Thanks for your reply, lots of useful info there! See my reply to Abraham re: the 80% stat. It's an approximation for the sake of discussion.
Luck has nothing (or very little) to do with success in the software projects, especially in the complex ones. You just don't happen to stumble by luck on the Linux kernel code :) In my personal humble experience I probably had to write (and subsequently delete) more than 100 lines of code before I arrived at the final 10 lines which I was happy to commit to the repo. See, again the "90% of everything is crap" rule is at work here :) It's inevitable.
Bad luck can be a factor in the failure of software projects, of course. There are a myriad of ways to fail in a software project and of them is pure misfortune, but there are very few ways to succeed and none of them is luck.
So, people deliver software projects not because they are lucky, but because they know what they are doing. The ones that get promoted and get to teach others are not necessarily the clever 20%. Plus the 80% of those "others" are the less clever bunch so they will inevitably misapply the methods preached by the gurus.
I agree that just having the right people is not enough. However it's 80% of success (please don't ask me from where did I get the 80% figure :)), because the right people will come up with the right methods (Gawande is a perfect example of the right type of person -- he saw a problem and he found a solution for it). They won't call it Agile or anything else with a capital first letter, they might just call it a checklist, and they will know not only what to put on that list, but also how to get it done.