I'm glad you posted this, because I've been thinking exactly the same thing for the last couple weeks. I haven't been part of any top-tier teams, so take my thoughts with a grain of salt, but I think there's several things at work:
- Powerhouse teams have a very strong understanding of the playing field, because they take time to build a nearly complete one and prototype with it. I can't tell you how many teams at the Oklahoma regional had bridge manipulators which were completely unable to push the bridge down, because they hadn't actually tested on a real bridge.
- Powerhouse teams are focused on performance. This seems obvious, but if you look at the top echelon of robots, they have one thing in common: they're fast. Unbelievably, jaw-droppingly fast. Great teams design their robots to do the difficult tasks quickly, and squeeze the most performance possible out of their motors and mechanisms. When a team moves from "what's the easy way to do this" to "what's the fastest way to do this", they move up in the standings.
- Powerhouse teams try a lot of things. For a while my perception was that great teams were great because they had a group of geniuses who could CAD the perfect robot and a machine shop to fabricate it. But as I've watched and read (particularly 148 and 254's build blogs), I realized there's a lot of prototyping, testing, iterating, and tweaking that goes on.
Obviously, these teams are "powerhouses" because they come up with these defining features year after year. But they develop these features because they are relentlessly pursing high performance designs. They aren't content with the first idea that comes to mind, or the idea that will be easiest to fabricate. Nor are they content with the first thing they build.
After watching a regional, most teams could go back home and design a really good robot, but by then it's too late. The trick that powerhouse teams have mastered is doing this type of learning and iteration during the build season - before they get to the regional.