My poker bot

[Greeksquared]

Over the past year I have created a poker bot that plays heads-up limit Hold'em. I used two free software packages (R and Tinn -R) to write it. There are about 5 or 6 different files that need to be downloaded before it can work.

I am writing this post to see if there is any interest in anyone playing it. As of today, the AI is very simplistic , but decent enough that it could hold its own against marginal opponents. I do have plans to up the intelligence quite a bit, but it will be at cost of running time.

I posted it here because I generally believe the input from this forum is usually best. I also want to know if anyone knows anything about poker bots that actually play online for actual money.

- Greeksquared

[oyvindgee]

I assume your interests in writing this bot is purely academical and that you have no intentions of actually using it to play for real money. Cause I think the vast majority of this forum is of the opinion that people with such intentions should be kicked once in the nuts for every dollar their bot has ever earned.

[SamIAm]

Are you inviting us to play against it for money? What table?
-Sam

[Shroomy]

I probably know as much about online bots as anyone who is going to answer you.

what do you want to know?

and oyvindgee .. I agree with you, despite having done work that I know was used for a comercial bot (but I am unsure of its success) but when the therapy for your 6 yr old daughter (autism) costs more than harvard does per year you would be surprised what you can live with.

[Greeksquared]

I just created this thing in my spare time. I actually created it just for fun, and dont really have any intentions of using it for a profit.

Honestly, it seems very possible to me that a bot can be created to beat low limit poker. I am very curious as to how the bots are created. What kind of intelligence is used to make decisions?

To me it seems like you just need a few things to make the bot pretty decent. First thing you need is some kind of preflop hand ranking. From here I list in order what I feel is most important

1. Number of all possible hands at each street that are better than hand
2. Potential hand value by next street
3. Overall strength of hand (which is basically a subset of 1)
4. Pot odds
5. Randomization of play

As of now my program just incorporates number 1, and seems to do alright.

So, how fast do the bots work? What is their win rate? How wide spread are they? Can they multitable?

Party poker actually has something of a bot. I believe they have a section in their software where you can play some bots. It even tells you your best possible choice at each decision point.

[NewUser2006]

Did you make your own interface for this bot or does it actually piggy back on the PartyPoker software?

[Shroomy]

[ QUOTE ]
I just created this thing in my spare time. I actually created it just for fun, and dont really have any intentions of using it for a profit.

Honestly, it seems very possible to me that a bot can be created to beat low limit poker. I am very curious as to how the bots are created. What kind of intelligence is used to make decisions?

To me it seems like you just need a few things to make the bot pretty decent. First thing you need is some kind of preflop hand ranking. From here I list in order what I feel is most important

1. Number of all possible hands at each street that are better than hand
2. Potential hand value by next street
3. Overall strength of hand (which is basically a subset of 1)
4. Pot odds
5. Randomization of play

As of now my program just incorporates number 1, and seems to do alright.

So, how fast do the bots work? What is their win rate? How wide spread are they? Can they multitable?


[/ QUOTE ]

a few random comments and answers.
How fast do bots work .. it obviously depends on the speed of the cpu and a bunch of stuff, such as if you are computing things based on simulations or full combinations. Typically there is a mixture of both depending on the situation and you can adjust the number of simulations to be reasonable.

Can they multi table.. certainly

The obvious thing I see you missing is the probability that the opponent has one of the better hands. The two basic things to think about are what are the chances due to hand distribution (ie AKs is more likely than AKo) and due to betting that has occured. (i.e. 96o is pretty unlikely if they raised preflop)

Randomization is not as important as you think, It will make enough mistakes to be random [img]/images/graemlins/smile.gif[/img] but yes incorporate randomness .. thats pretty easy.

You should look at things from an expected value point of view as opposed to pot odds and hand strength (since it takes both of those into account)

the PP bot is indeed the UofA guys, they have done some great work. (and they started poker academy pro)

What is the win rate? That varies greatly from consistant losers to big winners, but the rate has more to do with the number of tables that can be played and the strength of competition.

The big winners typically have 100 tables being played most of the day and are grinding it out.

[Yads]

[ QUOTE ]
1. Number of all possible hands at each street that are better than hand
2. Potential hand value by next street


[/ QUOTE ]

This is ok, but may lead you to the wrong conclusions. It's not just about how likely a hand is to win a showdown, it's also about how well does a hand play post flop. For example A2o is a better hand than JTs by your criteria, but JTs plays infinitely better and can extract more money post flop than A2o.