Brains vs. AI poker rematch coming to Rivers Casino - Page 10 - Poker News

Hi, I created Claudico (not Libratus). I wanted to say first that I am very impressed with what the team has done for Libratus. I agree that if they win, it is unlikely that they will do another challenge (why risk a "loss" after they have already "won"?)

I would like to run another challenge with my team against doug, sauce, ike, and jungleman in approximately one year after this, assuming we can work out a payment scheme both sides agree to and I can find a way to raise the money. My bot would not tank nearly as long as Libratus and would likely not use a supercomputer. I'd like to do this at Seminole HR, and I think it could be ready one year from now.

Should the humans lose this contest, it hope it becomes like when the US only sent college basketball players to the Olympics until they lost and then the next Olympics sent the "Dream Team" that slaughtered everyone.

I would pay to watch a bot vs a "Dream Team" of humans (any four of Doug, Jungle, Ike, Sauce, OTB) in a HUNL challenge and I believe others would too (but please make it easier to follow the action on the table). Hope it happens.

Quote

01-13-2017 , 11:10 AM

#228

watergun7

veteran

Join Date: Jun 2012 Posts: 2,796

Idk to what extent the bot solves turn+river on the fly. But if you hook up pio on every turn with full solving with 4 sizes and reasonable starting ranges (i.e. pre and flop strategies) I don't think any human can beat that...

If it was known in advance that the bot would take ages on every turn rather than every river like last year I'm sure the line for the match would have started very differently.

Quote

01-13-2017 , 11:14 AM

#229

SrslySirius

Animated Character

Join Date: Jun 2010 Posts: 4,357

Quote:

Originally Posted by restorativejustice

Comments during this challenge by the Professor and how he and the University interpreted AI's defeat last year as a "statistical tie" caused me to develop a very healthy skepticism about public science and experts and how both can be just as biased and corrupted by monetary/reputational rewards as the rest of us.

Same, kind of. I was already aware this is a problem, but seeing it play out with a subject I'm intimately familiar with has been disillusioning.

Quote

01-13-2017 , 11:19 AM

#230

mrno1324

veteran

Join Date: Feb 2016 Posts: 3,055

Quote:

Originally Posted by Sam Ganzfried

Probably a stupid question but would you be able/agree to pit your bot against Libratus to see which is better?
No needle, but everyone wants to see the dream team versus the best AI science can produce so even if they beat your bot it won't mean much unless you prove it's better than Libratus.

Quote

01-13-2017 , 12:06 PM

#231

getmeoffcompletely

old hand

Join Date: Aug 2012 Posts: 1,670

So interesting to see humans agonize over spots bot is putting them in.

Quote

01-13-2017 , 12:22 PM

#232

vrael111

veteran

Join Date: Dec 2013 Posts: 2,957

Quote:

Originally Posted by getmeoffcompletely

I'd pay to see Jungleman play a bot.

Hope the AI team insures their computers.

Imagine Cates just smashing keyboards and screens live on webcam in a casino after his 1st bad beat

Quote

01-13-2017 , 12:22 PM

#233

restorativejustice

veteran

Join Date: Aug 2014 Posts: 2,753

Quote:

Originally Posted by getmeoffcompletely

So interesting to see humans agonize over spots bot is putting them in.

A line I have heard more than once from separate human players in this contest is, "I don't know if the computer was value betting or bluffing there."

Quote

01-13-2017 , 01:36 PM

#234

swc123

grinder

Join Date: Jul 2013 Posts: 431

Quote:

Originally Posted by chinagambler

Quite funny that people wanted to bet on the AI at 4:1 and now the AI is a favorite. At 3k hands per day those odds estimate his winrate to be 5bb/100.

What site is that SS from?

Quote

01-13-2017 , 01:38 PM

#235

spewmonk3y

journeyman

Join Date: Mar 2016 Posts: 335

When some1 starts doing 5big blind 3bets OOP at hunl I know where they got that one.

Quote

01-13-2017 , 01:43 PM

#236

samooth

veteran

Join Date: May 2009 Posts: 3,350

Quote:

Originally Posted by swc123

What site is that SS from?

pokersharesdotcom (gl

)

Quote

01-13-2017 , 02:05 PM

#237

watergun7

veteran

Join Date: Jun 2012 Posts: 2,796

Why is jimmy 3betting hands like J6o 92o etc.

Quote

01-13-2017 , 02:30 PM

#238

AshleyC

veteran

Join Date: Sep 2005 Posts: 3,278

Quote:

Originally Posted by watergun7

Why is jimmy 3betting hands like J6o 92o etc.

Better hands I.e J7 are in his call range. He's 3betting his trash hands and value hands, calling in between and folding his very worst hands i.e. 72o.

Quote

01-13-2017 , 02:49 PM

#239

watergun7

veteran

Join Date: Jun 2012 Posts: 2,796

No human is 3betting 92o vs another human. I'm thinking that they thought the bot's fold vs small 3bet is too high but could also be them not taking these games too seriously.

Quote

01-13-2017 , 02:54 PM

#240

Ricky1231

Carpal \'Tunnel

Join Date: Jul 2007 Posts: 6,663

Quote:

Originally Posted by watergun7

Why is jimmy 3betting hands like J6o 92o etc.

Super standard - nothing to see here - move along pls

Quote

01-13-2017 , 03:01 PM

#241

boredoo

journeyman

Join Date: Jun 2013 Posts: 306

Looks like the humans are adjusting. Jason was on stream talking about how as they'll study they will improve far faster than the AI will. I suppose if the humans can pin-point weaknesses in the AI strategy, the AI will have a harder time unlearning its weaknesses than the humans will in patching their leaks and exploiting the AI.

So far today every human is up.

Quote

01-13-2017 , 03:17 PM

#242

restorativejustice

veteran

Join Date: Aug 2014 Posts: 2,753

Quote:

Originally Posted by watergun7

No human is 3betting 92o vs another human. I'm thinking that they thought the bot's fold vs small 3bet is too high but could also be them not taking these games too seriously.

Yes, all the tanking by the computer is a real disadvantage to the humans as it may cause them to make moves just to avoid delays on later streets.

Quote

01-13-2017 , 04:19 PM

#243

fluorescenthippo

Pooh-Bah

Join Date: Apr 2005 Posts: 5,848

Quote:

Originally Posted by restorativejustice

A line I have heard more than once from separate human players in this contest is, "I don't know if the computer was value betting or bluffing there."

There was a hand where bot hand J2s OOP in a 3bet pot vs Les where I was thinking the same thing.

Quote

01-13-2017 , 05:51 PM

#244

vrael111

veteran

Join Date: Dec 2013 Posts: 2,957

Anyone noticed swarm's 3b sizings with different hand strengths? He seems to be splitting the 3b ranges linearely, e.g. small 3b's with bad hands - big 3b's with best value hands.

Seen him do this a bunch times. 3x 3b with Q2o, 3.5x with AJ, 4x+ with AK

Quote

01-13-2017 , 05:58 PM

#245

pokermon!

centurion

Join Date: Jun 2014 Posts: 119

Quote:

Originally Posted by vrael111

This kinda makes me think that's he believes the computer may keep track of 3b frequency and will adjust to that without really differentiating between the 3b bet size.

Quote

01-13-2017 , 06:00 PM

#246

samooth

veteran

Join Date: May 2009 Posts: 3,350

goooo

Quote

01-13-2017 , 06:15 PM

#247

vrael111

veteran

Join Date: Dec 2013 Posts: 2,957

Quote:

Originally Posted by pokermon!

This kinda makes me think that's he believes the computer may keep track of 3b frequency and will adjust to that without really differentiating between the 3b bet size.

Yeah seems to be the logical explanation although it is hard to believe a bot like this would make such blatant mistakes

Quote

01-13-2017 , 06:32 PM

#248

Loctus

Carpal \'Tunnel

Join Date: Nov 2008 Posts: 11,161

Again, the bot is made to work towards a GTO solution to the game. This means it doesn't keep track of ****. It does not care what the human players do at what frequency and it does not adjust to what they do at what frequency.

The entire ordeal of the university, the AI, the everything, would be pointless if the bot did that. The end goal is not to create a bot which beats humans at HU NLHE, the end goal is to solve the game.