OpenAI and online poker
- Thread starter Madoxx
- Start date
PURSUIT MEDIA
Elite Member
- Dec 27, 2019
- 3,133
- 823
Good idea,open ai comes to gaming
- Oct 11, 2016
- 1,234
- 352
poker is already solved when it is 1v1 format, there is already a software that allows you to play perfect 100% of the time
ygg
Senior Member
- Feb 8, 2011
- 1,099
- 587
which software?
Morty073
Newbie
- Jan 2, 2020
- 18
- 13
I worked for over a year on beating online poker with AI, and my conclusion is that poker is very resistant against machine learning.
That's because unlike other games like chess you don't have the complete information given(other players cards), which is a weakpoint of current AI.
That's also why state of the art software relies on search trees rather than AI for solving poker.
On top of this(and which people often underestimate): poker is a game with extremely high variance, meaning you can play hundreds of thousands of hands against worse players and there is still a relatively high chance that you lose money, which makes evaluating your AI very difficult.
So i think it's very difficult to pull off, even if you only want to beat microstakes.
That's because unlike other games like chess you don't have the complete information given(other players cards), which is a weakpoint of current AI.
That's also why state of the art software relies on search trees rather than AI for solving poker.
On top of this(and which people often underestimate): poker is a game with extremely high variance, meaning you can play hundreds of thousands of hands against worse players and there is still a relatively high chance that you lose money, which makes evaluating your AI very difficult.
So i think it's very difficult to pull off, even if you only want to beat microstakes.
But my understanding of AI is that the most powerful aspect of it is exactly this - dealing with uncertainty. And yes since poker isn't pure calculation, it might even be good to have a tool that deals with uncertainty, yet doesn't have emotions.
If/else cannot deal with uncertainty and doesn't have emotions
Humans can deal with uncertainty and do have emotions
AI can have both...
... Or not
Maybe I'm mistaken
The point is why not try it in the beginning with chatgpt giving advise how to play, and later on developing a complete tool using openai api...
- Oct 11, 2016
- 1,234
- 352
- Oct 11, 2016
- 1,234
- 352
variance can be controlled depending on the style you played, so you will not lose over 100k hands, if you are losing over 100k hands, it simply means you are a losing player, it has nothing to do with variance
- Oct 11, 2016
- 1,234
- 352
there is already real time assistance software, are you in the scene? it seems that you are not if you are suggesting something like that
Ashk881
Elite Member
- Aug 24, 2021
- 1,695
- 2,071
Traditional Ai methods can't really deal with uncertainty. Infact, an ai is going to perform worse if a situation is not predictable.
All ai systems are what we call Universal Function estimators. You give therm some inputs and the ai figures out some estimated function that maps your input to a certain output. The less random your outputs, the better the estimation of this function and better the ai's performance.
The only good way an ai will perform decent on imperfect games is when you hone *intuition* in the model. This is usually done through deep reinforcement learning. The model sees millions of poker games and figures out an algorithm we don't understand well. And this shit is expensive and you can't really measure the performance of something like this.
This is why game theoratic solvers are most common for poker.
Are there any monitoring tools to figure out cheating?
As an experiment I mean...
Idk guys it was just a random thought, I see some people here claim to have or really do have much more knowledge on the subject
Morty073
Newbie
- Jan 2, 2020
- 18
- 13
Variance is exactly the reason why you can be a losing player over 100k hands.
The more hands you play the better your assumptions if the money you won/lost are an actual representation of your skill.
- Oct 11, 2016
- 1,234
- 352
I am just sharing that no winning players will be losing over 100k hands.
If anyone loses over 100k hands, it simply means they r a losing player and it is not vairance causing the losses
- Oct 11, 2016
- 1,234
- 352
many people have been accused of using real time assistance. monitoring tools will be useless. Impossible to gather enough evidence to say that someone is cheating
Roger Marquez
Elite Member
- Apr 17, 2017
- 5,126
- 7,816
I used to play online poker professionally for about 7 or 8 years and I am pretty sure that type of software already exists. The only difference is you need to press the buttons but the software will tell you what is the best either exploitable or GTO play (only two ways to play poker) for any given situation.
standardrecovery
Newbie
- Mar 30, 2023
- 20
- 4
Would be cool if it happened
I just spent 30 minutes trying to play against ChatGPT, but there was constantly a problem with it telling me the cards it has, not wanting to bet,check,raise...
Another conclusion is that 'Too many requests, try after 1 hour...' appears when you use ChatGPT for something like this. Sometimes I use it for coding help and spend hours without it kicking me out. Very interesting...
Another conclusion is that 'Too many requests, try after 1 hour...' appears when you use ChatGPT for something like this. Sometimes I use it for coding help and spend hours without it kicking me out. Very interesting...