Heads Up Hold'em Solved? - Poker News - News, Views and Gossip

Two Plus Two Forums Poker News & Discussion News, Views, and Gossip

Heads Up Hold'em Solved?

Post Reply Subscribe

...

Page 1 of 12

1 2 3 4 5 6 11

Page 1 of 12

1 2 3 4 5 6 11

10-09-2013 , 08:22 AM

Ziki6368

grinder

Join Date: Jan 2010 Posts: 416

Is 10bb 6 max satties "solved" ?

Quote

10-09-2013 , 09:02 AM

maton808

old hand

Join Date: Jun 2012 Posts: 1,230

Quote:

Originally Posted by Ziki6368

Is 10bb 6 max satties "solved" ?

Yes.

Don't think he's planning on telling anyone else how though.

[IMG]http://s13.************/477daou5z/jorj.jpg[/IMG]

Quote

10-09-2013 , 12:01 PM

ohsosick

newbie

Join Date: Oct 2013 Posts: 45

Quote:

Originally Posted by Ziki6368

Is 10bb 6 max satties "solved" ?

HU holdem limit was totaly solved this year .
HU NL 200deep is not totaly solved, but the best bot beats the second best bot with an edge of (only) 0.3bb/100 over 1 million hand.
(i will post the sick HM graph)
So i highy doubt that a humain can beat them.

6max is spared for the moment because it required more computational resource

Quote

10-09-2013 , 12:03 PM

DougL

Too helpful for this post

Join Date: Sep 2002 Posts: 21,810

Link for each of these?

Also to all of the people throwing around casual observations/accusations of botting for Doug and Ben, it seems to me like you should present some evidence to go with the random forum typing. Like or dislike who you want. These guys are professionals. The challenge is good publicity for poker. Randomly saying that they are horrible cheats is wrong.

Last edited by DougL; 10-09-2013 at 12:09 PM.

Quote

10-09-2013 , 12:17 PM

MastaAces

Carpal \'Tunnel

Join Date: May 2007 Posts: 10,008

yeah, i'd like to see some proof of a bot beating ike or some other hu nl beast playing 200bb deep (or even 100bb). definitely want to see the graph either way.

Quote

10-09-2013 , 12:29 PM

mperich

Pooh-Bah

Join Date: Nov 2004 Posts: 4,212

Ohsosick seems pretty half informed/wrong.

Quote

10-09-2013 , 12:39 PM

DougL

Too helpful for this post

Join Date: Sep 2002 Posts: 21,810

Quote:

Originally Posted by ohsosick

I'll post links that evening, I'm at work.

Definitely looking forward to that and plan on reading.

Quote:

I'm sure that sauce123 is not a cheater I have so much respect for him.
If he was cheating, he would never have agreed to have his name published on the annualpokercompetition site.

Sorry not to quote the people who were. I put two different things in one post and see how it could look like I was saying you were one the people of posting the accusations. I understood you weren't and again apologize for being unclear. At the time it was "hey, new paragraph, new subject". But, it doesn't really read that way.

Agree with you that there is not reason to think that either of these guys are involved in bad stuff.

Quote

10-09-2013 , 12:44 PM

druidfluid

adept

Join Date: Mar 2006 Posts: 992

Quote:

Originally Posted by ohsosick

How does it follow that since the best bots are evenly matched, they are likely better than the best humans? Seems almost irrelevant to me how they fair against each other as they are likely using similar systems and rules. It's not surprising (to me) that the two best are very evenly matched. I'd be surprised if they weren't. I'm fairly ignorant on the subject though so maybe I'm missing something.

Quote

10-09-2013 , 12:50 PM

Next Level

grinder

Join Date: Sep 2013 Posts: 644

Quote:

Originally Posted by druidfluid

I was thinking this exact same thing.Bot 1 having a .3 edge on Bot 2 tells absolutely nothing about what their BB/100 would be vs a human. Ide highly highly doubt if there were bots that could beat a good HU player at 200+bb.

When two people (bots) play someone will always have an edge, so thats not shocking or anything.

Quote

10-09-2013 , 12:53 PM

#10

DougL

Too helpful for this post

Join Date: Sep 2002 Posts: 21,810

I'm also dubious that HU LHE was totally solved. I'd believe that even the best humans can't beat some bot(s), perhaps with the bot given unlimited time or some other requirement to assist it. An actual final solution of GTO for a real poker game, is a much bigger claim. However, we'll see this evening. I've been wrong before.

Quote

10-09-2013 , 12:56 PM

#11

gregorio

Carpal \'Tunnel

Join Date: Jan 2007 Posts: 39,313

Quote:

Originally Posted by ohsosick

I highly doubt this is true

Quote

10-09-2013 , 01:01 PM

#12

ohsosick

newbie

Join Date: Oct 2013 Posts: 45

Quote:

Originally Posted by DougL

Definitely looking forward to that and plan on reading.

the link for the limit HE solved this year :

http://www.computerpokercompetition....owall=&start=1

Quote:

Team Members: Alexander Lee
Affiliation: Independent
Location:
Technique:
The bot was built using proprietary universal game theory methods applied to poker. We complete Fixed Limit Hold’em game tree search without approximation. Original AI utilizes own database of about 3TB and to comply with completion format our team provided special simplified version of Neo - Neopokerbot_FL2V.

Quote

10-09-2013 , 01:10 PM

#13

Next Level

grinder

Join Date: Sep 2013 Posts: 644

Mods, confirm the link is clean and im deff interested in looking

Quote

10-09-2013 , 01:14 PM

#14

ohsosick

newbie

Join Date: Oct 2013 Posts: 45

The graph beetween best no limit hu bot (200deep @200NL) and the second :

This variant is not solved, but the nash approximation used by the bots are so close to the real nash that the winner can only won 670bb on 102KH :
0.33BB/100
Std Dev= 131bb/100

Quote

10-09-2013 , 01:16 PM

#15

ohsosick

newbie

Join Date: Oct 2013 Posts: 45

Quote:

Originally Posted by Next Level

The edge is thin because their game are close to the GTO

Quote

10-09-2013 , 01:18 PM

#16

tultfill

veteran

Join Date: Jan 2009 Posts: 2,926

Quote:

Originally Posted by ohsosick

The edge is thin because their game are close to the GTO

No, that is not the only reason to why the bots could differ.

Quote

10-09-2013 , 01:19 PM

#17

Wotton

adept

Join Date: Sep 2012 Posts: 751

If they were playing both close to GTO wouldn't both red lines and blue lines be close to 0? (not rethorical, I'm really asking)

Quote

10-09-2013 , 01:20 PM

#18

MastaAces

Carpal \'Tunnel

Join Date: May 2007 Posts: 10,008

link should be fine.

Quote

10-09-2013 , 01:24 PM

#19

ohsosick

newbie

Join Date: Oct 2013 Posts: 45

Quote:

Originally Posted by Wotton

If they were playing both close to GTO wouldn't both red lines and blue lines be close to 0? (not rethorical, I'm really asking)

no

There are several equilibrium (even in smaller game like kuhn poker...)

Quote

10-09-2013 , 01:31 PM

#20

mperich

Pooh-Bah

Join Date: Nov 2004 Posts: 4,212

LOL @ using winnings at all to estimate how close they are to equilibrium. Double lol @ 100k hand sample.

I just coded 2 bots that are even closer to equilibrium for you. They both fold every hand no matter what the action. I played them for 100M hands (dwarfing your sample). Both winrates were 0. Zomg Nash.

Quote

10-09-2013 , 01:36 PM

#21

ohsosick

newbie

Join Date: Oct 2013 Posts: 45

Quote:

Originally Posted by mperich

LOL @ using winnings at all to estimate how close they are to equilibrium.

I use winnings AND deviation standard.
(ANd the edge on the other competitor)

Quote:

Originally Posted by mperich

Double lol @ 100k hand sample.

They played 1 million hand i think, but i 've only 10% on that sample on my HM.
All the hand are downloadalbe on le alberta website

Quote

10-09-2013 , 01:42 PM

#22

mperich

Pooh-Bah

Join Date: Nov 2004 Posts: 4,212

Quote:

Originally Posted by ohsosick

I use winnings AND deviation standard.
(ANd the edge on the other competitor)

They played 1 million hand i think, but i 've only 10% on that sample on my HM.
All the hand are downloadalbe on le alberta website

Winnings and standard deviation have nothing to do with nash (unless you are talking about winnings vs a best response strategy).

Quote

10-09-2013 , 01:56 PM

#23

ohsosick

newbie

Join Date: Oct 2013 Posts: 45

Quote:

Originally Posted by mperich

Winnings and standard deviation have nothing to do with nash (unless you are talking about winnings vs a best response strategy).

Imagine two different GTO strategy is played in NL HU 200 deep .
How will be the graph after 1 million hand?

And the high standard deviation amplifies this result ...

Quote

10-09-2013 , 01:58 PM

#24

NxtWrldChamp

Carpal \'Tunnel

Join Date: Feb 2008 Posts: 9,082

Quote:

Originally Posted by ohsosick

Imagine two different GTO strategy is played in NL HU 200 deep .
How will be the graph after 1 million hand?

How can you have 2 "different" GTO strategies?

If one is playing perfectly GTO and the other is playing a different strategy it would appear the 2nd is not playing GTO.

Also just showing us the results of 2 bots vs each other does nothing to show how good they are.