Poker-AI.org
http://poker-ai.org/phpbb/

HU Nash Table
http://poker-ai.org/phpbb/viewtopic.php?f=24&t=3254
Page 1 of 1

Author:  nefton [ Sun Sep 15, 2019 5:20 am ]
Post subject:  HU Nash Table

Want to repeat HU Nash table. (as a warm up before cfr)

Here is code:
Code:
class NashCalculator {
public:

   uint8_t PushStrategy[169]; //probability 1/0xFF parts to push
   uint8_t CallStrategy[169]; //probability 1/0xFF parts to call
   
   double PusherPushGain[169]; //gain from push action (only vs specific call strategy)
   double PusherFoldGain[169];
   
   double CallerCallGain[169];
   double CallerFoldGain[169];

   void CalculateGains(HU_LUT* HL, int small_blind, int big_blind, int ante, int stack);

};


Some questions appear:

1. As I see on nash table -> there are no mixed moves (with probabilityes).
So seems there are nash in non mixed moves?

2. My plan is:
- take some random or 50/50 strategies
- calculate pusher and caller gains
- then change strategies in some way (gradient? min max?)
- repeat until (dont know for now )

May be there are some no itterative methods?
Or may be I doing somthing wery wrong from a scratch?

Author:  spears [ Sun Sep 15, 2019 12:30 pm ]
Post subject:  Re: HU Nash Table

I would guess you can solve this problem with fictitious play or CFR, or maybe linear programming. At convergence the ev gap between the best response of player 1 and the best response of player 2 is zero, or very small

Page 1 of 1 All times are UTC
Powered by phpBB® Forum Software © phpBB Group
http://www.phpbb.com/