Post: #1

Libratus, developed at Carnegie Mellon University, recently beat 4 of the world's top Heads Up No Limit Holdem players in a 20 day challenge.

Libratus used a few different techniques/algorithms to 'learn' the game by playing over a trillion hands against itself over a number of months. The name means 'balanced' in Latin and that is pretty much how the programme plays. Of note are the massive overbets it makes to win small pots, sometimes betting 10x the pot on the river, and bluffing 'intelligently' while using blockers correctly.

It was also able to make adjustments to it's play according to it's opponents styles. Early in the challenge the pros found a weakness in Libratus' play and started exploiting it, however, over a few days Libratus was able to adjust to this exploit and the weakness disappeared.

Post: #2

(02-01-2017 12:29 PM)kavi Wrote:  Libratus, developed at Carnegie Mellon University, recently beat 4 of the world's top Heads Up No Limit Holdem players in a 20 day challenge.

Libratus used a few different techniques/algorithms to 'learn' the game by playing over a trillion hands against itself over a number of months. The name means 'balanced' in Latin and that is pretty much how the programme plays. Of note are the massive overbets it makes to win small pots, sometimes betting 10x the pot on the river, and bluffing 'intelligently' while using blockers correctly.

It was also able to make adjustments to it's play according to it's opponents styles. Early in the challenge the pros found a weakness in Libratus' play and started exploiting it, however, over a few days Libratus was able to adjust to this exploit and the weakness disappeared.

This could destroy the online poker industry. I know bots have been used in the past to beat fish, this will hit them much harder. The sites will try to catch the bots, but I'm sure it will be very hard for them to do. Not even pros will want to play against super computers if the poker sites can't catch them.

Post: #3

Nah. It won't. It just turns into a different game.

Look at chess. The first grandmasters were beaten in early 2000's and now, it's almost impossible for anyone to beat even a moderately strong chess AI. What happened? Nothing. Chess slightly transformed though. Now, most grandmasters use bots to discover new strategies/moves/variations.

Post: #4

(02-01-2017 06:32 PM)tapthatass Wrote:  Nah. It won't. It just turns into a different game.

Look at chess. The first grandmasters were beaten in early 2000's and now, it's almost impossible for anyone to beat even a moderately strong chess AI. What happened? Nothing. Chess slightly transformed though. Now, most grandmasters use bots to discover new strategies/moves/variations.

The amount of money bet on poker is many orders of magnitude higher than chess.

The poker sites depend on a small group of super high intensity (read: high rake rate) pros for a big chunk of their income.

Those pros will not keep playing once they realize the are probably sitting at a 6 handed table with 5 bots who all play better than them.

You can't be the only 'fish' at a high stakes table for long before you run out of money.

Post: #5

(02-01-2017 05:17 PM)ascotpudding Wrote:  

(02-01-2017 12:29 PM)kavi Wrote:  Libratus, developed at Carnegie Mellon University, recently beat 4 of the world's top Heads Up No Limit Holdem players in a 20 day challenge.

Libratus used a few different techniques/algorithms to 'learn' the game by playing over a trillion hands against itself over a number of months. The name means 'balanced' in Latin and that is pretty much how the programme plays. Of note are the massive overbets it makes to win small pots, sometimes betting 10x the pot on the river, and bluffing 'intelligently' while using blockers correctly.

It was also able to make adjustments to it's play according to it's opponents styles. Early in the challenge the pros found a weakness in Libratus' play and started exploiting it, however, over a few days Libratus was able to adjust to this exploit and the weakness disappeared.

This could destroy the online poker industry. I know bots have been used in the past to beat fish, this will hit them much harder. The sites will try to catch the bots, but I'm sure it will be very hard for them to do. Not even pros will want to play against super computers if the poker sites can't catch them.

No it wont, reasons why:
1. Ai beat good players but not best players.
2. It needs supercomputer to run.
3. Its a HU. HU was mathematically solved couple years ago. They wont solve fullring poker any time soon.

Post: #6

1. Everyone is saying no human can beat the computer
2. Yes, for now atleast.
3. HU Limit HE was solved a few years ago but not No Limit HE. Actually it's not about solving, the machine just learns by playing against itself. Solving would mean finding the GTO solution. But you are right that Full ring is much more difficult for a machine.

I don't think this affects online poker atm as there are already bots for that, this bot is clearly an exception in terms of complexity and not something someone can hack up at home.

Post: #7

It kills off poker because making decent money in poker these days has been reduced to bumhunting. Once fish know they are playing at tables full of bots they quit and go to the casino or do sports betting. Explains the rise of fanduel and draftkings. That's where all the 'easy' online poker money went.

Post: #8

I only play Live poker anyway. Online is completely different.

Post: #9

Online poker has been dead for years. The development of this AI is impressive though.

Post: #10

Eventually a light version of AI software will be commercially available. Assuming poker sites are either slow or unable to block the software, bots will multi table over multiple games and stakes to grind out X bb per hour. These bots will run 24/7 and capitalize on weak to moderate players which make up 90% of players online.

Once human are squeezed out, the game will turn into a software battle.

Post: #11

Can't imagine why anyone would play online poker.

1. You're not sure that the casino isn't using a bot against you for the other "human" players that may well be their own server. How would you ever know?
2. You'll also have no way of knowing if a human player unaffiliated with the casino is using some type of bot script run with AI to actually make their moves.
3. Real-life poker is where it is at if you have the skill. Not something easily developed, but can be done.

Post: #12

1. There's no way of knowing for sure but it would be a big risk for an online poker site to do this for obvious financial reasons.

2. Players have been using heads up displays for years already. HUDs aren't full AI but the data processing is a precursor to full AI.

3. I agree. Online poker loses what I consider to be the most vital and enjoyable part of the game; the psychology and to a lesser extend the social aspect.

Post: #13

3. - is incorrect
If you have an edge over someone you want to play as many hands as possible to maximize this.

Post: #14

(02-03-2017 07:29 PM)John Michael Kane Wrote:  Can't imagine why anyone would play online poker.

1. You're not sure that the casino isn't using a bot against you for the other "human" players that may well be their own server. How would you ever know?
2. You'll also have no way of knowing if a human player unaffiliated with the casino is using some type of bot script run with AI to actually make their moves.
3. Real-life poker is where it is at if you have the skill. Not something easily developed, but can be done.

1. Too risky for to little gain. Biggest poker sites make a killing from rake, rigging is counterproductive, why would they want you to lose all the money and not return to play again?

2. They do catch bots from time to time and reimburse the affected players. Again, much harder to do on big poker sites with good software.

3. Live poker has its advantages and its drawbacks. If you are good you will have bigger edge on average, but you will play less hands, so the variance can affect your winnings more, and it is less convenient then playing from home.

Post: #15

It doesn't wonder me at all. NHLE HU was solved quite a while ago.

Post: #16

There was an alleged pro poker player in one of my classes and we were talking about this story. He said that although the "bots" are getting better and better, this story is a little misleading because they computer did not beat the very best poker players in this particular style of poker, just some of the best overall poker players. He contented that had the organizers chosen specific champs of this style of poker it would not gone as well. But he also is aware of the larger point that the machines are getting better and better.