The key to the American team's recent success is forgetting about the IMO and just learning math. Instead of teaching mechanics-- techniques specifically for solving olympiad problems (as they had done a lot of before, and as many countries do), the idea was to just show interesting and fun things from real math and build strong intuition for mathematical problem solving through that. If you look at lecture's he's given in the past at MOP: http://www.math.cmu.edu/~ploh/olympiad.shtml, you can see they're mostly just highlights from his area of research.
When he was being considered for coaching he told the MAA he would run training for the IMO team in a very different way from how it was done and how and that choosing him would be a gamble: there was a chance that it would go spectacularly wrong. The gamble has seemingly paid off.
It's also nice that this is a quintessentially American way of doing things :) Just pursue what you find fun and it'll pay off. Don't over-optimize for the accolades.