Help me settle this discussion with a friend.


If one was playing against a pure gto bot and the only mistake made by the human player were mixing errors, would we still break even?

Yes you only lose when you make an EV mistake.

Why break even? Is the assumption that a GTO bot is unbeatable?

Yes, that is the essence of GTO

Presumably, a GTO bot would use solver solutions, which provide optimal strategies against optimal strategies. If you're playing against a non-optimal strategy, the solver output is not providing an optimal strategy. That's why nodelocking is a thing. That's why you don't 4-bet K6s against preflop nits. Etc.

Absolutely. But say you play against a GTO robot, and that we use the same solves that it uses, only that we don't mix optional actions but just choose one or the other 100%. Will we then be losing or breaking even vs the bot over an infinite sample? Hte robot will not adjust to our imbalance.

Bot strategy is making those actions indifferent(-ish), so you may pick either of them for the same(-ish) outcome.

Breaking even. But if the bot realizes you don't use GTO frequencies then it could adjust its strategy to exploit your imbalance (and then you are now both playing non-GTO, and you lose).

Theoretically yes, if you don't take any 0% frequency lines
In practice no, you will end up taking some 0% frequency lines

A GTO robot won't adjust. A bot that adjusts needs AI for its engine and would destroy humans much harder than a GTO bot ever will.

The ideal (non existing) GTO bot has an answer for every possible size from preflop to postflop, every situation on every board with every possible stack size. It would take a gazilion hours and a trillion gigabytes of data hence it doesn't exist. It won't adjust tho, you just can't go Stefan on its ass.

The ideal poker bot vs humans uses GTO as a baseline but it can invent adjustments that it keeps learning on small to big data sets. It can extrapolate from a single showdown where it should adjust other parts of the game tree and it can of course exploit based on extensive tracker like databases.

I agree. However, the question I was debating with my friend was whether a GTO bot (non adjusting) would win against a human player who employs the same strategy but with significant mixing errors. So the human is theoretically unbalanced/exploitable, but one side thinks as long as the inbalances are only mixing errors, the bot will remain neutral EV against the player in the long run and not come out ahead.

If you are still taking GTO lines but making frequency mistakes then the bot would only be able to beat you if he knew those mistakes.

Yes that, and a bot that makes mixing mistakes would be beatable by a human in theory.