Poker-AI.org

Poker AI and Botting Discussion Forum
It is currently Mon Nov 13, 2023 5:20 pm

All times are UTC




Post new topic Reply to topic  [ 5 posts ] 
Author Message
PostPosted: Tue Mar 18, 2014 10:49 am 
Offline
New Member

Joined: Sun Mar 16, 2014 12:53 pm
Posts: 4
I want to build a poker game tree of small size, such that it fit in memory. What are the state-of-the-art approaches? Before you ask about the gametype: it depends. If a fullring NL game tree fits in 8 GB of ram I would prefer this.

The next question is, how do I apply the lossless abstractions? Preflop card isomorphisms are pretty easy but how do I build the tree once a common card has been dealt?

EDIT:
In "Measuring the Size of Large No-Limit Poker Games" Michael Johanson states that with card ismorphisms FLHU has the following sizes:

PrF: 169 -> BF=169
PoF: 1,286,792 -> BF=~7614.15
T: 55,190,538 -> BF=~42.89
R: 2,428,287,420 -> BF=~44

In a FL HU game I have per betting round 29 nodes; 9 paths continuing and 10 terminal. Hence

169*(20+9*(7614.15*(20+9*(42.89*(20+9*(44*28)))))) which is 4.9657757 × 10^13

nodes, which is still not practicable. Is this correctly calculated? Is FLHU poker without lossless abstraction possible at all? I don't know how to start. How do you guys master the game?


Top
 Profile  
 
PostPosted: Tue Mar 18, 2014 3:30 pm 
Offline
Site Admin
User avatar

Joined: Sun Feb 24, 2013 9:39 pm
Posts: 642
http://poker.cs.ualberta.ca/publications.html


Top
 Profile  
 
PostPosted: Wed Mar 19, 2014 12:57 pm 
Offline
New Member

Joined: Sun Mar 16, 2014 12:53 pm
Posts: 4
Thats a nice bookmark thanks. I read a few of those publications, but as a matter of fact it is impossible to read and understand all of them (as a non native english speaker) in acceptable time.
I am starting at point zero and want to build my own bot with state of the art approaches, hopefully to be able take part at the ACPC. But at the moment it wont make sense to read all papers as some of them are certainly outdated.
So I am asking you for some hints on what to read exactly. Just some catchwords of techniques that are absolutely necessary would be fine. My current state is that CFR is the most efficient (better least inefficient) poker algorithm, with fictious play I approximate the ε nash-Equilibrium, and I need lossless card isomorphisms and lossy bucketing to fit the gametree in memory.
Are there other things that I have to take into account to create an agent that performs well?


Top
 Profile  
 
PostPosted: Wed Mar 19, 2014 1:08 pm 
Offline
Regular Member
User avatar

Joined: Tue Mar 05, 2013 9:19 pm
Posts: 50
Two papers (at least) have been published in regards to reviewing all the methods being used at the time:

Read them over, or at least the introductions, to get a good idea of what's currently doing well.


Top
 Profile  
 
PostPosted: Wed Mar 19, 2014 2:55 pm 
Offline
Site Admin
User avatar

Joined: Sun Feb 24, 2013 9:39 pm
Posts: 642
manuel wrote:
Thats a nice bookmark thanks. I read a few of those publications, but as a matter of fact it is impossible to read and understand all of them (as a non native english speaker) in acceptable time.
I am starting at point zero and want to build my own bot with state of the art approaches, hopefully to be able take part at the ACPC. But at the moment it wont make sense to read all papers as some of them are certainly outdated.
So I am asking you for some hints on what to read exactly. Just some catchwords of techniques that are absolutely necessary would be fine. My current state is that CFR is the most efficient (better least inefficient) poker algorithm, with fictious play I approximate the ε nash-Equilibrium, and I need lossless card isomorphisms and lossy bucketing to fit the gametree in memory.
Are there other things that I have to take into account to create an agent that performs well?


I find them nearly impossible to read and I am a native English speaker. :D The more recent papers usually reference earlier ones so there aren't many short cuts. A key concept in memory reduction is imperfect recall. Clustering is used to categorize boards. Different sampling techniques are used: public chance sampling, chance sampling, external sampling

I can't really help much more because I use my own technique.

https://www.google.co.uk/search?q=site% ... 2&ie=UTF-8

https://www.google.co.uk/search?q=site% ... ect+recall


Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 5 posts ] 

All times are UTC


Who is online

Users browsing this forum: No registered users and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Powered by phpBB® Forum Software © phpBB Group