this game how bad all the big men were, and how good Seth and Tyler were.

ice-9
01-12-2011, 03:14 AM
This would probably be better served in the cumulative +/- thread tacked at the top of the board. Since pfrduke is still on his honeymoon and it is not being updated and I doubt that many will look at it, I put it in this most recent +/- thread.

In his blog today, Pomeroy looks at plus/minus (http://kenpom.com/blog/index.php/weblog/a_treatise_on_plus_minus/) from the college basketball perspective and basically declares it of little value. Not just from a one game sample as many of us (including Indoor66 three posts up) have noted but from a multiple (20 game) sample. He crunches some numbers and finds that the variance is so large that it is quite misleading. He states that adjusted plus/minus used in the NBA might be relevant on a season long basis but that the adjustments that the NBA uses don't apply as well to the college game.

I would love for someone to explain to me what Pomeroy did in his example, because I was rather confused by his methodology.

tele
01-12-2011, 05:54 AM
It is also an example of why charting has very limited usefulness on a one game basis. IMO it is only meaningful with many games of charting - like maybe 20+ games. Short of that it is very misleading.

It's also of limited usefulness when you have two players playing 40 minutes, or even 3 players playing almost the whole game. like last year. Then for all other players you are limited to seeing how well they play with these predominate players. And comparisons of individual players is also limited since the players playing the whole game will always have the same plus minus, the final score margin. So they may have a low plus/minus relative to another player, even when they are clearly your best players, over the course of a single game or a season.

I do think that plus/minus has some interesting sidelights vs the spread, but probably just in rare cases.

CDu
01-12-2011, 09:29 AM
I would love for someone to explain to me what Pomeroy did in his example, because I was rather confused by his methodology.

It looks like he did the following:
1. Created a 70-possession game and had the random player in for half of those possessions. In other words, tagged each possession with a flag that indicated whether or not the player was in the game for that possession or not, with an equal number of times when the player was in and when he was out.
2. Assigned the following probabilities for each possible outcome in a possession: 52% for 0, 3% for 1, 30% for 2, and 15% for 3. The probabilities were the same for both the offense and defense and the same regardless of whether the player was in or not. This means the player had no impact on the probability of the team's outcome on either end of the court.
3. Simulated 20 games by generating an outcome for each possession based on the probabilities listed.
4. Estimated the +/- for the team with the player "in the game" and with the player "out of the game." In other words, estimated a +/- for the possessions which were flagged to indicate the player was in, and a +/- for the possessions which were flagged to indicate he was out.

The expected +/- should have been zero for both the "in" and "out" scenarios, since the probabilities of a basket were the same for both offense and defense with and without the player. Obviously, there will be some variation, but the idea is that over a reasonably sized sample, the outcomes will converge to the expected average (in this case, zero).

The fact that he sees so much variation leads him to the conclusion that +/- is just subject to too much variability to be a reliable tool over even a 20-game sample. As Pomeroy notes, there's a lot of randomness that occurs that people tend to overlook, and +/- doesn't really do any job of filtering out that randomness.

Jderf
01-12-2011, 10:08 AM
I agree, it doesn't really seem to have much predictive value at times. However, when looking at the lines, you can truly see what lines were effective in a specific game which can allow you to validate previous judgments.

Basically, it's a fun statistic. A good bit flawed, for sure, and not something to be taken as the be-all end-all. But it's fun to see. For example, it does capture in this game how bad all the big men were, and how good Seth and Tyler were.

I have never been a fan of the +/- numbers, especially as a metric for the influence of individual players; yes, it gives credit where credit is due, but it also gives credit where it is not due. While other stats like Points Per Game may not tell the whole story, at least you know that each point gained is a point earned. With +/-, it is never that simple, even in the long run (as kenpom points out).

Nonetheless, I think loran makes a good point here. While +/- stats might be useless in terms of evaluating a single player, they are definitely helpful in evaluating a lineup as a whole. This makes sense, since it is really just the scoring margin for specific lineups. Perhaps that should be the focus of any +/- discussion, rather than individual performance. That way, we could avoid confusing discussions like the one regarding Miles' stellar +/- stats from earlier this season.

CDu
01-12-2011, 10:22 AM
Nonetheless, I think loran makes a good point here. While +/- stats might be useless in terms of evaluating a single player, they are definitely helpful in evaluating a lineup as a whole. This makes sense, since it is really just the scoring margin for specific lineups. Perhaps that should be the focus of any +/- discussion, rather than individual performance. That way, we could avoid confusing discussions like the one regarding Miles' stellar +/- stats from earlier this season.

Actually, I think even this isn't really the case. I'd caution in using the +/- to evaluate a lineup as a whole. The +/- simply illustrates the results achieved by a particular player or lineup. But as Pomeroy just illustrated, there is so much random variation inherent in basketball possession-by-possession that individual +/- are largely meaningless unless you have a REALLY large sample. Well, lineup +/- have even smaller samples and are thus subject to similar problems. You eliminate some of the variables present in the individual +/-, but also greatly reduce the sample of information.

The +/- can give you the exact outcomes observed for a particular lineup or individual. But just as the +/- is a poor evaluative tool at the individual level, it's probably as poor a tool at the team level. It's a data point, but you'd need to consider a lot of contextual information that +/- can't provide. And even then you'd need to account for the problem of randomness and outlier occurrences.

Basically, I think +/- is a nice concept and fun for discussion. But as an evaluative tool (especially at a single-game level, but even at a season level) it is just too flawed, just like the other more standard metrics. This is true both at the individual and lineup level.

NSDukeFan
01-12-2011, 02:33 PM
It's also of limited usefulness when you have two players playing 40 minutes, or even 3 players playing almost the whole game. like last year. Then for all other players you are limited to seeing how well they play with these predominate players. And comparisons of individual players is also limited since the players playing the whole game will always have the same plus minus, the final score margin. So they may have a low plus/minus relative to another player, even when they are clearly your best players, over the course of a single game or a season.

I do think that plus/minus has some interesting sidelights vs the spread, but probably just in rare cases.

I would think that +/- would actually be a little bit more valuable if two players are playing 40 minutes as that would take away two confounding variables when measuring player's impacts.

I can understand Pomeroy's points that there is a lot of variation and randomness in basketball, but I still believe +/- has value. There just isn't a better way (at this point, that I know of) to measure some of the "Battier" things that players do such as communicating and helping on defense, screening and moving the ball on offense. Victories and +/- are the best ways that I know of to measure some of this stuff, realizing that randomness is a confounding factor and the eyeball test has some value, assuming a non-ESPN-highlight-searching set of eyeballs.

loran16
01-12-2011, 03:07 PM
+/- is fun. As an evaluating tool and predictive tool, Ken is correct, it's value is very little.

Essentially, what Ken is talking about should be familiar to anyone who keeps track of hockey, where +/- has been a statistic for AGES. Anyone who pays any attention to the sport knows that it's a misleading statistic....a great player who is traded from a bad team to a good team suddenly has a huge change in +/-....not because he's better, but because of his teammates.

What some NBA guys use is ADJUSTED +/-, where you adjust +/- for who is ON and who is OFF the Court when a player is on the floor. The problem with this in college is several fold:
1. Less Games to accumulate Data, each of which is 8 minutes shorter
2. College players will play all 40 minutes and have much smaller benches.

This 2nd point is key. To use +/- to show a player is better than a second player, you have to compare two lineups which consist of the same FOUR OTHER PLAYERS and the two players you're comparing (This is called WOWY or With or Without You). With more lines, this becomes easier to do, as you can eliminate noise by repeating WOWY evaluations with different four-other-players.

For Duke however, you're gonna get some of the same guys in those four players all of the time. So you'll get Nolan-Singler-Scheyer last year ALWAYS among those 4 guys. And last year in particular, you rarely had straight subs of say...Lance for Miles, but instead subs of two different players at once (Zou and Lance would go out together, Miles and Mason would come in). Thus it was basically impossible to use +/- to show that LANCE THOMAS was a better player than Miles or Mason in the starting lineup. You could show that Zou and Lance were a better duo than the Plumlees, perhaps, but even then there's a lot of noise because you always have the Scheyer-Smith-Singler trio on the floor.

Thus instead of +/-, we need to use our eyes and other stats. It's still very fun, but it's not incredibly USEFUL.

NSDukeFan
01-12-2011, 03:30 PM
+/- is fun. As an evaluating tool and predictive tool, Ken is correct, it's value is very little.

Essentially, what Ken is talking about should be familiar to anyone who keeps track of hockey, where +/- has been a statistic for AGES. Anyone who pays any attention to the sport knows that it's a misleading statistic....a great player who is traded from a bad team to a good team suddenly has a huge change in +/-....not because he's better, but because of his teammates.

What some NBA guys use is ADJUSTED +/-, where you adjust +/- for who is ON and who is OFF the Court when a player is on the floor. The problem with this in college is several fold:
1. Less Games to accumulate Data, each of which is 8 minutes shorter
2. College players will play all 40 minutes and have much smaller benches.

This 2nd point is key. To use +/- to show a player is better than a second player, you have to compare two lineups which consist of the same FOUR OTHER PLAYERS and the two players you're comparing (This is called WOWY or With or Without You). With more lines, this becomes easier to do, as you can eliminate noise by repeating WOWY evaluations with different four-other-players.

For Duke however, you're gonna get some of the same guys in those four players all of the time. So you'll get Nolan-Singler-Scheyer last year ALWAYS among those 4 guys. And last year in particular, you rarely had straight subs of say...Lance for Miles, but instead subs of two different players at once (Zou and Lance would go out together, Miles and Mason would come in). Thus it was basically impossible to use +/- to show that LANCE THOMAS was a better player than Miles or Mason in the starting lineup. You could show that Zou and Lance were a better duo than the Plumlees, perhaps, but even then there's a lot of noise because you always have the Scheyer-Smith-Singler trio on the floor.

Thus instead of +/-, we need to use our eyes and other stats. It's still very fun, but it's not incredibly USEFUL.

I don't disagree with most of your post, but for statistical purposes wouldn't it be easier to get more meaning from +/- if you always had the Scheyer-Smith-Singler trio or this year always having Smith-Singler? At least for other players, it would be easier to compare since you would have fewer variables as opposed to adding noise.

loran16
01-12-2011, 04:27 PM
I don't disagree with most of your post, but for statistical purposes wouldn't it be easier to get more meaning from +/- if you always had the Scheyer-Smith-Singler trio or this year always having Smith-Singler? At least for other players, it would be easier to compare since you would have fewer variables as opposed to adding noise.

Yes and no. Yes, you want to have enough data for two lineups that feature the same grouping of players other than the two you're comparing. Thus the trio last year does help.

But you'd like to compare more than just two lineups, because with one such set of lineups, the constant players could be the ones dominating play.

In other words, you'd like to have:
Scheyer, Smith, Singler, Lance, Zoubek
Scheyer, Smith, Singler, Lance, Mason

BUT ALSO:
Davidson, Dawkins, Singler, Lance Zoubek
Davidson, Dawkins, Singler, Lance, Mason

and

Smith, Dawkins, Lance, Miles, Zoubek
Smith, Dawkins, Lance, Miles, Mason
etc. etc.

You get my point. By having more lineups to compare, you avoid the trap of simply comparing lineups led by dominant constants. When you do that the result is once agian often just noise resulting from just random variation (Mason isn't doing anything better than Zou, but he's just on the floor when Scheyer has two 11-11 shooting nights, causing him to look better).