agnar22 / pluribus Goto Github PK

I'm surprised that I am the only one who saw this code and thank you, you saved my life (can I offer you something?) because I have reimplemented it.
I understand how mccfr works pretty much well but I don't understand how real-time search works. Is it used lcfr for that?
I thought that the only reasonable way of doing that is to use the pre-trained regrets and sets the hero game steps fixed for one player.

For example, if the action is:

• sb/bb: 1/2$
• Player 1 raises 4$
• Player 2 calls 2$
  Flop:
• Player 2 checks
• Player 1 bets 6$
• Player 2 raises 18$
• no solution

Then in the lcfr (?) do I have to set fixed this first actions for n° time iterations?