Quote:
Originally Posted by LektorAJ
One difference though is the different branches of the chess tree can be solved independently of each other. The correct move in a given position doesn't depend on what you would do at earlier points of the tree and under what circumstances you would get there if at all, it only depends on the evaluation of the later positions. In a chess computer the task can be sent off to a different processor in the super computer to get on with. In poker you need to solve the whole tree together because you can't evaluate moves with particular holdings till you know what other holdings you get to a particular point with, which you don't know until you've solved the tree - making it a circular problem - which is why poker computers are way behind chess computers.
There's some truth in this. However...
HU FLH is essentially weakly solved, despite it only being studied for a couple of decades. HUNLH will be 'weakly solved' in the next few years. Chess is
nowhere near solved, despite the game being studied for centuries, and with computers running sims for decades. (Only endgames with a limited number of pieces are completely solved, but there are almost infinite ways to arrive at those endgames).
The Alberta Uni AI team chose to solve HU FLH poker because it's an easier problem to solve than chess.
In poker, everyone knows that AA is not just a better hand than KK, it's a lot better than
every other hand. In chess, no one's even sure what the best first move is! Since the game is so large and complex, we likely never will.