I once saw somebody say that you don't need to worry about stacks in NBA because players don't vulture each other too much. I never checked it out, but saw something in data I was looking at recently for cbb.
Martin and Mickey are highly negatively correlated (-0.76, p < .0039) since conference play. The effect is still negative but not significant if you run it for the entire season, but I'm not sure why. It could be that they were playing soft competition in a lot of those games and just beasting, a fundamental change in their playstyle, or maybe just noise. I tend to put more stock into recency and conference play because it seems like a lot of players reach their "true" level in the conference grind, but maybe that is just my biased thinking. The data is pretty highly influenced by times when Mickey blew up and Martin did nothing (see scatter below). I've looked at pairs/triplets of players from other teams with multiple studs and have found mixed results, with some being very negatively correlated, some have no correlation, and others being positively correlated.
The takeaway is that some players *might* be strongly negatively correlated, so stacking both in gpps could be a bad idea, similar to stacking QBs/RBs in football. My intuition is players of the same position could limit each other (F/F) while a (G/F) stack might not. However, the effect could be offset with the possibility you hit OT minutes or some insane total when you stack. It could make for a fine cash play to lock up the value.
Another thing I see people say a lot is that certain guys are matchup proof. I was doing some research on Sir'Dominic Pointer earlier and decided to dig a bit deeper. I regressed his FanDuel points output onto site (home/away), spread, total, KenPom AdjD, and KenPom AdjT for all games played so far in 2014-2015 season. For Sir'Dominic, the only significant factor was the home/away split, and even then it was marginally significant (p < 0.10). A common theme is that St. John's is awful on the road, and at least for S'DP, the data seems to support that a bit. The coefficients on all of the other variables were so small that I feel pretty safe saying that he is indeed matchup/tempo proof.