VIEW: People are too dependent on solvers
People are putting in a lot of work with solvers but aren’t grasping the rationale behind the solver recommendations. It seems everyone makes a play because it’s “solver approved” or “what the solver says” instead of considering WHY the solver says that.
This severely limits people as it is impossible to memorize every solver approved play you might encounter, but understanding the rationale for why the solver recommends it affords you the flexibility and adaptability necessary for a winning session.
They don't "play GTO". I can guarentee you that at the high rollers, they are looking for exploits against players and population tendencies.
The exploits you would use at 1/2 and in a high roller are different, they play completely differently, obviously. The biggest difference is that GTO will work MUCH better at 1/2 because the exploits are much bigger.
Can you find exploits without solver study? Of course, if someone never folds, you never bluff, you don't need a solver for that. But there's
If you are playing $1/$2, then why in da fuq are you studying charts?
That is silly.
Fundamentals and reads are far more important.
People misimplementing solvers today is no different to the guy who would raise-call off 25bb with AJ from the button against the 72 year old woman with shaking hands in the big blind in 2011 who has a range of QQ+AK+ and be like 'meh standard cooler'
They're a tool that make good players much better but obviously the average recreational player trying to implement them into live play is going to fail miserably
Hi Everyone:
I'm going to come back and go through some of the posts in more detail. But I wanted to make a few basic comments.
Recently, I've read three poker books where the author is often looking at solver results, and they tend to repeat them without understanding exactly what the solver is doing and/or why the solver does what it does. For one example, you can look at my notes for the book 100 Essential Tips to Master No-Limit Hold'em by J Little which you can find here: https://forumserver.twoplustwo.com/33/bo...
The solver can do a lot of things which if you're trying to compute the underlying Game Theory Poker Math will just be too difficult to do. But if you understand how the game theory poker math actually works, you'll be able to understand better why the solver does exactly what it does, and this should help your overall strategy.
Here are a few examples:
1. As the defender, the solver sometimes doesn't seem to call with enough hands. What's often happening here is that the defender is working in some raises, which includes bluff raises, so the defender doesn't need to call as often as appears necessary.
2. The aggressor sometimes makes a range bet on the flop. What's happening here is that the defender won't have enough legitimate bluff catchers in its range, so the solver will now have the aggressor bet every hand in its range (for a particular flop) and these bets will usually be small.
3. The solver sometimes holds back legitimate value hands by not betting them. But this is only done when the bet is large compared to the size of the pot (and when the bet is large the percentage of bluffs is also large). But what's not understood is that when the bet is very small compared to the size of the pot, it can be correct for the aggressor to make negative expectation bets because by checking the expectation will be even more negative.
4. The size of the bet that the solver prefers is linked to how many very strong hands the aggressor makes on the flop. So for an UTG aggressor, his bets will tend to larger on an AK2 flop than a J52 flop.
5. But when the aggressor goes to a smaller bet, he'll frequently bet more hands for value (but a smaller bluffing percentage) than when he makes a large bet (relative to the size of the pot).
And this list is not complete.
Mason
1. How do you determine what calling range appears to be necessary?
2. I don’t think that’s why or certainly not the only nor main reason
3. Legit value hands, how do you determine that ?
4. Preflop aggressor loses to many more hands on J52 than AK2 in a single raised pot. I believe that’s the main reason the size of the bet changes. Solver plays boards and possible run outs more than anything Else
Rock Paper Scissors isn't the best analogy because in RPS it's impossible to make an EV mistake against the GTO strategy. In poker if you don't know GTO you are going to be making EV mistakes all the time which gives the solid, non-exploiting player a winrate
Very good point, it's vs the player who never misses the correct combos but makes frequency mistakes all over the place that you can't expect to make a winrate.
You misspelled "Nash equilibrium." If one player is far from Nash, the game-theory-optimal play is to exploit the living **** out of them.
Agree about my wording, although I never made an argument that max exploiting even the slightest deviation isn't the optimal play. That's exactly what the solver do, for example, in that river toygame nuts-air vs bluffcatcher if you nodelock any deviation no matter how small.
Hi Everyone:
I'm going to come back and go through some of the posts in more detail. But I wanted to make a few basic comments.
Recently, I've read three poker books where the author is often looking at solver results, and they tend to repeat them without understanding exactly what the solver is doing and/or why the solver does what it does. For one example, you can look at my notes for the book 100 Essential Tips to Master No-Limit Hold'em by J Little which you can find here:
Two other things I've seen that just came to mind:
1. GTO allows you to take advantage of your opponent when he calls your value bets and also allows you to take advantage of your opponent when he folds to your bluffs.
2. On hands where you mix between a bluff and a check your EV is exactly the same whether you bluff or check.
3. GTO maximizes your expectation. (This one I'll answer. GTO maximizes your "minimum expectation" or "minimizes your opponent's maximum expectation" depending whether you're the aggressor or the defender.
Mason
People are putting in a lot of work with solvers but aren’t grasping the rationale behind the solver recommendations. It seems everyone makes a play because it’s “solver approved” or “what the solver says” instead of considering WHY the solver says that.
This severely limits people as it is impossible to memorize every solver approved play you might encounter, but understanding the rationale for why the solver recommends it affords you the flexibility and adaptability necessary for a winning sess
You wrote "This severely limits people ..."
While I agree with your basic point, I'm not sure how much it severely limits people. It seems to me that there are a fair number of people who are having at least some level of success who don't have a good understanding of GTO poker math. I know this from the books I've read recently.
Mason
This is pretty much what the majority of players "using solvers" are doing. I'm not saying that these mixed frequency combos aren't mixes, I'm saying that if your input sizes only included 33% and 50%, but that the actual optimal sizes vs an opponent would have include 41% that both 33% and 50% could only ever be "close enough," never actually giving you the highest EV strategy.
We've put this is a couple of our books. But when you look at statistical distributions, minimums and maximums tend to be broad. This means that if you're reasonably close, there won't be much difference from being perfect.
Mason
That sounds more like some sort of exploit because from what I've seen solvers bluff their asses off. Humans even the best in the world bluff way less than the little guy in your computer so bluffing less would be moving away from the solver
But solvers aren't playing a tournament, and in tournament play there are many spots where reducing your EV a little if it reduces your fluctuations by a lot is well worth it. Solvers are playing each hand independently from any other hands, and this might be the explanation for what you're seeing.
Mason
[QUOTE]Say, the solver says bet 70% and check 30% and you say "Bet". Well, if your opponent's strategy is fixed, those two options are the same EV.
Suppose your opponent calls 100 percent of the time. Is your EV the same when you bet versus when you check? I don't think it is.
But if you always bet 100% of the time, your opponent could adjust their strategy to make your bet exploitable.
This is true. But in your example above if you always bet 100 percent of the time and are already against an opponent who calls way too much, he won't have to do any adjusting.
So it's not "perfect" in any sense of the world. And in a world of people using trainers poorly, you might want to specfically study adjustments against people doing pure strategies where they're supposed to mix.
But I'm also not surprised that it happens because as a software developer and PM working on GTO trainer, you have to give people what they want. And people ask for a trainer where they "play a hand of poker" and then they either here it's "Perfect" according the solver, or it's not.
Minor point. If a solver could talk, I don't think it would use the word "perfect." It would say "optimal" and that's something different.
Mason
Guilty of this. When I first started playing, a chart of those groups pretty much served as an early guide in online NLHE SNGs and MTTs. Probably not that egregious when stacks were deeper but it was also informing my push-fold decisions if/when we got to shorter effective stacks. You can immediately envision some spots I must have had flat-out wrong – e.g. shoving or calling off with some hands that should have been folded, or folding hands with which I should have jammed.
It did not take long
The Sklansky Hand Groups were designed for limit hold 'em games and not for no-limit hold 'em games.
Mason
Incorrect. The whole theory of gto is to remain unexploited whilst exploiting villain to the maximum.
It's defensive in that it defines its offense first and then works back from there to plug as many holes as it can. Which is just good solid play anyway.
No. What GTO does is that it maximizes your minimum expectation or minimizes your opponent's maximum expectation. See the Minimax Theorem (Maximin Principle) and get a visual view of what a saddle point is:
https://en.wikipedia.org/wiki/Minimax_th...
Mason
[QUOTE]Here's a mathematical fact : there exists a GTO solution that maximally exploits the chicken farmer's strategy.
No. There exists a GTO strategy that maximizes your minimum expectation against any specific strategy, including the chicken farmer.
The best way to find that exploitative, GTO , chicken-farmer-crusher strategy is by doing solver work.
No. The best way to find that exploitive strategy which works best is to understand how the chicken-farmer's strategy differs from GTO, and then to think about what the best exploits might be.
Exploitaitive strategy means solver work, but it's been co-opted by people who mean "making **** up / playing by feels". Why? Because studying is hard. And there's a massive market of people who don't want to study but want to feel like theyre learning anyway.
I agree.
And sure you can beat the chicken-farmer without it but you'll beat him for a lot more if you study actual GTO/exploitative poker to figure out how.
I think it depends on how badly this opponent plays. But what you're suggesting certainly won't hurt.
And if you actually hang out in the poker community long enough, it's obvious that the vast majority of top players like Haxton, Chidwick, Foxen, Polk have all talked solvers quite extensively while the dudes hawking "you don't need GTO to beat 1/2" are still playing 1/2 instead of , you know, beating it and moving up.
Perhaps. But the strategies that should maximize your EV against bad players in small stakes games probably won't do the same against better players at higher stakes.
Mason
I feel as though there is a lot of talking past each other here. People saying the same things but just in different languages.
GTO is nothing more than taking the fundamentals of poker (as expressed by Sklansky decades ago) and computing them perfectly.
Take the infamous chicken farmer from this thread. Let's say his deviations from perfect strategy is that he calls too much (both pre and post flop). If his tendencies and ranges are input correctly, a solver is going to spit out play charts that
This is a very good post.
Mason
I actually tried playing GTO as an experiment heads up and got my ass handed to me. In the end there is nothing better than adapting to your opponent. Whoever adapts the quickest to what their opponent is doing wins.
Playing GTO can often increase your fluctuations. So, you may not have been playing as badly as you think.
Mason
I like turtles
has anyone considered the perspective that... 'People DON'T use solvers often enough'?
