Player Exploit AI Models
Hi everyone, I've been working on a new project that I am looking for some beta testers for to try it out and gather feedback.
It is essentially an AI model that completely replaces the use of hand histories!
It was trained to accurately mimic how players tend to play in different spots and estimate what their ranges are.
So for example, you can provide it any hypothetical scenario in a hand and it can give you the estimated ranges of each player based on their actions/stacks/etc. It also gives you accurate estimates of their fold percentage to different bet sizes, etc.
I've tested it on a large number of hands and it is very accurate and well calibrated.
The problem with normal hand histories is usually sample size. So let's say villain opens 6BB RFI (huge), and makes a 50% c-bet on the flop. What is their range? What percentage will they fold to a raise?
This would be difficult to estimate with hand histories because the sample size of similar hands in your database will likely be too small. But the AI model has no problem providing accurate estimates of all these important properties and more.
If you're interested in trying it out (for free) and providing feedback for me as I continue to develop it, then please message me 😀
4 Replies
so how does it work?
I'm not sure exactly what you mean. Are you asking on a technical level or a usage level?
From a technical perspective, it is composed of multiple neural network models trained jointly on a large dataset of real hands.
From a usage perspective, it essentially predicts what a players range is and what they will do.
So for example, let's say you've played 100 hands with someone. The model will analyze those 100 hands, and can infer things about that player. It might see a single hand where they showdown with a weak hand that should have folded preflop and never called flop.
So it can make predictions based on all that data, and it can estimate how often they will fold on a dry flop to a 1/3 pot bet, for example.
With only 100 hands on a player, it is hard to get much info/exploits using traditional statistics. But with a large complex model that's seen hundreds of millions of hands, it can easily predict that they are a loose calling station and how often they will fold on certain flops, etc.
Hope that helps provide some context! Nobody has reached out yet to try it lol. You can also message me on discord at Ty4Reading
I added you on discord, but you don't take my friend request. (you have an avatar with two dogs right?)
Added you! Still looking for anybody interested in beta testing (for free) and providing feedback.
