Talk:Pseudo-optimal

From PokerAI

Jump to: navigation, search

Ups! We got duplication. We have Pseudo optimal (notice the lack of "-"), refering directly to E-Nash equilibrium. Doublecheck -- How these two relate?

Pseudo-optimal typically refers to strategies that are derived from optimization on a simplified or abstract game while epsilon-Nash optimal refers to how exploitable a strategy is.

So, pseudo-optimal is Nash equilibrium in abstract game (which will also be E-nash in theoriginal game, but E-nash is basically everything, where E simply says how far it is?)

E-Nash is basically measuring how exploitable something is - how bad the worst case scenario is compared to the equilibrium. One of the approaches to making poker AI is to work at shrinking the exploitability rather than using an analytic solver.

I'm not really sure what pseudo-optimal refers to, but I'm thinking it means bots that try to play Nash equilibrium strategies rather than exploitative ones. (The UofA people use 'optimal' to refer to equilibrium attempts and 'maximal' to refer to exploiters.)

Retrieved from "http://pokerai.org/wiki/index.php/Talk:Pseudo-optimal"