penbeast0 wrote:. . . then do understand that +/- stats are really orthogonal to the rest of statistics, and the rest of statistics are not only weighed more heavily, but are much more widely known among this group.
Oh thanks, Doc, that's really helpful to those who aren't really into statistics.
lol, I'm sorry beast. I don't want to launch into stuff if people aren't interested, but any time someone is sincerely interested I'm happy to delve.
I'm going to reference a blog post I wrote. The full article is here, and if you want to know how I think about +/-, it's the place to go:
http://asubstituteforwar.com/2011/03/26 ... tatistics/But let me try to be really tailored to the context here:

The validity of a metric is how closely it aligns with what you actually want to measure.
The reliability of a metric is how consistent it is.
The ideal metric is both valid and reliable. In basketball, we basically don't have one of those.
Box score based stats score well on the reliability side of things. They're pretty consistent. If I see a PPG or a PER for a season, I can be pretty confident that the adjacent seasons will look similar unless something important changes.
However, those stats are not ideal on the validity side of things. No matter how much data I collect from box scores, there are still chunks of the game I'm not learning enough about (man defense, on court leadership, etc). These metrics there for have a bias that keeps them from being centered at the bull's eye.
Enter +/- stats. They have no such biases. All parts of the what happens on the court are factored in to the extent that they help a team. Hence, +/- stats are valid. Having this new valid stat to be used in conjunction with our existing stats is fantastic because it helps let us know where the bull's eye actually is.
However, +/- stats are much weaker on the reliability side of things. Most people grasp this, and are right to do so, but they often make the mistake of throwing the stat out because of this because they don't understand its unique value as a valid stat. Reliability for reliability's sake is worthless. While it would be wrong to say box score metrics have no validity, people able to grab on to a new stat with more validity is an opportunity no one should pass up. At the same time, the lack of reliability in +/- is one of several reasons why it would never do to abandon box score metrics, so both should be used.
Now, as you can see from the dartboards, one of the things about validity is that if I keep uping the sample size, I effectively increase reliability. This is the reason why we start using multiple seasons for +/-. More data drives us close to the ideal stat. From the other side of things, more data on a reliable stat doesn't improve validity, so there's much less value to doing it.
Of course, adding more seasons to the mix adds other variables that need to be factored in for individuals. If a player gets hurt, then his data is going to get thrown off. More generally, in sports that have brief peaks, you probably wouldn't be able to get anything great out of this. Fortunately, basketball has really quite good player consistency from year to year, so in general, we get very good data out of all of this.
So, TL;DR: When I say orthogonal, picture X & Y axis. Reliability is one axis, validity is another. You want to improve both. In your analyses, using box score stats helps more with reliability while +/- helps more with validity. Using both together will lead to the most robust conclusions.