Let me first say that I'm by no means a game theory expert, so if what I write below is incorrect please let me know. We are all here to learn after all! I think that a GTO strategy is only mathematically guaranteed to work in the way you describe against a single opponent (once everyone else have folded with the ranges you both have then as a starting point).

As I remember it, when reading about game theory in the past, there are some issues with trusting GTO to much at lower stakes multiplayer games:

1. In multiway pots, GTO still works well as long as everyone else also tries to play as close to GTO as possible. But if a recreational opponent joins the game, he can make mistakes which may cost both him and you (playing a GTO-strategy) money. This money gets won by a third player. In order to avoid this it can be correct to deviate from GTO in multiway pots (as long as you do it in an intelligent and correct way).

2. (EDITED) The relatively high rake at micros and low stakes means we are not playing anything close to a zero-sum game (the game theory math behind the solvers converging towards GTO-solutions is based on zero-sum games).

3. I don't own MonkerSolver yet, but I think it uses different sorts of approximations and abstractions in order to be able to "solve" PLO. Thus, I think we should be somewhat careful with assuming that the outputted results are exact GTO-ranges IMHO.

I hope this helped at least somewhat! Since I'm only just now starting to play PLO and bought access to this course recently, I unfortunately can't comment on the specific starting hands you mention.

Just catching up on this forum and I wanted to add something as this is something I struggled with for a while as well.

The statement preceding this sentence is not correct.

"by definition, a GTO strategy is such a strategy that if our opponent deviates in their strategy (by "calling too much") then our expected return would be better than if they played the original (GTO) strategy."

The ONLY thing that a GTO strategy will do is guarantee a minimum amount of EV. It will not maximize your EV against an opponent who is playing poorly as it is not designed to counter act whatever specific mistakes your opponent is making. Only an exploitative strategy can maximize mistakes. Since this is such an important distinction, I recommend an exercise. If you have a NL solver input the following (if you don't just pony up the $80 and get GTO+ as understanding this concept will make you considerable amounts of money.) Set stacks of 100 for both IP and OOP, input a bet size of 100 for both IP and OOP. Input a board of 222 2 3 and ranges of all pocket pairs 55+ for both players. Run the sim and then notice the EV for the node where OOP has checked and is facing a bet and has to either call of fold. From there, node lock IP's betting strategy when checked to and then change OOP's calling strategy to either pure fold or pure call the hands it mixes with and run again. You will notice that the EVs are IDENTICAL regardless of whether or not you chose call or fold for the hands it mixes with. Once you have done that, go in and node lock OOPs strategy to calling all hands it mixes with and unlock IP's betting adjustment. Note how IP's betting strategy changes. Finally look at how IP's betting strategy shifts when OOP folds all hands it mixes with at equilibrium. Make sure to notate the EVs in all of the above scenarios. You will notice that IP's EV will dramatically increase once it is allowed to adjust whereas it will stay exactly the same if it sticks to an equilibrium strategy.

In small and mid stakes PLO games, your opponents are NO WHERE NEAR FUCKING CLOSE to playing a GTO strategy. If your opponents were playing GTO you would pick up the blinds almost exactly 1/2 of the time that you raised from utg at a 6 max table. Does that ever happen in ANY game you have ever played? No, me neither. At equilibrium, if a gto utg opener tightened up its range such that it only played hands that were massively profitable at equilibrium, then it would never get any action and its overall EV would decrease. The range that a solver calculates is structured such that it has achieved a finely tuned balance between strong hands and just enough weak hands such that it can maximize it's ev across all potential scenarios. A major part of this balance is having the ability to get action with its strong hands. In poker geek terms, it structures its range such that it gives its opponents some incentive to give its stronger hands action. The opponents you are playing with do not need that incentive and are willingly going to give you action even if you're the biggest nit in the world. They may say differently, but let's face it; degens gonna degen. As a result, because most of these games are so wild, you can exploitatively fold a good chunk of the bottom of a gto range (I have data to suggest that it is upwards of 30%). This bottom chunk would actually LOSE money even if a perfect GTO bot were playing it. And here's the other kicker. You are not a GTO bot. Neither am I. Neither is Sauce/Galfond/Whomever the best player is in the history of history. As humans we need to rely much more heavily on hands that realize their equity and have playability. The AAA combos are pretty much trash in a loose game. While the 2 aces are indeed very strong, the third is a dangler unless you have opponents who can actually regularly fold about a J hi flush. By electing to vpip these combos before about the cutoff, you are willingly sacrificing a card and effectively taking a knife to a gun fight.