I think youre wrong, man. Solvers teach you what you should be doing, but people dont do what they are supposed to do. Just watched a RIO video from a 500z reg that he analyzed his own session, he said at least 4 times " this is a call, if I fold here I would be overfolding", then he folded in the recorded session.

Recently a reg 4-bet called 97s vs me on 200z for 100bbs, which probably every player in the world knows its a -ev play, even whales know its terrible.

Imo poker will always be beatable because we are humans, we tilt, we are afraid. And if you say bots can be a problem, any site can detect them if they really want, just gotta put more effort in security.

And tbh, a gto bot would be losing or a small winner at best at 50z on stars with 8bb/100 in rake and 0 rakeback. Explo poker has never been so alive since PIO was released.

Nobody is saying poker will be unbeatable because of the solvers teaching people to play perfect. People do dumb **** all the time, even extremely smart people. That's never going to change. Live poker will still be profitable.

But online poker will die like the dodo because of gto bots. And live games will be at risk when ever smaller devices can be used to solve poker.

It's a losing arms race for the sites to detect bots. How are you going to detect that a human is playing on one device but using another to assist with decisions? And the financial incentive simply isn't there for the sites, cause most customers don't care enough to pay for extra security. So I doubt they will solve the bot issue.

Don't forget we have already run this experiment before and we know how it ended. Backgammon used to be fairly popular as a gambling game, yet there is basically no real money online backgammon being played. Why? Because they've had close to perfect bg bots since the late 90s.

A gto bot will have a higher winrate at 50z than a human playing explo poker. Sure, the value of being balanced doesn't matter that much at 50z, but don't forget all the small gto mistakes that even an expert is making with every decision. This adds up real fast.

Youre wrong, with such high rake you need insane margins on people to crush the game, such margins are very low when you play a proper gto strat because a lot of the gto plays are made to defend the bot from getting exploited.
Just as an example, AA is a frequent(>20% probably) check back OTT on 8237r after cbetting flop. That play is very bad vs population on 50z. Theres no need to defend your x back range, specially when people arent floating flop as they should and wont turn hands into bluff enough.
And poker sites are professionals, they have people that are specialists in that area, botters are probably random guys that are good with programing and poker. About assist tools, the user will have to input every flop on every hand they vpip, it would be impossible to put some decent volume with that.

Of course this is nonsense. I would be surprised if a gto-bot would have less than 30bb/100 vs. the average heads up 50z pool on stars, pre rake. Libratus' winrate was 15bb/100 vs. really good humans. Those 15bb/100 comes from the humans making playes that are strictly worse than an alternative. Really good humans do this fairly infrequently. A 50z player does this all the time. If a strong human goes to the 50z pool on stars and tries to mimic GTO play completely, it would be a bloodbath. If you put loose the actual GTO strategy it would be a bloodbath of epic proportions.

I'm talking about a 6-max table, so the game is quite different in there. I've played a lot at 50z and a game that would mimic GTO would perform pretty marginal in that game.

Who do you think has a higher winrate on 50z, otb redbaron or the best reg at 50z?

Rapidesh,

You're not understanding how crushing a perfect GTO strategy is vs any non-GTO strategy. Every single hand, every single play, flop, turn and river, multiple times per street, the human is making plays that can deviate from GTO. That gives an edge in favor of the AI. The AI never gives back even a sliver of edge. So when you add up all those decisions you end up with a monumental edge. Not to mention the AI never tilting, misreading board, misclicking, etc.

You don't need volume if you are crushing your opponents for a huge winrate. Just single table 500/1k+. Or you can run a video camera that captures your screen, calculates each table for you and tells you the move for the current table. Then you can 20 table NL500 or whatever and crush.

Realistically the only way to prevent cheating by software assistance is live play.

Who cares? It's two flawed humans. Neither of them is playing even close to GTO compared to AI.

50z regs rating themselves better than the baron itt, probably time to lock it up

Don't be ignorant, man, otb focuses on beating the best poker players online, he trains a totally different kind of skill than people playing micro and small stakes. Beating donks at a higher winrate won't make one a better player, otherwise hellmuth would be considered one of the best in today's games, even with his super ******ed plays.

By comparing otb vs a weak reg, I was comparing how explo could be better in a game given game than a GTO-based strat. And do you know that there is way more than 1 GTO solution to poker, right?

GTO isn't a static strategy that does only 1 thing vs all opponents, but deviates depending on how each opponent plays. The input to that software would be way more than the board and pot sizing, but villain's range, tendencies and your range at a given pre-flop scenario.

It's a bad comparison because otb isn't close to gto. And the answer is "no", a human playing explo will never do better than a perfect GTO AI in any game.

ah man, you were starting to make some sense, and then you had to write something that couldn't be farther from the truth.

A given gto strategy is 100% fixed. It does the same thing vs every opponent. It has zero interest in it's opponents ranges and playstyles. In fact, a gto strategy can tell its opponent exactly what it does in each situation (essentially give away its entire playbook), and the opponent still can't beat it. That's what makes it gto.

you are utterly confused about the way this stuff works

great posts Wolfram. and commendable patience responding to the uninformed opinions being thrown around by especially 7OAD and Rapidesh123

even in the face of this information and knowledge they will probably continue to try and argue their confused points though

I'm not an expert on GTO, but this doesn't sound right. From what I know, GTO is a defensive strategy that guarantees no loss against any other possible strategy, it doesn't say anything about winnings. Consider a game of rock-paper-scissors. If you know someone likes to choose paper a lot, then you choosing more scissors then what GTO says (33.33%) should be a strategy which will produce bigger winnings. I don't see why the same wouldn't apply to poker. Imagine if you play against a passive fish who almost never raises and mostly just checks, but loves to call. What do you think is the best thing to do with a middle strength hand out of position?

I'm not so sure GTO would do so great at smaller stakes where it would do a lot of bad bluffcatching and miss value when trying to check-raise instead of just going for a bet. Sure, it should be a winning strategy but I doubt these winrates would be something out of this world.

We should probably move the general GTO theory to another thread but I'll respond to this one here.

Yeah, it is defensive in the way that it doesn't attack a specific weakness. However, It has two guarantees. (a) it never loses to any other strategy, including GTO strats, and (b) it wins money from any non-gto strategy.

True. But in this case, you know what the GTO strategy is, i.e. choose each option 1/3 of the time. If you know the GTO strategy already and you have extra exploitative information on your opponent, then you can maximize your profits from deviating from the GTO strategy in the correct cases. However, in poker, we humans don't know how to play GTO.

A fish makes mistakes in every single hand they play. A human explo strategy catches some of those mistakes and exploits it, but nowhere near all of them. A human explo strategy is also randomly giving value back to the fish by misplaying many spots.

A gto strategy never gives away any value. It just doesn't fully exploit some spots, but it crushes in all others. It will never make "bad bluffcatches", just neutral ones.

To be the ultimate winner though, it would be best to know what the GTO play is as a baseline and then deviate when your read tells you that you can exploit. But we're still missing the first part as humans, and might never get there.

I was actually in your boat on this before, and this point was argued theoretically back and forth. But now we have experiments (Cepheus, Libratus...) that show what a near GTO strategy does vs human opposition. And it is basically "out of this world" (15bb/100 vs experts, no human comes close to that).

This is actually false. It only wins money against strategies that sometimes selects a strategically dominated option. And that's significantly different from what you said.

I didn't know that. So a non-gto strategy can go even vs a gto strategy. I guess that makes sense in some cases, like rock-paper-scissors. A 100% rock strategy will go even vs the GTO 1/3 rock, 1/3 paper, 1/3 scissors.

But does it also hold for a game like poker?

In rock-paper-scissors, the GTO-strategy won't have an edge on any other strategy, because there are no available moves in the game that are strategically dominated. In poker, there are many moves that are strategically dominated. Assume the GTO play with a certain hand in a certain spot is to call 82% of the time and raise all-in 18%. You would not lose money against the GTO strategy if you raised all-in 15% and called 85%. The only way you would lose money against the GTO-strategy would be to do something other than raising all-in or calling; such as folding or raising to a bad size. The fact that the GTO strategy is sometimes raising all-in and sometimes calling, means that calling and raising all-in have equal EV. And the GTO-strategy does not deviate to exploit your exploitable frequenzies. This means you can be vastly overbluffing/underbluffing/overcalling/overfolding in many many spots and still not lose money to the GTO-strategy. Yet it's extremely hard to never select a strategically dominated move, but nowhere near as hard as executing the GTO-strategy. And I allready commented on, earlier in the thread, what winrates I think implementing the true GTO-strategy would yield.

it seems you should play more 50z/100z/200z to see what a "bad bluffcatch" looks like. There are spots, specially OTR where vs population almost any bluffcatch is losing at least 10bbs, and Im being generous, vs some guys out there it's more like 30bb

Also youre underestimating the differencies of a 6-max game and HU game. On 6-max a lot of tight strategies are now unlocked. You would be surprised at how people play close to gto at a 6-max game today. As an example, usually OTR most bluffs should be close to 0 EV, while some are +ev because of strong blockers, lets say the gto bot plays vs a table filled with nits that underbluff, the nits strategy will be somewhat close to a gto strat. Ofc it will leak ev in some spots, but it wont be a bloodbath like you say.

Ive watched a podcast from a very gto german mtt reg and he said erik seidel plays very well, even after pio his game is still very strong without studying with the new technology. In HU ofc humans will have a tougher time, because they have to play loose and have way less viable tight strats available to them. Its also way easier for a human to play a correct tight strat than a loose one.

This thread is for discussion of all things related to GTO in poker.

I will move a bunch of posts from another thread (the Brains vs AI Challenge thread) into this thread to get us started.





Edit: Perhaps confusingly, those prior (moved) posts will appear above my post since posts appear in chronological order.

The same impending demise of online poker that everybody has been talking about for years and years? It's true that I can still find some awesome games online, but I'm quite sure you know what I'm talking about.

Second time you've trotted out this bull**** which gives zero responsibility to the people responsible. You're just like, welp, these awful people are gonna do these awful things, we can never stop them, we just have to learn how to take it in increasingly unpleasant ways!

Once again, GTO people/bot developers et. al could very easily construct a game to test their software, and they could make this game explicitly as simple/complex as needed, but they instead decide to help destroy the livelihood of poker players around the world because more rando people (who aren't gonna care much about GTO or solvers or whatever) are intere$ted in the$e guy$ $olving a game that they know about.

OP raises a legitimate point. Exact same thing went on with Starcraft AIs. They win matches mostly by achieving far more actions per second than any human can do. As a result they have a bigger army and more resources to win with. They'll be beating the best humans long before their strategic game is superior to the best.

Jesus guy, get off your high horse. You are a poker player that (presumably) makes a living exploiting people's weaknesses for money. The bot makers are pursuing their own interests and there's nothing immoral about what they are doing. They are the iphone to your Nokia, simply just out-competing you in life. They don't owe you a god damn thing and their work is infinitely more likely to be good for society than yours.

History is on the side of the innovators, not the buggy-whip manufacturers.

no, i really don't. games are good live and online despite proliferation and advancement of bot and solver technology/software- the main barrier to access and growth for the online poker industry is market segregation/closure AINEC

so, this "impending demise" you speak of....you need to show your work, or i'm just going to continue to scoff at your hyperbolic assertions

i see this got moved to a new thread, i probably won't stay involved in this discussion

I am a poker player that makes a living playing a strategy game. "Exploiting people's weaknesses" is a really bull**** way of putting it.

lol @ "out-competing you in life" -- wtf are you talking about? I'm not playing the same game as these *******s. I'm not competing with them, I'm wishing that they all went away.

GTFO with this Social Darwinism. Strategy games deserve a place in this world, and they should be protected from destruction. And destruction, fwiw, is exactly the implicit goal of these bot/GTO developers. You yourself mentioned how nobody can play backgammon for real money online. No tears, however, get shed for the people who dedicated a chunk of their life to understanding backgammon and/or depended on it for their living, they just hear "PULL YOURSELF UP BY YOUR BOOTSTRAPS"! by you and the rest of the idiot libertarian/crypto crowd that is everywhere in poker.

look, I also think that games are still fine and I'm usually the one scoffing at THE SKY IS FALLING people, but it's a straight fact that online poker games in general ARE indeed getting harder (to the inevitable point at which they'll become unplayable) and you not admitting otherwise is just you being willfully obtuse. I'm sure lollive poker is as absurdly soft as it's always been, but I'm not talking about lollive poker.

Just look at backgammon or chess for examples; nobody plays these games for any significant money online, mainly because there are engines for these games that are stronger than any human and they became advanced enough to run in real-time. This will certainly happen to online poker as well.

this is, of course, not what he is saying