^^

Tuomas Sandholm did, however, state in 2015, before the match, that he was thinking about getting the public to access Claudico as a training tool. After being defeated then, there was no more mentioning of that, and I haven't heard a comment after their win against the humans this time around. If they do decide to publish "Snowie" style tool, it would be great for learning though.

I think the point has been missed here. It's not that Libratus is going to be put to work in the online games. It's more about how the top end of the relevant technology has taken massive leaps in the last few years.

This will inevitably have a trickle-down effect, which is pretty troubling seeing as bots are already taking hundreds of thousands of dollars out of the game. It's already perfectly clear that you don't need anything close to Libratus's level to crush small-midstakes ring games. Further increases in technology in this area might be great for mankind, but pretty terrible for online poker.

And yes, it would be a great learning tool if they developed something like that, but I wouldn't get too excited about it because it will be ****ing horrendous for the games

thats how progress works tho, you can either embrace it and step your own game up or you can be the old man yelling at cloud, getting dragged kicking and screaming into the future while the rest of the world keeps it moving all around you

such is life

Ignorant post

i'd ask you to go on and explain, but i dont want to give aids to everyone itt

there are other threads for bummy regs to rage about coaching sites and training tools

This is false. Chess is far more complex than poker. There have been threads on this topic on this forum in the past. I predicted years ago (with an old screenname I don't have access to) that computers would easily beat all the top human players in poker.

There would be no advantage for "human + Libratus 1" vs. Libratus 2 in your example, because if libratus is playing GTO poker, there is no improvement that a human could provide. Take a look at AlphaGo, eg. The computer would give the probability that a human would make any particular move. One of the startling moves it played, it estimated that a human would play 1 in 10,000 times. So the computer is considering everything that a human would consider, and then makes its decision.
This is false. Kasparov did not have any access to a computer. Deep Blue not only had every one of Kasparov's games stored in its database, but also millions of games played by grandmasters; probably all significant games that had ever been played. That is why Kasparov's victory in 1996 was so extraordinary. Not only did Deep Blue have a huge database, but could calculate millions of moves ahead, while Kasparov only relied on his mind.

Between 1996 and 1997 when the computer won, IBM hired a chess grandmaster to work full-time with them. This was a huge factor in Deep Blue becoming much stronger. Besides that, all of the database is from prior human games, so there was significant human-derived chess knowledge built into its system. As far as I know, no poker pro was needed to assist the Libratus team.


But this match seemed very biased in favor of Libratus. Using Doyle's game rules (stacks reset after each hand) is not like playing a cash game and not like playing in a tournament either. So the poker pros are not accustomed to it. The strategy is obviously very different, and much more complex, to play tournament style vs. Doyle's rules.
Pots were split when both sides were all-in, depending on the probability of each side winning. This is not a standard poker rule anywhere. So basically they have "invented" a very specific form of poker that their computer is designed to play, and are subjecting the humans to it. It should have been the other way around. Knowing more about this match, I don't consider it as significant an achievement as was touted. It's a bit surprising to me that the players not only agreed to these rules, but also the number of hands and hours that they are playing. The computer doesn't get tired or go on tilt, but it's a huge human factor.

Are there any computers that have demonstrated playing high-level poker with standard rules?

Your post is littered with errors and while I'm sure your insight into the chess side of things adds something to this thread, if you can't see why the strategy required for Hold'em with all in equity chops is identical to Hold'em letting the board run out (for a large enough sample of hands), then you're probably not qualified to talk about the poker side of things.

Before more people pm me: The screenshot hand histories don't show up properly on 2p2 anymore but they are still there, simply open the link and get rid of the .jpg at the end.

I don't have any hand histories other than the ones posted in here to share, unfortunately. Besides CMU, only the Empire (Doug Polk & friends) has access to the hand histories. This was set up this way in order to 'protect the players', but ofc the opportunity to monetize this exclusivity wasn't left out.

Becoming a programmer who writes better code is not "stepping up your game" its playing an entirely different game

Prof. Sandholm and Mr. Brown are doing a reddit AMA on 12/18.

https://www.reddit.com/r/MachineLear...fessor_tuomas/

"Did you beat them or was it a statistical tie?"

Got em, upboated

Maybe now that the players involved in both challenges are basically retired they will release some hand histories. Always thought it was a scummy move to keep the AI play behind closed doors.

Here's the newly published paper revealing details of the bot:

http://science.sciencemag.org/conten....aao1733?rss=1

AMA is live on reddit:

https://www.reddit.com/r/MachineLear...fessor_tuomas/

If someone can tldr any interesting stuff in layman language that'd be great

Why would that be scummy?
Isn't Doug selling these HHs on Upswing? Pretty sure releasing some HHs is against his interest.

+1 to mrno's request.

TL,DR:
Next step Settlers of Catan.
Libratus won't crush 6max because filthy humans are colluding against bots.
He will challenge anyone who thinks he has a gambling problem HU4ROLLZ.

Excellent! I've been waiting for this paper for a while.

I didn't read this whole thread but heard about the new study where the AI beat the humans. Didn't realize it was only heads up.

Don't know if this is any better - or 100% accurate - but it was my attempt to write up the AI approach in simpler terms...

First up, the duo set up a simple version of the game – an ‘abstraction’ – and part of the program then ‘computes game-theoretic strategies’ for this smaller version. This solution becomes the ‘blueprint’ for later stages of the AI development – ‘an approximation for how to play in the more numerous later parts of the game’.

Secondly, Libratus’ 2nd module refines this blueprint strategy and solves it in ‘real-time’ - a necessary part of competitive play in almost any game, poker of course included. The advantages of their new approach means that ‘Whenever the opponent makes a move that is not in the abstraction’ - the simpler version in part 1 - a ‘subgame is solved with that action included. We call this nested subgame solving.’ Basically refining the process and solution as it goes along.

Part 3 of Libratus’ innovative and now-proven technique is called ‘the self-improver’ by the computer science geniuses and is designed to ‘enhances the blueprint strategy’.

The simpler version could be imagined as having far more intricate ‘branches’ to become the full game, and the self-improver module ‘computes a game-theoretic strategy for those branches’.

Because it would become way too complex to solve all of these possibilities, as seen by the numbers given above in the Match Poker post, the authors of the paper on Libratus decided to ‘tame this complexity’, by using ‘the opponents’ actual moves to suggest where in the game tree such filling is worthwhile’. Not every branch is useful in practice, so the AI works on what it faces in reality.

Q:
Libratus obviously needs a supercomputer to run at the moment, do you think that it’s possible to make it efficient enough to run on regular computers or servers?

NoamBrown:
Yes I absolutely think it's possible to make a slightly weaker version that can run on regular computers or servers. I also think that as the algorithms improve, less and less powerful hardware will be needed to achieve the same performance. I think we'll see this stuff running on smartphones within 5 years.

Lets see if we're still around in 5 years

The end of online HU in 5 years then ?

fwiw, the published paper was an excellent read for poker players.

Online poker will be done in 5 years. It will still be played but its not going to a profitable game.

Early this year we got the solvers that can solve multiway pots. Currently you cant run the solvers that fast in real time but 5 years from now computing power will be way stronger and the first generation solvers we have now will be way better. As these guys are saying you'll be able to run libratus on your smart phone.

Even the fishy online poker players will have their smart phone app running the future solvers in real time. The edges over players will be so small that beating rake will be nearly impossible. The fun factor of the game will be gone. They'll obv be some hardcore guys who love poker no matter what who'll play but its not going to be a game where any new player ever wants to play.

Disagree with this for a variety of reasons, there is no reason that crushers wouldn't be audited and forced to film themselves playing to show they're not using software.

Furthermore what's stopping the biggest websites from coming together and purchasing these AIs not allowing them for public consumption.

Finally, the same problem could be said to exist in chess where you can run Stockfish and beat the best players in the world (especially online where there aren't as many GMs (highest rank)) eeezzz, there is software to detect if you're playing too perfectly though and you'll quickly get ban hammered.

Where poker is going that websites will likely force you to stream it so they can review if you cheated, built in calculations to how perfectly you play and if you get too close to solvers, you would need to explain yourself or be banned.

And if India and China ever come online and inject 10+ million players into the field, that won't be a problem at all.

Hoping the poker sites are going to band together and payoff all the programmers to not release this type of software is wishful thinking. That would just lead programmers too spending more time to develop better software and bots to make the pokersites pay them more or threaten to release it public. Its just not going to be worth the resources you're discussing.

The poker sites cant stop all bots and cheating going on now. Players have to catch this stuff alot of the time. Its not going to magically get better in the future when the tech gets better for those who cheat. All the top end high stakes online pros have spent alot of time with solvers already. If theirs a solver app on the phone that could be used by everyone in real time the strategy used would resemble a close enough game to a decent portion of the mid-high stakes pros already.

Currently pio is confusing for a noob, and you'd crusher most games harder playing a exploitable strategy. The solvers in 3-5 years will look way different and work much much faster then they do now. Who knows what type of features they'll have. I could imagine them being much more noob friendly and cheaper priced, taking your cards on the flop in real time through a picture on your smart phone, using your hud stats to anticipate a near optima; strategy that exploits weaknesses all in real time, even randomizing the decision so you dont have to use a randomizer.Worried about getting caught cheating? Change the bet size off randomly, make mistakes here and there etc. People who want to cheat will learn to cheat better and better.

It sounds sci fi but the way computers improve drastically with time its possible. Years ago nobody would of thought a bot would conquer heads up. Now its happened. Look at how poker software keeps improving as time goes on. If you talked about pio around blackfriday it would sound like scifi.

Yeah the China and India going to cause another boom thing. Been hearing about that since black Friday. China isnt the most internet friendly country they also have tons of gambling laws. They do have some unregulated Chinese apps you can play on now but many say there rife with cheating. Its not going to be 10mill Chinese playing with you on stars in a big public atmosphere.

Live poker isnt going to go anywhere anytime soon. You could easily just ban phones at the table in that setting. A future game similar to no limit online could also work with some diff rules that would make it harder to beat for bots. As the developers brought up in that AMA a variant of poker where players trade cards could be a good way to slow down the AI and super software takeover.