Poker-AI.org

This is not for sale anymore

why do you call it GTO bot? it is surely not a gto bot if you used deepstack implementation.
Also you claim that it is undetectable on pokerstars using VM which obviously means that you didn't run it long enough on pokerstars.

How is DeepStack not a GTO bot?

DeepStack's method is proven to converge to e-nash equilibrium given enough granularity/resources.

Can you post your huber losses?
DeepStack's river huber loss is essentially 0 since they did not use a river net.

As the other users have explained this is a GTO bot. Of course this bot has not completely solved HUNL Poker. However with 10 million samples, it is an extremely close approximation of a nash equilibrium and theoretically the algorithm could eventually reach a nash equilibrium.

This bot setup is UNDETECTABLE. Using my setup Pokerstars is unable to see any bot activity. The only thing running on the VM will be Pokerstars itself. Everything else will be running outside the VM making it undetectable. Also you will need two mice and two keyboards. The VM will be configured such that one mouse and keyboard will not be recognized(so no suspicious clicks on the actual machine are detected). I want to make clear that it is physically impossible for Pokerstars to even detect any bot activity using this setup.

Regarding deepstack being gto bot, i was not aware that it converges to nash equilibrium. According to what i read and watched videos how it plays, it uses a lot of approximations and plays quite non standard.

First of all VM is already a red flag for pokerstars. Almost none of real users use VM to play on pokerstars so it is a good reason to look closer in your account.
It is easy for pokerstars to detect that it is running inside VM.

2nd problem is that they can detect that the mouse and keyboard input is not coming from hardware but from software. Some people solve this issue by having own mouse/keyboard hardware which can take input from software and send click commands.

3rd issue is that you need human like mouse movements because if you move mouse artificially it is easily detectable. One way to solve is to record human mouse movements and replay them.

4th issue is that such bot is very famous and produces identical stats after even small sample of hands (like 10k hands), also uses very non standard ranges, bet sizes, lines which is easily detectable.

How many hands did you play on pokerstars undetected and how do you approach above problems?

can somebody please provide link it is explaind why deepstack is a gto bot?

Go to deepstack.ai download Paper pdf and it should be clear that this is a GTO bot. Now lets not waste our time with this futile discussion about GTO or not. Deepstack has beaten poker players by 50 big blinds /100. That is an amazing winrate. As long as it is better than humans we are fine.

1. It is actually not that easy detect if something is running inside a VM or not. Even if they can detect it a VM alone is not a reason for a ban. You could be a linux or a OSX user.

2. This is false, the VM is setup such that you have a keyboard and mouse for the VM so the clicks are coming directly from there.

3. This bot is a live assistance tool and does not play by itself. This means a human is actually clicking every click. Making your point invalid.

4. This is completely incorrect. I am the only person who has this exact model. The Stats are not identical and 5 people using the same bot is completely irrelevant with 100000 users on pokerstars.

You have incorrectly assumed this is a 24/7 bot that plays by itself. Such a bot is extremely detectable. This bot is a live assistance tool which means you will click the actions yourself. Therefore it has human behaviour.

I hope your points were answered.

Hi yes you answered my questions. i didn't know that it is asisstance tool. That makes it undetectable actually because human is clicking so you don't have any issues with mouse or keyboard or any artificial behavior.

My point regarding having identical stats is that if the bot plays gto, it is having very typical stats. Your bot is not the only one that plays HU gto. There are few gto solvers asisstance tools that are able to play gto. All of those will produce identical statistics.
Sure if you are the only one to use it no issue. but now you shared the code, meaning that likely multiple users will use it, so all those accounts will have identical statistics even after very small sample.

I have a rule based bots that uses statistics to adjust to different opponents. If i run multiple accounts on the same site i can already after small sample see that stats are very similar making it obvious bot ring. so i have to modify preflop charts in order to produce different stats. still there are many similar things that those accounts do.

There is one company, "Safe Poker" that compares player statistics, bet sizes, and lines they take, and that way they easily detect bot rings. They claim on 2 plus 2 forum that they sell their service to multiple poker sites. Likely poker stars is able to do the similar thing.