The comment that "lowering your death rate" will have the most impact on your W/L stats is probably conflated. If you lose, you will almost certainly die (my experience has been there are very few cap/objective wins in QP/Group Queue), so you can't just decide to play differently to lower your chance of dying. What you're really doing is playing to improve your chances of winning, which implies solid team play among a group of relatively skilled teammates.
In fact, I'm reasonably sure that the most gaudy k/d and w/l ratios come from players who are both skilled and play group almost exclusively with other skilled teammates...
I say this part to rationalize my own performance, which is "meh" on the stats front, but I do derp around in QP a lot - let me have my egotistical illusions, please.
Also, the comment about the spread of performances being way too big and the indictment of the matchmaker as a result I think is related to the above. We know there are some highly skilled players in the game. If those players play a lot of QP, their stats probably aren't terribly gaudy, but the ones who focus on group queue and only play with their similarly skilled teammates can generate some pretty eye-popping stats.
That isn't a shortcoming of the matchmaker, though. If you let players play in groups (and I think that's a good thing), you're going to have situations that the matchmaker can't possibly compensate for. The only solution in the statistical analysis would be to only count QP drops, but as I understand it, the leaderboards are counting both QP and group queue, so you can't separate this variable out of the data right now.
Edited by Khereg, 14 July 2016 - 01:16 PM.