1. I don't know what is all of this panic about. Did we know it was coming? Yes! The only surprise is how they managed to make it work with minimum hardware resources, but that's few years difference max, computer chips improve exponentially. It was going to happen anyway, it only happened little bit sooner that we all expected.

2. Rob Yong said he has unbeatable plan to exterminate bots completely. Well, now he'll have his chance, lets' watch and see.

3. In the meanwhile, I have to say it again - there will always be live poker. Personally, I learned to play online, played live only a few times, but it was f...in refreshing experience. I got to play 5 times higher stakes with weaker competition then my regular online game. So, multitabling effect eliminated!
As a customer I get unlimited free coffee whole night, garage parking ticket for a day, get to talk to interesting different kinds of people, from some kind of mobsters, retired wealthy guy, drunk kids sitting to play with you at 4am after night out and many others... And what an experience it is when someone shows you J5o BB defend against UTG raise!

There is also possibility that improvements in AI will enable poker sites to detect bots in their sites. Nobody knows if this is going to be the case but hopefully.

I think HUD's should also be disabled. There is nothing wrong with studying the game off the table but when in the gameplay you should be on your own without any supporting software.

I mean if you understood anything about AI you'd know

No I'm sure everyone else in the world is a starfish iq simpleton in the presence of great mastermind xeno who instantly knows the GTO solution, as well as how computers play it and how high stakes players play it and how far away each is from equilibrium.

You only don't play high stakes cash because you have better things to do, I guess, like trying to prove how great you are 24/7 to randoms on tryhard forums.

I do play high stakes cash though

also I don't instantly know the GTO solutions, I ran it? ....... obviously?

lol

it's also not even about knowing the solution but about knowing how it gets to said solution

Not a chance.

not a chance of what?

I'm interested what you are referring to. Could you elaborate please?

"even if a Nash equilibrium could be computed efficiently in a game with more than two players, it is not clear that playing such an equilibrium strategy would be wise. If each player in such a game independently computes and plays a Nash equilibrium, the list of strategies that they play (one strategy per player) may not be a Nash equilibrium and players might have an incentive to deviate to a different strategy."

so yea, given that pluribus does not try to find GTO solutions and considering that it is unclear that such an unexploitable strategy even exists for games with more than two players, it is kinda obvious that pluribus could in theory be exploited.

Don't mind.

Just struggling microstakes players getting upset.

that's true, I was refering to the QJ hand which was HU postflop though, although it's not impossible that it RNG'd an extremely low % of river check and then played accordingly against the medium bet size

I was just thinking if we compare with PIO's outputs then it's very similar except it's been using extremely low fequency plays (sub 1%) and this usually happens when you don't let AI play long enough against itself. I think if they had kept playing then eventually pluribus would always bet river with QJ instead of check raising if they'd encounter this exact same spot

Does anyone else think that calling 65s HJ vs LJ 2.25x open (first hand in the video) is probably a losing play? Maybe a low frequency play (or one that should converge on zero% but hasn't yet)?

Would be more interested in seeing some of its preflop ranges (as these are necessarily related to the fact it's 6max rather than HU) and maybe some hands where it donks multiway. Who cares when it goes HU to the flop, pips someone and "wins much more than a human player" taking what is probably a low % line...

Was thinking same thing about the 65s cc in HJ. Also wished LJ had A4dd in that hand and snapped, but then I guess the hand wouldn't have been in the video.

Mostly agree with your 2nd point too, though I do think the QJ/QT hand is at least mildly instructive.

I could be wrong but I think this one would not go to 0%, think it would be low freq like 67s 78s 89s but it would be there at a low freq for board coverage purposes? then again I have no idea about this one, but fair to think it may be going to 0%, wouldn't be a complete surprise if it was so

Does anyone have link to full results for each pro in this study?

Why do you think Katya is a woman? Just because of the sn?

This bot vs Stu Ungar. Thoughts?

Hey guys- is anyone with a bit of coding knowledge able to turn the hand histories from Pluribus into something intelligible by poker software? I can do it for individual hands but it takes forever. All 10,000 hands played in the 5 humans and 1 AI challenge are available at the link here. I've managed to combine them into one file and format the hands to make it a bit easier to read, but I don't know how to do more than that. I think they'd be very useful to learn from but I can't do as much as I'd like to with the format that they're given in.https://science.sciencemag.org/conte...ce.aay2400.DC1

the real question I am asking here is what what pluribus have down in the the QJ hand if the villain shoved in that spot - I assume call but I've been wondering how it would respond to real strength after it assumes it has the winning hand and really tries to get max value

I've been learning regex for notepad++ specifically to try to be able to reformat all the hands so PokerTracker can read them. I think I will be able to do it but it's quite tricky. I'll share resulting hand history files here if I can get to that point in the next few days.

Someone good in cash games please explain the logic behind the A7s hand (especially the pf part)

Snowie flats in that exact spot at a mixed frequency. It does’t flat any other suited connectors in that spot, literally only 65s which I assume it does for board coverage.

The thing that makes me skeptical is the vast difference between Pluribus which uses a $500 CPU and Libratus from 2017 which used a supercomputer and was tested on 120.000 hands. Pluribus is 6-handed and Libratus heads-up. That is some jump in performance!
Assuming that Pluribus' achievements are real, there is today no way of translating its knowledge into strategies that benefit players anyway. Can we learn from its play? Maybe, but I suspect some of the success it had comes from its vastly different strategies which confuses human players. Change your strategy from the norm and you will initially win until the humans figure you out and your winrate drops.

An important thing that is lacking in this article is the word 'solve' which is key in this field, the researchers doesn't claim poker is solved in the way checkers is solved, they just improved a poker-bot. So, no need to look for a more complex game in the near future in my humble opinion.