RealGM

I did not mean to hijack supersubs thread so I decided to put this in its own thread. As it can merit discussion

The original from 82games.com


http://img240.imageshack.us/img240/6507/5menunitspx5.jpg

Adjust per minute and sorted by (+/-) per minute.


http://img239.imageshack.us/img239/6492/5manunitwv3.jpg

-I notice that Delfino without Moon is usually in least productive lineups.

-Calderon is more productive as a starter than Ford ( as far as +/- is concerned)

-Andre is not the liability that posters emphasize.

-Graham is.


NOTE: The 1st most productive lineup with Martin+bench is our blow out close out line up... these guys usually play against the other teams bench.

That's good to see for the people who are hating on Bargnani.. even though he's in a slump he is still being some what productive..

Here is the URL if you are having trouble seeing the graphic. (it's all distorted on my screen)

http://www.82games.com/0708/0708TOR2.HTM

There is also one thing that I wanted to point out. 82games only lists 20 of the most used linups.

So this means that combinations involving Graham, Martin that did not make the top 20 are not displayed.


So for instance we only see 36/112 minutes for Martin. We also only see 26/165 minutes for Graham.

My conclusion is that the above numbers do not judge players that have received limited playing time fairly. Martin and Joey specifically. In the case of Martin he is artificially high, in the case of Joey I suspect he is too low.

These numbers are a lot more accurate for players with significant minutes.

dinosaur_dan wrote:Here is the URL if you are having trouble seeing the graphic. (it's all distorted on my screen)

http://www.82games.com/0708/0708TOR2.HTM

Ok try this link for the second chart. Adjust by per minute.

http://img239.imageshack.us/img239/6492/5manunitwv3.jpg


I get the same thing but when I use [CTRL]+[mouse wheel] I can enlarge the pictures. If anyone knows what I am doing wrong I can edit the post so that they display properly.


(Weird the pictures come in fine with Opera, but distort with IE)

from24ft wrote:Ok try this link for the second chart. Adjust by per minute.

http://img239.imageshack.us/img239/6492/5manunitwv3.jpg


I get the same thing but when I use CTRL-wheel on my mouse I can enlarge the pictures. If anyone knows what I am doing wrong I can edit the post so that they display properly.

Ohhh, that second chart has all the good stuff in it. Nice work!

I looked at Kapono's lines. Except for the garbage-time line (1st in the list), Kapono doesn't appear till the 9th best line. I wanted to see how his lines did defensively. I expected them to have higher scoring but worse defense. The average defensive score is 2.03 d/m. Kapono was on 4 lines higher than this, and 4 lines lower. So, he doesn't push the team too far one way or the other, statistically. Offensively, the average score was also 2.03 d/m. Kapono was on 3 lines higher than, but 5 lines below.

So, this tells us that Kapono doesn't have much impact at all. He never really hurts the team or helps the team in any big way. Sounds like the guy you bench whenever another guy (parker, delfino, moon) is having a big game.

At the same time we haven’t been using Kapono efficiently when he is on the floor. Too many times I see him log 10 minutes with no shot attempts.

I think plays have to be drawn up specifically to employ Kapo, otherwise his best talents are wasted.


It's like Smitch puts him out there and expects him to contribute without lunching any threes, very stupid.

You shouldn't trust +/- in small sample sizes. Individual +/- over the course of the season will tell you something about a player's worth .... but there are just too many 5 man combinations, so each one doesn't get enough playing time for the statistic to be much more than noise. It takes a lot of data to get a good result because so much depends on the competition on the floor, the energy level of the unit, and pure luck.

For example: Darrick Martin is a +/- beast this year. How can we explain that he is, by far, our team's leader in +/- per minute? He's just on a lucky run, and he hasn't played enough minutes (only 116) for the statistic to tell us anything about him. Eventually the fact that he is a liability will show up, but it takes time.

So if Darrick Martin's 116 minutes aren't enough to give us a reliable statistic, it's easy to see that the five man unit +/- is statistical noise. The one with the most minutes has only 198.

If you want to look at +/-, you should look at individual players with a lot of minutes ... but even then I would be very cautious in drawing any conclusions. Here's a link to the Rap's individual stats:

http://www.nba.com/statistics/lenovo/le ... am=Raptors

pierrot wrote:You shouldn't trust +/- in small sample sizes. Individual +/- over the course of the season will tell you something about a player's worth .... but there are just too many 5 man combinations, so each one doesn't get enough playing time for the statistic to be much more than noise. It takes a lot of data to get a good result because so much depends on the competition on the floor, the energy level of the unit, and pure luck.

For example: Darrick Martin is a +/- beast this year. How can we explain that he is, by far, our team's leader in +/- per minute? He's just on a lucky run, and he hasn't played enough minutes (only 116) for the statistic to tell us anything about him. Eventually the fact that he is a liability will show up, but it takes time.

So if Darrick Martin's 116 minutes aren't enough to give us a reliable statistic, it's easy to see that the five man unit +/- is statistical noise. The one with the most minutes has only 198.

If you want to look at +/-, you should look at individual players with a lot of minutes ... but even then I would be very cautious in drawing any conclusions. Here's a link to the Rap's individual stats:

http://www.nba.com/statistics/lenovo/le ... am=Raptors

I don't think 20 top line-ups are a small sample size. For instance our starting line-up is fairly static, our line-up when Ford is healthy is up there too. These numbers don't just show one player, they show the efficiency of the group as a whole.

I agree that picking just one linup without looking at the minutes may be fallacy, but looking at combinations and comparing them as a group is quite telling. For instance the Rosho/Bosh line-up vs Bosh/Andrea line-up shows what we know and has a huge sample size. There is a small boost in offence and a small drop off in defence.

I think these numbers do have some merit when analyzed by an analytical eye. The +/- is not individual, and it certainly is not compiled over one game.

from24ft wrote:I don't think 20 top line-ups are a small sample size.

Yes, they are. Look at the minutes .. only two of these line-ups have been on the court for more than 100 minutes total. Most have only played between 20 and 30 minutes of actual game time over the course of the whole year. In order to draw conclusions about future performance, you would need a lot more data. You're better off trusting what you see on the court than these numbers, because they are close to meaningless.

from24ft wrote:The +/- is not individual, and it certainly is not compiled over one game.

Individual +/- is better. Some of our players have been on the court for over 1000 minutes this year. 1000 minutes is a much better sample size to work with than, say, 20.

No one ever said that these +/- numbers were compiled over one game. I am aware of how the statistic works.

from24ft wrote:For instance the Rosho/Bosh line-up vs Bosh/Andrea line-up shows what we know and has a huge sample size.

200 minutes is much too small a sample size to tell us anything.

And if your criteria for evaluating statistical performance is "does it tell me what I already know," you might as well not bother.

from24ft wrote:I think these numbers do have some merit when analyzed by an analytical eye.

The Romans believed that an "analytical eye" could tell the future by examining sheep guts.

If you trust these numbers, you should be prepared to say that Martin-Delfino-Kapono-Humphries-Bargnani is one of our best units. Will you make that claim?