There's already prompters and bots out there that are close enough to GTO to beat almost any game. imo.

It will be available or some variation of it, just a question of time. The author of piosolver was a poster in ycombinator iirc. Monker already uses an higher abstraction to solve for multiplayer games. And they want us to think that having hand histories or pokertrackers are the bad thing for the future of online poker. Welcome to the anonymous gto bots reality everyone!

Pretty disappointing to see Linus participating for such a low amount.

Also they could at least invited some woman players, like Katya18 for example.

Or maybe the battle of the bots, and play against 5 VM's of oborra.

Or some russian players, they have a long history of playing against AI in many types of games. Maybe against zaruba, he plays better than the bots.

GG skynet on the way.

based on a seven hand sample size bot is a wild maniac and also never folds any piece to anything ever.

I remember that bot a lot from about 2005 and not so much after that.

the QT vs QJ shown hand is very cherry picking to me since there aren't many worse Q's that are gonna call, heck QT might have folded there.

I have an issue with this feat.

From the AMA with Noam Brown (one of the authors of the bot) when asked about the bots performance in multiway pots :
I believe performance in multiway pots should be the key thing to test in this challenge. Libratus already established that humans are no longer a match vs AI in heads up formats. While proving so again this time with varying pre-flop ranges is of course impressive the really interesting question is whether the AI can come up with a better post-flop strategy multiway than top humans. It is conceivable that what happened here is the bot murdered in the heads up pots, but did relatively poorly in multiway pots.

I also don't understand why the author says 'It's not really possible to measure the bot's performance just in specific situations'. Why is not possible to compare the results for multiway pots?

Well, tournament results are usually reported in big blinds per 100 hands, so it is obviously SNG. Also, Linus is knowns as a best 6max sng tournament player.

It is. It would.

QJ has >93% equity against BTN river betting when I simulated it with standard preflop ranges & overbets possible. This is a hand that looks insane to humans only because humans suck at getting thin value. In equilibrium BTN is almost never checking a better hand on the turn and can't improve on river, and is mostly trying to get value with a 2nd pair type hand with the river bet.

Probably this, yea. People really don't check back hands that are stronger than QJ on the turn there sooo that basically makes QJ the nuts on the river I guess. Which the bot figured out because bots.

I think people with QJ there would be like "well he's not calling a river c/r with worse so I snapcall his bet" while the bot is like "well he never has QJ beat here so I 7x the pot for value"

It's possible. We had this discussion before so I won't come back to it. Just because their implementation is such that it's hard to measure doesn't mean "it's not really possible".
If we ever release multiway solver there will be a way to measure exploitability of the solutions even if it takes a lot of time to get the estimation to good precision.

I don't know, there were winning bots in 2009 playing Party 6max games. It's 8 years later now so you can imagine the advancements since then. Another group of programmers kinda getting there (I mean, taking 20 seconds per decision is still not usable online so they would need to optimize things) doesn't change the prospects for online poker in any meaningful way.

10
lol
yeah, major world-changing advancements in AI over the past decade including decimating the top pros with their PIOsolvers, and finessing .5bb/100 out of the BEST pro we have, in HUNL and now 6 max, doesn't change the prospects for poker in any meaningful way, at all

so much denial ITT

IT'S O.V.E.R.

Just pay linus to play it for a bunch of hands, we need to know if it’s over, I’m skeptical of their variance reduction techniques

A facebook funded AI team that can make a bot which beats Linus probably know what they are doing in this regard. You can read more about the technique here: https://arxiv.org/pdf/1612.06915.pdf

fwiw I don't know **** about it and I don't understand it. This quote from the paper about the bot is interesting: " Due to the extremely high variance in no-limit poker and the impossibility of applying AIVAT to human players, the win rate of individual human participants could not be determined with statistical significance."

So seems like they only apply it to the bot

edit: Another interesting chunk from the paper:

" Pluribus confirms the conventional human wisdom that limping (calling the “big blind” rather than folding or raising) is suboptimal for any player except the “small blind” player who already has half the big blind in the pot by the rules, and thus has to invest only half as much as the other players to call. While Pluribus initially experimented with limping when computing its blueprint strategy offline through self play, it gradually discarded this action from its strategy as self play continued. However, Pluribus disagrees with the folk wisdom that “donk betting” (starting a round by betting when one ended the previous betting round with a call) is a mistake; Pluribus does this far more often than professional humans do."

Fwiw I'm pretty sure that donk-betting isn't at all considered "folk wisdom", I think it's pretty well known that donkbetting is part of a GTO strategy. It's just considered too hard to balance well for humans so it's removed all together from most human strats to simplify the game tree. Only talking about the flop here ofc, because I think people donk bet turns and rivers

OBORRA was likely banned for some kind of realtime GTO advisor years ago, and he was regarded as one of the best players. Is this really that big of a revelation? And some of the best players today have also been accused of cheating. Pluribus got the mainstream attention, but OBORRA got the money.

"And some of the best players today have also been accused of cheating."

Its true for almost any sports best players.
Edges are so small people risk steroids. But there they could get 4 years penalties and even lifetime ban. Which means cheaters are under constant pressure can lost their income, lose their image and so on. Despite the hars penalties almost everyone does it in top10.

In online poker someone cheats, gets banned and money confiscated. But cheaters dont keep big amounts of $ on poker accounts anyway. 1 room bans the cheater, but he can buy someones ID and utility bill for 150-200$, sometimes even less, changing up mouse patterns etc. Rooms sharing database of cheaters wont cant stop this until we have some system to combat this.

Also couple reasons why stealing/ cheating can be bigger in poker than other conventional sports, sitting in front of computers, not seeing / knowing other humans.

https://www.youtube.com/watch?v=XBmJay_qdNc

When OOP dont have donking range on certain runouts (especially turns and rivers) loses lots of EV. Also depends a lot on stack sizes, Im more of an expert of 0-50bb. Just compare couple different board texture: A) no donks at all B) can donk every street.

The variance-reduction techniques are sound.

I created the first variance-reduction system for poker, called DIVAT (aka Darse's Ignorant Value Assessment Tool). [see https://webdocs.cs.ualberta.ca/~dars...ivat-icgaj.pdf, or Chapter 5 of my PhD thesis, https://poker.cs.ualberta.ca/publica...lings.phd.pdf] Morgan Kan and i built implementations of DIVAT, and used the tool in developing programs that won the 2007 AAAI computer poker competition. [see Morgan's MSc thesis for details]

Martin Zinkevich proved that DIVAT is a statistically unbiased estimator of the actual EV. The baseline strategy used for comparison of each decision was hand-crafted by me (based on my understanding of a game-theoretic strategy), but the differentials in expected values are unbiased, so it is a completely fair assessment method. [Loose analogy: Measure distances using a crooked stick. As long as you use the same stick, you can tell that distance A is bigger than distance B (and roughly by how much).]

Michael Bowling wanted a fully automated method for variance-reduction, since no one else on our team had the domain knowledge to hand-craft a baseline strategy for other games. All of the subsequent methods (MIVAT, etc.) were weaker than DIVAT in terms of variance reduction, but they had the significant advantage of being purely algorithmic. AIVAT is the latest iteration in that line.


The Carnegie Mellon result for 6-max is laudable, if it is real, but i have learned to be skeptical of all their claims. For years they have been making over-inflated claims, without giving proper academic credit to the University of Alberta Computer Poker Research Group (CPRG), which has been years ahead of them in every area of poker research. [I even remember Tuomas making grandiose claims in a talk at AAAI, about an hour before their program was crushed and humiliated (again) in the computer poker competition. For the record: the first poker program to beat top humans in heads-up Limit Hold'em was from the UofA. Then Mike, Neil, and Michael epsilon-solved the game. The first poker program to beat top humans in heads-up No Limit Hold'em was from the UofA. None of the press coverage bought by C-M to advertize their advances has ever mentioned those prior results.]

If their program is better than all human players in 6-max NLH, then i congratulate them (but sorry, i'll need to see some actual proof first). To be honest, i expected that result to come much sooner (like maybe 5 years after i left the research in 2007). [Perhaps the advances would come sooner if the researchers took a few weeks to actually learn the basics of the game... it can really help in gaining key insights. ;-)]

zaruba with the help of his own AI or without it?

how long till they sit down next to you in the casino in a realistic looking human suit

i actually think this is a good thing, we need to mainstream awareness that bots are real and can crush the best out there for people to care enough to not play on sites that don't fully police them

once players start voting with their wallets, the sites will adapt and put in measures to truly ban them effectively, until then the bots will slowly bleed online dry until it's no longer viable

Very interesting stuff, thanks for the info.

I have no doubt they know what they’re doing, just wary of them exaggerating how effective it is to make the bot look better.
Guess the skepticism is because of that Claudico statistical tie thing lol.

Lol, 6max poker is about your overall performance. Sample size is absurd in multiway pots since its so few of the % of pots in a non ante game. Also preflop Ranges would make a drastic difference. One player or bot could have higher winrate in mw pots if he decide to give up overall ev to play more strong hands preflop passively.

I have done a lot of db studying and agree it doesnt make sense

So nobody cares about (mostly) MTT players (of variable skill) playing 100bb cash, no rake and fixed stacks each hand?

This is like putting NL regs in PLO vs PLO bot and bragging that the bot managed to win (yes, this comparison is a hyperbole).

If he's pushing you around and your stack starts to wither/
Check back the nut flush he'll check shove the river

this bot isnt gto btw right and wouldnt striving for gto begin to have less of a return the closer you get ? , its max solution? also 10k hands split up between that many different players with a style that used high variance isnt enough to determine if its even winning?

also would someone like linus be able to exploit its baseline strategy after he had a large enough sample to pour over? i believe it said it categorizes players from one of four baseline strats and then doesnt deviate from that baseline after that so to me that sounds exploitable but i dont know enough to say.