Confusion between different GTO solutions
Posted by SpankoPita22
Posted by
SpankoPita22
posted in
Low Stakes
Confusion between different GTO solutions
I've been toying around with the GTO Wizard trial playing against the NL50 solutions and I've noticed significant differences between GTO Wizard and a solver app I use called Solver+. So far it seems GTO Wizard has preferred the more aggressive actions on the river. I'm wondering why there's such a discrepancy? Is it because the rake is higher at NL50?
I also noticed GTO wizard has no restrictions on sizings used vs other presolved libraries. Does this make it a bit less practically applicable? I would have thought they should have narrowed the solutions down to the best sizes in any given spot?
I'm just confused about which solutions to trust, while they are mostly the same EV wise the frequencies differ wildly.
As an example:
My solver says that checking the river here is the highest frequency action (but all other bet sizes are the same EV as checking). But GTO Wizard says the check is a mistake and that I need to overbet the river. Solver+ also prefers a cbet on flop whereas GTOWizard checks
Loading 7 Comments...
Did you already compare the defined preflop-ranges?
Not familiar with solver+ and which openingsize this tool uses.
So if we have different open sizes and therefore different preflop-ranges, the solutions will be quite different already.
Yes, basically what raoulflush said. All of the following will change the solution:
I might add that the further into the game tree you go the bigger the discrepancies are likely to be. So, in the example you've provided where it's on the river in on a four flush paired board it's not particularly surprising to see different solutions.
GTO Wizard is significantly higher quality than solver+. The solver+ is fine for the kind of analysis you do on a phone, but GTO+ has more accurate preflop ranges, more robust solutions, and much bigger trees.
Thanks guys. You're right in that the rake is higher for the GTO Wizard ranges and the opening ranges are oddly also wider
Keep in mind, GTOwizard looks like it has 6 different sizings on that river. I really doubt that the solve you are doing has all of the same allowed sizings. Basically, unless the ranges are exactly the same and all post flop settings are the same along with rake settings and exploitability settings (accuracy of sim), then there are going to be differences in the solver outputs and as I said above these differences will be the most noticeable further into the game tree or when exploring lines that occur in one sim, but not the other.
Like other's have said if your parameters are different the results are gonna be different. If the asteroid that killed the dinosaurs hit the moon instead Earth would be very different and that is one parameter in the history of earth, mind you a big one. Basically its the butterfly effect.
Solvers are great tools, but Tools they are. Most if not all people have very limited solving capabilities unless you are gonna invest thousands of dollars in ultra high end computing power. Along with the amount of time it is gonna take to solve each spot. So people using solvers are using simplified solutions to approximate the out come.
It's a tool and should be used as such, I'll give two other examples. Haralbos has a detailed computer for NBA sports betting, But he still watches as much as he can and uses the computer as a guide but knows it can be dangerous and very wrong at times. Solvers are no different.
The other example is Chess engines, They play very strong Chess Stockfish is almost unbeatable for a human player(but can be beaten at times) But Stockfish which is a great tool for chess pro's gets crushed by AlphaZero. So what that means is Stockfish is most of the time finding the optimal play but it does make mistakes and is not 100% the stone cold answer. There are many examples on stream where Stockfish has shown a position equal that is totally lost. Also, Stockfish has different depths it can be set to and different levels of the program, GM's use a cloud based one that has far more computing power then the free version on lichess. Does this mean the free "solve " is bad no, its a great guide to what the right move might be, but its solved less precisely so it maybe slightly in accurate.
Use them as tools to help you construct ranges that are gonna be hard to exploit and make you money. Don't use them as the supreme answer. We can all come up with spots in solvers that the population plays very differently(exploits) that drastically change what we should do with certain hands.
Be the first to add a comment