I think there is a lot of fundamental misunderstanding of solvers in this thread.

i think its funny when someone claims a solver couldn't beat their donk infested 1-2 or similar stakes game because its gonna get confused by the wild play or whatever nonsense. Even a non node locked solver would crush due to all the mistakes being made. It would passively exploit the hell out of the game. A properly nodelocked solver would crush even harder. Not sure why people like OP believe that if someone could actually sneak GTO+ or pio or wizard to the table it wouldn't send every other player to brokesville if they played long enough

Publicly available solutions actually are really far from live 1-2 play.

In live games people often open to 6-10x ... 8 people limp and someone isos to 20x ... GTO Wizard solutions open to like 2-2.5x ... and doesn't even have preformulated solves vs limpers ... so assuming you could even find the spot you're in quickly, it would be kind of hard to use GTO Wizard since none of the bet sizes or ranges are going to match what your opponents are doing.

Yeah, if a solver could solve on the fly and with inputs that are even close for whatever spot you're in - you'd crush. I'm not saying solvers are outputting the wrong solution - and I'm not saying what people do in live games is good ... because it's definitely really bad - just saying it's a solution for completely different situations compared to what you commonly face in live play.

kevmode are you the guy who wrote about a stabbing in the poker room but more of the post had to do with your choice of wearing cargo shorts instead of the stabbing or am i confusing you with someone else?

You are misrepresenting what people are claiming.

I think a solver can easily beat a donk infested $1/$2 game. No doubt.

I also think a player with a solid understanding of poker fundamentals will crush the game far more than a solver that isn't focused on beating a silly mess.

Yes, solvers will do fine. Eventually they will break the others at the table. An exploitative player will do it faster.

Maybe you believe the solver can win but op clearly believes a solver wouldn't be efficient against themselves or Chris Moorman I take not efficient to mean the solver would lose. Which is of course non sense. If the solver is just playing gto it crushes. If it is properly nodelocked even more so, only way solver loses is if you force it to play a horrible strategy

That's effectively what we mean here. We don't mean players who play like computers (impossible), but those who emulate the known solver outputs in a wide array of spots.

My point is that even emulating GTO at live games really isn’t possible. Too many people are opening big, have limping ranges which affect their open ranges, and take many other actions which create spots that either aren’t available in published solutions or aren’t similar enough to published solutions to be that useful.

If you have solved and are familiar with the actual spot you’re in (such as iso’ing a limper, flatting a 5x open etc.) and the ranges for your solve are close, then that would be different - but it’s not GTO at that point either - just an optimal response to a suboptimal play.

Recs are doing so much weird stuff you will never be able to solve or come close to memorizing what the solver did in every spot … but having a good grasp of math & theory can get you closer to playing optimally in unfamiliar situations.

Using a BTN flatting range for a 2x open size from GTO Wizard wouldn’t be great (could even be -EV) if someone is opening 5x, depending on what their range is.

This. There isn't a single game in the world which you would want to play in where you would ever want to play "GTO" as opposed to the old "look at what they are doing and exploit their mistakes" as has been understood in all games since the dawn of time

Correct. That doesn't stop the swath of internet players doing exactly that when they play live though. Exploitative play will always be king.

I think it's hilarious how every player at my table just outs themselves as GTO wizard users by using the exact same sizing schemes as it does....

Yes so you would need to use a native solver and run sims where people do stuff like open 5x. Sure you can never emulate the solver but you can still learn generalities and heuristics. Remember though op is actually claiming they can beat a solver. That's the only issue I have. If you wanna say you can beat a bunch of bad players for a higher rate than a GTO strat fine, Not sure if I agree but it's at least a reasonable statement. The claim that a solver can't help you make adjustments is also false. Solvers are quite good at making adjustments. You just need to give it good info. Which admittedly may be difficult when dealing with a bunch of yahoo's playing half the deck and most every flop being multiway

Getting back to opening sizes a little common sense goes a long ways, if someone opens 5x UTG, the first adjustment would be to fold all your 0ev calls and even a lot of your hands that are marginally +ev

This statement shows that you fundamentally do not understand what GTO -- game theory optimal -- means.

GTO play is fundamentally exploitative, sometimes massively so.

For example, if we somehow know that the villain who otherwise plays well is bluffing on the river a small percentage less often than the Nash equilibrium frequency, the GTO play is to Stop. Bluffcatching. Completely. The response to a tiny deviation from equilibrium can be a massive one.

Solvers find solutions by having the simulated players exploit each other as much as they possibly can, until ultimately the play converges to the Nash equilibrium, where there is no more edge to be gotten by either player trying to exploit the other.

Do not confuse GTO play with play at Nash equilibrium.

The only real point I'm trying to make in this thread is that published solutions won't help much in live games ... because most of the time before the action gets around to you, someone has already done something that deviates from published solutions in a big way, sometimes in multiple big ways ... enough that you won't be able to use a solution to effectively play against their flawed strategy using any published solution.

However if you have solved the spot and are close to what they're doing ... then that would of course be very effective.

I make zero claim to what the OP believes.

I also know that "nodelocked" in the crapshoot games that are $1/$2 or $1/$3 is a meaningless term.

Yes, a GTO player wi win in these strange games just because it won't do something stupid, but I also contend that that if you say you are playing GTO, yet are including exploitative elements 99% of the time are you really playing GTO or are you just playing "poker"?

This brings up an interesting question I've had for a while. We know in an equilibrium that both strategies and ranges are known perfectly. So what, if any, +ev plays a solver is doing in the equilibrium become pure mistakes because of wrong assumptions about the strategy and ranges used?

For example, let's say a solver is assuming that AK/QQ/JJ is pure 4bet in a spot, but the player actually flats those hands a high % of the time and even KK and AK some % of the time, making their postflop range much more uncapped. I fail to see how this couldn't lead to a solver making pure mistakes and even losing money overall depending on the intensity of false assumptions, but maybe I'm wrong?

Depends on how you define "pure mistake"


If pure mistakes means that the solver's action is -EV, the answer is none. A bot playing GTO at equilibrium will always make at worst neutral EV decisions no matter the opponent.

But it wont always take the highest EV line against a specific opponent. So you could consider that to be a mistake if your goal is EV maximization.

But the equilibrium is computed given assumed hand ranges (perfect information). When those ranges change, doesn't the equilibrium and the strategies defined by it change too? For example, let's assume the solver picks a bluff combination on the river partly because it blocks some of the opponents value range. What if that value range isn't even in the opponents hand range (it pure folds pre)? Isn't the solver then making an error?

Yes, it is only an equilibrium under the given assumptions (ranges/bet sizing). This is referred to as unsafe subgame solving. There has been some work on safe subgame solving but it turned out to not really be necessary. It turns out to be quite difficult to exploit unsafe subgame solving if the root of the subgame is at the start of a round.

Yes, the maximum EV strategy will change. But that doesn't mean that the equilibrium strat becomes -EV. Thats why I say it depends on how you define mistake.

Only in the sense that if the solver knew that villain couldn't have specific value hands, it may bluff more often and value bet other hands more aggressively and for larger sizes, with would increase its EV. But the solver is still going to be balanced with bluffs/value and put villain in an indifferent spot. Its going to make a 0EV bluff at worst, not a -EV one.

So this is more what I mean. In the first photo, we've computed an equilibrium using an assumed 3bet range from our opponent and the solver tells us to do this facing a 1/3 pot cbet:



Now just imagine that our assumptions about our opponent are wrong, and his 3bet range is not that but more looks like this:



We re-compute the equilibrium for his actual range while keeping ours fixed and this is our new response to a 1/3 cbet:



We can see in this second image that our response differs significantly based on our changed assumptions. Not only is our ev lower across the board, there are quite a few hands that originally mix in the first equilibrium that now pure fold. A pure mistake in this case would be to take an action the solver never suggests. If the action had the same ev as folding, the solver would mix, and because the solver doesn't mix on these combos we can safely conclude that taking the original raise/call mixed actions are not 0ev, but -ev instead (JJ,TT,77-22, etc). This is what I mean by assumptions being wrong messing with the whole model. Different equilibriums will be generated from different assumptions (including preflop ranges), so I don't see how playing like a GTObot can't result in -ev decisions when our assumptions are wrong, at least sometimes or on some combinations. I'm open to being proven wrong of course.

Ok I see what you are saying. Maybe specific nodes of the game tree can become -EV, but the entire strategy (sum of all nodes) will still be 0EV at worst.



now how does villain's strat change effect other parts of the game tree. I would imagine that if you are getting 3bet less, then the hands that you do play become higher EV pf, and the EV of the IP player in single raised pots will increase as well.

Oh for sure well if villain's moving 8% of his 'should 3bets' into his call range or not defending wide enough as well, then we either make more money IP vs. a wider calling range or steal his blinds more often, which makes up for our reduced ev in 3bet pots. 100% with you

Thank you for this I have never heard of it and will look into it.

the EV loss from the player not 4 betting everytime they have the pure 4bet hands is going to essentially be larger than the mistakes the solver made on subsequent streets

but yes naturally a solver will punt money in some later part of the game tree if earlier part of the game trees assumptions are off, if it never did it would mean its exploiting perfectly every opponent