If you have a table full of clowns who are playing a range of strategies (ranging from being spiteful to just all-ining every hand, to barely knowing how to play poker), you will very quickly go broke with GTO, because you didn't adjust.
If a GTO player has negative EV in heads up poker, he is not playing GTO.
You don't understand multi-player poker and I see this meme repeated on many poker forums. It's possible for GTO to be worse than other strategies, based on suboptimal strategies of other players at the table.
Ignoring the fact that there is no known GTO solution for more than 2 players, GTO only applies under the assumption that everyone is acting independently for some fairly formal and strict notion of independence. This requirement is broken in many real world games with more than 2 players potentially playing loose and wild, where they may not actually be colluding or coordinating with one another, but they behave as if they were.
Imagine a situation where the collusion was explicit, you are playing GTO against two opponents who are colluding with one another. A Nash equilibrium will still exist in that situation but your EV will be negative. Now explicit collusion is considered cheating but you can imagine a scenario where two players play in such a way that even though they are not explicitly colluding with one another, they behave the exact same way as if they were. This happens among a group of loose players where their actions as a group rewards the group as a whole against players who are acting independently.
In other words, just because a strategy is optimal does not mean that it results in a positive expectation. You might be in a situation where playing perfectly just minimizes your losses, but you will still lose anyways.
If there are weak players they can effectively end up colluding against a stronger player who is playing "More Optimally".
Understanding the table, is a key skill. There's still a human element in the ring game.
I agree with you that you can probably construct some pathological multiplayer situations where collusion forces negative EV on a "GTO" player - assuming GTO exists in this setting.
Going back to the original parent comment:
> If you have a table full of clowns who are playing a range of strategies (ranging from being spiteful to just all-ining every hand, to barely knowing how to play poker), you will very quickly go broke with GTO, because you didn't adjust.
I feel pretty strongly that this is not representative of practical reality. I don't believe that caveats about implicit collusion against a GTO player should prevent one from calling out a harmful belief that is also provably wrong in the heads up setting (the setting where GTO is actually known to exist).
In practice, I have never seen anyone who played remotely close to Nash have negative EV against very bad players. On the other hand, I have heard a hundred times someone complain that they are losing money because their opponents are "too bad" (for their presumably excellent strategy to work as planned). This is close to ridiculous in most practical settings and also quite self-limiting.
In heads-up, no-rake, NLHE implicit collusion is not possible, so yes playing GTO has a non-negative EV.
>I agree with you that you can probably construct some pathological multiplayer situations where collusion forces negative EV on a "GTO" player - assuming GTO exists in this setting.
GTO always exists, but in the presence of implicit collusion it may result in a negative EV.
> feel pretty strongly that this is not representative of practical reality.
GTO is not representative of practical reality. Very very few people, if anyone, really plays GTO poker consistently. It's an excellent theoretical result, and using GTO solvers and simulators can build up good intuition, but the idea that there are these poker savants playing GTO poker in 6+ handed games is mostly a myth.
Also your claim that you have never seen anyone play against poor players have a negative EV kind of gives away that you are likely overconfident and possibly naive about your position.
I highly doubt you have been in a position to calculate someone's EV while playing poker, or assess whether someone is playing GTO poker against multiple opponents who are bad. And assuming you did possess this amazing power, GTO poker does not guarantee against losses against bad players over any particular night or week of poker. So if you've never observed that then you are likely biased in your observation. You forgot/underweighted the scenarios where someone playing really well lost money to bad players and overweighed the scenarios where someone playing well made money against bad players.
My experience is that talk about GTO poker is mostly used by people to add a more sophisticated and technical dimension to poker, people like to say that a poker play was GTO because it makes it sound authoritative and you can geek out over it, so to speak. People do the same thing across many hobbies including sports.
In reality the best advice for playing poker if you want to be profitable is to find fish and play against them. The sad truth is that there's no shortage of gambling addicts to take advantage of, or people who overestimate their abilities but lack discipline, or heck are simply rich and play poker socially and have no problem losing money and have fun doing so.
That's far more likely to yield profitable results than pretending that you can play an incredibly volatile and dynamic game like poker with anything remotely resembling game theory optimal.
I see good poker players lose to bad players in a single session all the time. Even if you are crushing with a +15 BB/100 hand winrate you’re still going to lose 30-40% of your sessions. That is not an indication of negative EV.
I’ve never seen someone who understood poker very well have a losing record over a long period of time against bad players. I don’t disagree about humans never playing actual GTO, which is why I use phrases like “remotely close” to optimal. My point is that the bar is low enough that any reasonable deep attempt at learning and approximating optimal play should clear the bar to be +EV against bad players.
Parent comment says “if you’re playing GTO against clowns you’ll go broke because they are too bad and you are too GTO.” I strongly disagree with this - the reason you have negative EV is not because you are “too GTO.” If you losing money in expectation you are just not very good at the game yet. What is needed is to understand the game better (and moving closer to optimal play is a reasonable goal). I suspect you agree on this - any reasonably good player should have positive EV against fish, regardless of whether they go out of their way to exploit.