|
hello, I am trying to model my opponents using neural nets. My plan was, that i have a model for action prediction which is based on observed things like board texture, past actions etc. Then i use the showdown hands to predict the range of the opponent at showdown.
What is confusing me now, is that many people in this forum write, that they are using things like EHS in their action prediction model.
However, as i can only observe the hole cards at showdown, how can you predict ranges on flop for instance?
I can only think of 2 options:
first you use the estimated cards based on the Showdown hands, which would imply a huge bias, but maybe it is good enough to improve the cction prediction?
second you use some rule based approach, like on the flop you say: OK,pf he has opened from the button so i guess he has top 30 percent of his range etc.
i would be very thankful, if someone could point me in the right direction.
|
Login
Register
FAQ
Search
