[KJ o]

I thought this forum was chock full of Computer Science majors who actually learnt something, but this is obviously not the case. (Or rather, they have all learnt to skip ignorant ramblings like these.)

First of all, let's separate two things
1. A poker algorithm that can't be exploitable (i.e. beaten in the long run) (and remember that the long run is a *long* time in poker, and not a short tourney like the WSOP ME.)
2. A poker algorithm that beats particular opponents in a particular game at a rate reasonably close to the maximum possible level.

1 is difficult, but it's not *that* difficult, in partcular if one adds "beaten for more than the rake". It is almost definitely easier to write a poker bot which will not allow any human to beat a full ring of said bots for more than the rake (assuming fairly low limits, i.e. a significant rake) than it is to write a chess program that beats Kasparov.

Why, then, has the second happened, but not the first? For the very simple reason that hundreds (thousands) of times as much effort and money has been spent on chess compared to poker up until today.

2. is significantly harder, as it is basically 1. plus the learnings and adjustments. And it's debatale if it will ever be +EV to do this properly. It's not too valuable against world class opponents anyway and it is hopefully difficult to make anhy money on it against poor opponents.

This all hinges on one thing: that there is a game theoretical approach to poker that is not exploitable. This is not completely obvious, since the relationship between algorithm strengths could be non-transitive even for optimal algorithms.

I have seen several more or less convincing arguments for why there are optimal algorithms that are not exploitable, but nothing even remotely convincing suggesting otherwise. But I'd love to hear about any such argument!