Hi,

Chess engines comfortably beat the top humans but one of the few weaknesses they do have is poor positional play. So for non chess players, chess engines cannot calculate deeply enough to “understand” and use abstract concepts.

Is there a similar weakness for solvers in poker? E.g. maybe solvers are crushing in river play but have small leaks preflop?

Lol at your first paragraph, and lol at trying to compare games of complete and incomplete information

What an awfully long way to say “I don’t know”

Do you get style points in chess for the use of "abstract concepts"? If you don't, it's not really a weakness, is it.

Or a very short way of saying 'It is a question of inaccurate assumptions and lacking in precision, rendering it almost totally unanswerable'

If I were going to elaborate on Sixfours much more succinct response, it would go something like....

Can you be specific with regards to why you consider AI (and which specific AI enigne...not all programs use the same approach) to have an exploitable wekaness if they can beat the best human players? Can you describe the weakness in more detail beyond a vague reference to pisitional play and abstract concepts? Can you tell us which poker AI approach you are are trying to compare to (neural network based machine learning, or mathematically deduced GTO models)? Can you tell us why you think poker and chess are equivalent, even though they are very fundamentally different (aside from the complete versus incomplete information, there is also the matter of number of players, and variance.)

This seems like a silly, pretentious question that really cannot be answered in its current form.

Is that even true?
Even if it was, there are no points for getting into a "strong position". The engines just want to win the game. They are quite good at doing that.

As for poker solvers, there is the GIGO problem (garbage in, garbage out). If you input a terrible pre-flop range for one of the players, the solver will work out how to play that range in a way that maximises EV, but the strategy as a whole won't be as good as one with a better pre-flop range.
To use a silly example, everyone knows that 72o should use the line that gets maximum value on 222xx, and a solver could work out the optimal line, but 72o shouldn't be in the pre-flop range except in specific circumstances (e.g. in the BB facing a limp).

Humans should play more closed, positional vs. engines, but it might not be 100% clear that engines are weaker at those positions, it just saying that humans are weaker at tactics compared to engines.

Engines pick the most +ev positions in chess and most +ev lines in poker. In chess they calculate the material and positional plus and minus factors and pick the most +ev one (the best position that as so has the biggest positive expectation).

As so, I don't really think that either one thinks in terms of GTO, unexploitable play, it just ending as so in poker, while I don't see any GTO in chess in terms of "unexploitable," but maybe one can point out some balance for the defense and attack that needs to be there (relative to the ev one has).

Similarly, I don't think there is any specific street or position as a significant weakness. Analyses is made based on whatever street or position. The machines and humans do their best in both cases and one can think if the machine has some weakness or a strength compared to humans somewhere.

If we take the chess example, then humans would like to avoid the wildest, most dynamic streets (and poker forms) vs. engines; is that the flop? While the most closed, positional could be the first (positional, opening) and last (endgame) streets.

The engine will be up to perfect in the endgame, but I guess I would take my chances against it there rather than in the middle game where I would likely get killed more often, having some chances in the endgame.