Poker Bot - I've coded it... but it's not winning - need AI guys. - Page 3

axioma · #21 10-04-2004, 01:20 PM

you got to realise that this subject is going totouch a nerve with many posters here.

i actually have a lot of AI experiance, incuding coding neural nets. i might have been interested if i wasnt already busy, but the tone of your post makes me realise you are in this for the wrong reasons.

if you had been comming from a more accademic point of veiw, and were interested in the theory behind this kind of project, rather than trying to cheat for money i might have been interested myself.

as is, i doubt you will get any positive responces here.

dokomoy · #22 10-04-2004, 01:53 PM

I did this with blackjack for ap computer science 2 years ago, and I've wanted to do the same thing for poker for a while now, any possibility on either releasing the source or the executable?

edit: Because this clearly isn't popular I'd like to add that I have no intention of using any bot, but I do think it would be immensely intresting

axioma · #23 10-04-2004, 02:02 PM

correct BJ strategy can pretty much be summed up in a 1 page chart, poker is a little deeper... [img]/images/graemlins/wink.gif[/img]

bisonbison · #24 10-04-2004, 02:57 PM

Unsurprisingly, this thread sucks.

I think bots are unethical, but I find myself unable to convince the congenital idiots who keep bringing them up, so let me spell out the practical impact of bots as clearly as possible:

If casual players believe that bots are a widespread problem, they will stop playing online. Without casual players, there are no games.

Without effective security to prevent their use, bots will destroy the environment they are designed to exploit.

jdarwin · #25 10-04-2004, 03:18 PM

Well i've tried 1-2... but maybe i didn't put in a big enough bank roll for the swings... either way... it's not that simple... there are ALOT of variables... i need opponent modeling so I can remove much of the guesswork that the bot would be facing... and this is possible with the implementation of neural networks...

anyway... some people have expressed interest in coding this with me... so we will see where it leads.

LikesToLose · #26 10-04-2004, 04:13 PM

I have a lot of history at the 1-2 games and can probably tell you where your bot is making mistakes. PM me your player name and I'll look up 'his' play in PT, or just send me your hand histories.

I don't think it would take too much to beat those games.

Praxx · #27 10-04-2004, 04:46 PM

[ QUOTE ]

If casual players believe that bots are a widespread problem, they will stop playing online. Without casual players, there are no games.

Without effective security to prevent their use, bots will destroy the environment they are designed to exploit.

[/ QUOTE ]

With any luck, we can rely on alot of amateur programmers, wanting to get rich quick, to lose alot of money trying to write a bot that plays well (which is difficult.)

I wonder how much this happens already! jdarwin of course, is a good example. [img]/images/graemlins/wink.gif[/img]

meow_meow · #28 10-04-2004, 05:26 PM

[ QUOTE ]
Well i've tried 1-2... but maybe i didn't put in a big enough bank roll for the swings... either way... it's not that simple... there are ALOT of variables... i need opponent modeling so I can remove much of the guesswork that the bot would be facing... and this is possible with the implementation of neural networks...

anyway... some people have expressed interest in coding this with me... so we will see where it leads.

[/ QUOTE ]

I assume that by "opponent modelling" you mean changing your play based on previous play be a specific opponent. If you cannot beat 1-2 without the above, then I think there is a big problem with your decision rules dealing with what you are holding vs. the board - hand reading basically. If you've also taken into account position and # of players remaining in the hand (as well as pot size), and still can't beat the game, I don't see that "opponent modelling" is going to make a very big difference. Maybe I'm underestimating, and just do this unconsciously, but I'd guess that the read I have on a player actually make me change my action (from what I would do with no read), in less than 5% of the hands I'm involved in (or not much more than 1% of total hands).

dokomoy · #29 10-04-2004, 05:50 PM

[ QUOTE ]
correct BJ strategy can pretty much be summed up in a 1 page chart, poker is a little deeper... [img]/images/graemlins/wink.gif[/img]

[/ QUOTE ]

And thats exactly what makes it intresting :P

Cosimo · #30 10-04-2004, 06:02 PM

[ QUOTE ]
But philosophically, what's the objection?

[/ QUOTE ]

Philosophically, the objection is that it will destroy the games.

I think the dog analogy doesn't match the situation. A bot can play 24/7. A bot can play at every single table that sites host. A handful of people running bots means that there are no players left to take money from.

As a game programmer, the AI challenge interests me. However, if this code got out, it could quickly destroy online poker. No more games, no more money. I have no interest in playing at a glacial 30 hands an hour, if I can get in up to 500 an hour playing online.