Yobz
04-01-2005, 04:34 AM
Quick background on myself:
I am a CS major and consider myself pretty decent at poker. At the end of last year, over winter break, I was really bored and decided to create a poker bot that could actually win something. The GUI was hacked together in about a weekend (with the HHs and a couple of API calls, it was a joke). The bigger problem came with the poker bot itself: the AI behind it.
I decided to go non-traditional and created a preflop table (simple to adjust, etc) to get the bot up quickly. Postflop play was more complex:
I created an interface that dealt random cards and had 1 neural net hook into the same table 10 times and I had it train (with the pot size being the marker of success). After a couple of days of training, the bot played damn well. I did the same with 9 instances, then 8, etc down to 2.
Starting in Early January I had the bot play party 2/4 4-tabling 7 hours a day, 4 days a week. If the table became 9 handed, the 9-handed neural net would take over, etc.
After a win rate of 3.24BB/100 and nearly 75k hands, I have over 10k in the account (including the initial investment).
Anyone who says bots cannot play for a profit is wrong. I have posted PT stats and other info here (http://www.veganstreet.com/news/aprilfools.gif) for the naysayers. I will not release the source code, so do not ask, I am merely pointing out that bots are out there and taking your money.
Have a happy april /images/graemlins/smirk.gif
I am a CS major and consider myself pretty decent at poker. At the end of last year, over winter break, I was really bored and decided to create a poker bot that could actually win something. The GUI was hacked together in about a weekend (with the HHs and a couple of API calls, it was a joke). The bigger problem came with the poker bot itself: the AI behind it.
I decided to go non-traditional and created a preflop table (simple to adjust, etc) to get the bot up quickly. Postflop play was more complex:
I created an interface that dealt random cards and had 1 neural net hook into the same table 10 times and I had it train (with the pot size being the marker of success). After a couple of days of training, the bot played damn well. I did the same with 9 instances, then 8, etc down to 2.
Starting in Early January I had the bot play party 2/4 4-tabling 7 hours a day, 4 days a week. If the table became 9 handed, the 9-handed neural net would take over, etc.
After a win rate of 3.24BB/100 and nearly 75k hands, I have over 10k in the account (including the initial investment).
Anyone who says bots cannot play for a profit is wrong. I have posted PT stats and other info here (http://www.veganstreet.com/news/aprilfools.gif) for the naysayers. I will not release the source code, so do not ask, I am merely pointing out that bots are out there and taking your money.
Have a happy april /images/graemlins/smirk.gif