I don't necessarily agree. If it has an approximation of GTO strategy and it recognizes that the players are deviating from that strategy in consistent ways it should be fairly simple to make some adjustments.

For example, its initial strategies may be based on assumptions about pre-flop raise sizing. When the humans all start raising 6x preflop it can calculate counterstrategies to that overnight that it likely can't in real time.

no one is calling jungleman a fish except you. and isildur1 is the endboss at HUNLHE, maybe not so much other formats.

id bet my money on jungleman vs any of these 4 players.

The AI had poker player assistance at the end of which session to help on the adjustments.

What impressed me the most, is realising that if I play "very good", I could have a gorgeous edge of 15 bb/100 on some of the top 10 HU players of the world. This just means there is still a lot of edge in poker to be exploited. Maybe I should stop watching videos of RIO (they probably will be losers), and start working with Libratus to try to implement its 10x 3bets, 6x overbets on the turn, etc.

It's only a valid win when the ai can beat the brains without human assistance during the competition or has some kind of an auto correct feature that cleans up its play at the end of each day's play (designed before the competition started).

This was just brains vs brains + ai.

When Kasparov lost to the AI back in 1997, the machine had chess players assistance before and in during the match, just like any other programmer.

During one particular game Kasparov was even claiming the AI had Chess player assistance DURING THE GAME just because the machine made a very human move during one game that Kasparov lost. You can google it.

Because of the suspicion Kasparov asked for the logs of the game but IBM never complied.

isi is better at huplo than he is in nlhe nowadays. On top of that, of all the variants in 8-game or mixed holdem´s(NLHE&LHE) is his weakest. Gladly get corrected on this though.

Your levels are the sickest dude

Heh, that is an intersting perspective.

I agree, I also think it shows similar things. We are far away from a perfect solution, and knowing that should motivate people to work hard to get to Libratus level of playing

Scientifically, that's a statistical wow.

Idk if this is serious but the whole point is that it's impossible for any human to balance those massive overbets and you would just be owning yourself no matter how hard you tried.

Another point of reference:

Although Prof. Sandholm said that Libratus never played Claudico, it did play Tartanian 8, which was, according to them, stronger than Claudico. He declined to divulge details until after the challenge, but on the last day as the challenge was winding down on the Twitch stream he said Libratus beat Tartanian 8 by ~10.5 bb/100, so it beat a stronger bot than Claudico with a greater winrate than the first group of players had against Claudico.

Just the opposite, it is exactly not the point.

Maybe humans will be able to beat this version of the bot sometime. It's not crazy to suggest these guys would have a 14bb/100 edge over the top 2012 regs, so maybe 2022 regs will be able to beat this bot. On the one hand, it's harder to improve since the lowest hanging fruit is gone, on the other hand, they can use vastly more efficient study tools (eg the bots themselves).

I also find it interesting that many of the bright flames in poker history caught the poker world's attention with their very unconventional bet sizing, notably over betting. (Prahlad, Tom, Viktor, Dan C. all come to mind)

Not meant as a troll and I realize that several of these flames have burned out (and i realize their strategy was far less complex obv) but it's neat to think in a world of 2.6x open, 3 cbet sizings of 27%/54%/73% pot at various frequencies, 2 turn sizings of 63%/78% pot at various frequencies, etc., that Libratus comes in and be like "oh, we can bet any amount we want? Neat." *proceeds to 10x raise*

What if Libratus found that for example 45s is 200b 3b-(massive overbet)shove after preopen and gives best +EV?
Humans cannot never see it. No isildur no Durr
Another intresting point was strong opening 4x, weak opening 3x. It doesnt care if we know it!


I think algos were quite simple and basic when aa vs xx we 3 bet flop were 222. Im sure Libratus found at least million possible ways in database how the hand should continue. And if there wasnt enough data it made some algos abbrevations
Mayby data says that we must 3x over bet aa in 222 flop?**** that science is so beatiful . I hope they give some nice repsort about the game

Funny, that mankind think that optimal pre raise should be 2.4...

https://www.cmu.edu/news/stories/arc...oker-pros.html

The event was surrounded by speculation about how Libratus was able to improve day to day during the competition. It turns out it was the pros themselves who taught Libratus about its weaknesses.

“After play ended each day, a meta-algorithm analyzed what holes the pros had identified and exploited in Libratus’ strategy,” Sandholm said. “It then prioritized the holes and algorithmically patched the top three using the supercomputer each night. This is very different than how learning has been used in the past in poker. Typically researchers develop algorithms that try to exploit the opponent’s weaknesses. In contrast, here the daily improvement is about algorithmically fixing holes in our own strategy.”

Sandholm also said that Libratus’ end-game strategy, which was computed live with the Bridges computer for each hand, was a major advance.

“The end-game solver has a perfect analysis of the cards,” he said.

It was able to update its strategy for each hand in a way that ensured any late changes would only improve the strategy. Over the course of the competition, the pros responded by making more aggressive moves early in the hand, no doubt to avoid playing in the deep waters of the endgame where the AI had an advantage, he added.

Sandholm will be sharing all of Libratus’ secrets now that the competition is over, beginning with invited talks at the Association for the Advancement of Artificial Intelligence meeting Feb. 4-9 in San Francisco and in submissions to peer-reviewed scientific conferences and journals.

Throughout the competition, Libratus recruited the raw power of approximately 600 of Bridges’ 846 compute nodes. Bridges total speed is 1.35 petaflops, about 7,250 times as fast as a high-end laptop and its memory is 274 Terabytes, about 17,500 as much as you’d get in that laptop. This computing power gave Libratus the ability to play four of the best Texas Hold’em players in the world at once and beat them.

Look at who retweeted..

This computing power gave Libratus the ability to play four of the best Texas Hold’em players in the world at once and beat them.

Is this a true statement ?

That article is terrible

Are my betsize shortcut buttons not standard?

#UnlimitedHoldem

Anyobne has good scientific article. All i have googled are almost copied emotional ****. I hope professors give some hinsight

OTB is libratus!