I can't help but think that the extent of your analysis there is that Asik is another gawky looking rookie straight out of the euro leagues and therefore can't possibly be a good defender. "Correlation doesn't equal causation" ... ahem. David Thorpe midseason rookie report: http://insider.espn.go.com/nba/insider/columns/story?columnist=thorpe_david&page=Rookies-110126 [rquoter] Best Defender: Omer Asik, Bulls Part of the reason why Asik gets the nod is the lack of good defense being played overall by this class. But he's deserving as well, thanks to his attentiveness to scouting reports and his willingness to move his feet and contest every shot. He does jump a bit too much for the shot block, costing him valuable rebounding position, but for the most part he's an excellent pick-and-roll defender and a tough guy to score on inside.[/rquoter]
He can move his feet and read scouting reports? Impressive. The Bulls are crazy for getting 38 yr old Kurt Thomas and starting him over Asik the elite defender. I also wonder why Thomas was unemployed until very recently since, according to the #s you threw together, he's apparently an elite defender too, lol. Okafor and Marc Gasol can only wish to be what Asik is defensively. In sum, correlation does not equal causation. This is an elementary rule of statistics, despite what you may think.
"Not equal" does not equal to "not related." If two things are highly correlated, you try to find out the causation, either between the two things or some other factors. The alternative is believing in pure luck, which is much less rational. You know, people noticed the correlation between smoking and lung cancer way before they found out the cause. This makes no sense. You don't think leadership can be measured by how early someone shows up to practice, then how do you know leadership? By results? What results? Can you measure it? If you can't measure it, how can you tell it is there? You label people as "dogmatic" about statistics, but in fact it is people like you who are dogmatic against statistical analysis because you do not understand how it works and how it is used. You are really beating the strawman.
Where did I say that Asik is a better defender than Kurt Thomas because of the results of my formula? I did not, could not, and would not conclude such a thing based simply on a statistical model. Especially one that is fairly simplistic and imperfect, like the one presented in the first post. At best what I could say that is the model indicates or suggests that Asik has done a good job on the defensive end. That is not confusing correlation with causation, despite what you might think. "Model indicates X" does not equal "X is true". For most people, that should go without saying. For others like yourself, I guess it needs to be spelled out explicitly.
The problem with +/- is there's always nowhere to draw a proper line for causation, the possibilities are almost endless, and can change scenarios repeatedly. Take a look at Dwight's +/- last game, -12, the second worst on his team, despite playing one of the most dominating games you'll see from a centre, he was in total control of the paint on both ends of the floor for 40 minutes, he fouled Thomas out of the game, he treated Asik like a childr and then proceeded to beat Boozer around like an infant. Then what do you do? drop games like that? but which ones? or do you just hope that it evens itself out over the year. Another problem is that the stas are incomplete, the numbers of blocks doesn't tell you much, a block you keep control of the ball is worth something like 3x the value of a block where they get it back, is there a stat for controlled blocks? no (and hell for Rockets fans, that shouldn't be rocket science)
I never said this wasn't true, but you have a habit of putting words into people's mouths so it doesn't surprise me. At least you're consistent when it comes to putting words into a person's mouth. If you really believe leadership can be measured by how early someone shows up to practice, why don't you create a data set that presents us with this information so we can analyze it? Here, I'll use a different word instead of dogmatic, I think people like you are naive and don't realize the limits and flaws of all this +/- stuff. So it looks like it's you that doesn't understand how this sort of stuff works. Durvasa doesn't acknowledge it either until someone challenges him on it. I know a lot more than you think. Your formula and "model" did. I definitely agree with that. Then you should have been more explicit and said this in your original post. You can't just throw out a bunch of numbers and then not provide any sort of real accompanying analysis, which is what you did. Well if you're going to present us with quantitative "research," you should accompany it with analysis, which your original post lacks. For most people, that should go without saying. For others like yourself, I guess it needs to be spelled out explicitly.
It was so great having Yao and Dikembe on the same team. No one could score on us in the paint a couple of years ago. Now we have neither of them and that is our biggest hole on the team.
You are familiar with how linear regression works? You understand that the result of a regression formula has an error associated with it? That the estimates here are quite fuzzy? Given the "diff" column I provided in all the tables (and explicitly referred to), that should have been quite clear to people. Let me spell it out again: just because the formula predicts a certain +/- figure, that doesn't mean it is spot on. OK:
I said real accompanying analysis indicating that your analysis fell short given the amount of info you were putting out there. Only the last few sentences you bolded in your previous post begin to analyze the data. For starters you could have provided us with better analysis of the standard deviation or standard error of the mean. Instead of saying that maybe it indicates this or maybe it indicates that.
This sort of analysis? http://bbs.clutchfans.net/showthread.php?t=199587 http://bbs.clutchfans.net/showthread.php?t=180272 I guess I should have posted some fantasy numbers for a few players, added "these guys are BALLIN'" and "these guys are SUCKIN'" and that would have satisfied you. Sorry I could not meet your high standards. I'll try better next time.
Durvasa, I like the analysis. I also agree that I think +/- values have too much variability and too many causes for error. I'm sure you considered it, but I was wondering what you thought about using the Player's Defensive Rating or Player's Defensive Rating with some correction for the Team's overall Defensive Rating?
Part of the motivation in deriving a "statistical +/-" estimate for +/- is to stabilize the rating and smooth out the variability you refer to. You can see that's exactly what it does in the first post, though by doing so in eliminates other information that isn't captured by the boxscore. I mentioned that this approach is nothing new. I think the first description of using statistical +/- to stabilize adjusted +/- player ratings was given by Dan Rosenbaum in 2005. You can read more details about the approach in his article on his player defensive ratings here.
C'mon man you can do better than that. Anybody with an ounce of humor in their body could see that I was being less than serious in both of those threads. Nevertheless, even if I was being dead serious, you're comparing apples to oranges. I did not present a massive amount of quantitative data in either thread. You, on the other hand, did just that and your analysis amounted to only a few sentences. Then you seem to get upset when you're called out on it. I guess asking you to explain and analyze your own findings is too much. Here's a couple of books that will help you with that: Levin, Jack, and James Alan Fox. 2004. Elementary Statistics in Social Research, The Essentials 2nd edition. Boston: Pearson Education. Lewis-Beck, Michael S. 1980. Applied Regression: An Introduction. Newbury Park:Sage Publications.
If you believed the commentary in the first post was insufficient because instead of merely giving numbers for 1 or 2 players I took the trouble to post it for a large pool of them, if that was your point from the beginning, you could have said so directly in your first post. You could have asked for more details, and where I could provide them I would have, or you could have asked about my interpretation on some of the results. Instead you wrote the following in your first two posts, even though I very explicitly said that the formula was for making an approximation of defensive impact, provided the errors for each prediction on current players, and even critiqued the results on Marcus Camby and Ben Wallace: That, if anything, is what upset me. Snide remarks taking pot shots at posters while misrepresenting their views and ignoring what they wrote. I seem to remember this being an issue with you before.
