Has this actually been verified, at least in the modern era (last 10 or so years)? Yes, you see the same guys on TV and to a lesser extent, final-tabling multiple tournaments. But these guys are often on TV because they are entertaining, not necessarily great (or consistently-winning) players. They typically all have external sources of income (sponsorships, poker training, commentary, casino ambassador, company owners, etc) and most reduce volatility by buying and trading pieces of each others' action. In addition, for tournaments, some guys are able to simply enter way more events, which naturally leads to more final tables, but you're not seeing all the losses and bust outs.
All I'm saying is that it's not the same as the 70s through 90s where the skill gap was so huge that yes, certain players could (especially by going on the road and playing private games) absolutely dominate. I think winning consistently today is more of an illusion and some short-term luck.
Winning consistently today relies on being as close as you can to this "game-theory optimal" strategy. As skill gaps decrease chance certainly plays a larger role, but AI analysis of poker has indicated that there are ways to improve your odds and gain an edge.
That's a good point about the cash game format.
But tournament format often forces players into multiple all-ins over the course of the tournament. This leads to a much more luck based outcome.
While poker does have a significant element of chance in a given hand, it's no accident that top players consistently make more money over the long run.
Knowing how much to bet based on your odds of winning a hand, figuring out what your opponent's hands could be based on how they're betting, etc. all take skill. There's a reason poker AI can consistently beat humans over the long run, a feat that would be impossible if poker was all about chance.
In fact, poker AI was a difficult area to create software in precisely because the game is one of imperfect information.