Originally Posted by Dumpy

OK. If I may point out a deficiency with this sort of analysis, it ignores the impact of the absence of other players. For instance, using Carter as an example, I suspect that most of the time that he plays with Kidd, Jefferson is also on the floor. However, I also suspect that when Carter plays withoutKidd, he is also playing without Jefferson. Coach Frank would often go with a lineup at the end of the first quarter and beginning of the second quarter that consisted of four reserves and either Carter or Jefferson, but never both. So, captured in the Carter analysis would be the impact of playing without Jefferson.

To really, truly, get a sense of how Carter would perform without Kidd, you'd have to look at how Carter played with Jefferson, Collins, and Krstic (the other starters), but not Kidd. That may actually never have happened unless Kidd was injured for a game or two. I suspect you could get to the same results by looking at how Carter performed with and without the other starters, and use those results to adjust the with/without Kidd numbers. Obviously, this would be more work than is really worthwhile, and in the end all you'd get is the difference in the overall impact between Kidd and the backup PG on the Nets roster, and it isn't as if Jacque Vaughn would ever be the Nets' starting PG if Kidd was hit by a truck or something.

[and don't take this the wrong way, I think it is fascinating, and good work, but we just need to be aware that you might not be measuring what you are trying to measure]

Anyway, I guess part of my point is that the absence of Jefferson and Krstic will affect those numbers just as much as the absence of Kidd, and I suspect that there is a high covariance (in other words, Carter either plays with all of them or none of them).