It can't be optimal because the inputs the solver is receiving aren't correct. In fact it's impossible for it to have the correct inputs unless if its playing another bot. So basically it's always going to be solving a problem with inaccurate information.
The more I research ab out this topic, the more I realize how non-optimal GTO is. Like it's not even close against a lot of players. Your own intution is capable of giving you infinitely more EV than the computer could ever dream to achieve. Only people it's gonna give you an edge against are people you probably don't want to play anyways. So imo, it almost becomes irrelevant.
Quoted from Expert Heads Up No Limit Holdem Vol 1 by Will Tipton (I take no credit):
"Two players playing a zero-sum game will adjust and re-adjust until both reach a point where neither can exploit the other to profit.
Both players are aware of what is happening when they reach this point in their strategies.
Both players are experts at the game they are playing.
Neither player can do better by deviating from their respective strategies.
Both players can only do worse by deviating from their respective strategies.
Therefore, both players are incentivized to remain at this "equilibrium" point.
This concept of "equilibrium" is called the Nash Equilibrium.
The Nash Equilibrium is known as the solution concept for the game.
Equilibrium strategies are called "unexploitable" strategies.
Non-equilibrium strategies are called "exploitable" strategies.
The words "equilibrium", "game theoretically optimal (GTO)" and "unexploitable" have all been used interchangeably to describe the strategies to reach the Nash Equilibrium.
Of course, for a human to play a perfectly GTO strategy is currently impossible.
Therefore, we must seek to identify exploitable tendencies in our opponents in order to profit.
GTO is a conceptual approach to the game.
Therefore, GTO is a viable initial strategy vs an unknown opponent.
However, once we deviate from the GTO conceptual approach to the game, meaning we exploit our opponent, we ourselves become exploitable.
Also, in reality, no opponent is truly unknown."
A solver is just a fancy calculator.
A human inputs open-jamming range "x" at 10bb effective in a HU match, and solver calculates output calling range "y".
A human inputs open-jamming range "z" at 10bb effective in a HU match, and solver calculates output calling range "a".
Whatever the human input is will dictate the solver's output.
But a solver can also calculate the following:
Equilibrium open-jamming range "q" at 10bb effective in a HU match, and equilibrium calling range "r".
Values of "x", "z" and "q" will not necessarily be equal to each other.
Values of "y", "a" and "r" will not necessarily be equal to each other.