Hello,
I would like to know what is the algorithm that is used to create Nash Equilibrium tables like these:
http://www.holdemresources.net/h/poker-theory/hune.html

Some values of those tables look strange to me. (E.g. "push" with 54s is 10 times better than 54o. Why? Probability to hit flush is only ~2% higher)
I would like to verify that Nash tables are correct.

Waaaat

Enviado do meu LG-D405 através de Tapatalk

Hello, my name is Elder Price

Hi.

There are many algorithms for finding Nash equilibria and even the simplest ones require knowledge of linear programming so I'm not quite sure you wanna do that.

There are two basic reasons why the table says you wanna shove with 54s up to 20bb but only 2bb with the offsuit variant:

First, the tables are simplified - there exists no pure equilibrium in poker (meaning you take specific action in specific situation 100 % of the time), only mixed equilibria (you fold x % of the time, shove y %). The tables only represent a pure strategy that is very close to the optimal, mixed one (that's not neccesarily a bad thing, it's hard to work with frequencies right).

Another reason - you kind of want to be able to cover a lot of cards, obviously only shoving 5 when you have another 5 to it doesn't get you much value from hand like 22. (Think of it as balancing ranges - what is the best card I can add to 5 to be bluffing with?) You might want to shove 54 ~20 % of the time, but that's how often it's suited - why bother shoving a (marginally, true) worse hand and folding a better one?

I am a programmer, I would like to see some links to "many algorithms for finding Nash equilibria".

Well, this depends on whether you use a normal form or an extensive form notation. While most of the poker games are too large to work with them as normal games, push/fold holdem is reasonably small for this and the normal form algorithms are much simpler.

Then you do linear programming.

Nash isn't even that usefu above 6-7bb HU poker, it's already been calculated and you already know how to code (i.e. not doing it to learn how to code).

Why waste your time with this?

The tables you linked were calculated using Fictitious Play:
https://en.wikipedia.org/wiki/Fictitious_play

CFR+ is a newer alternative:
https://arxiv.org/pdf/1407.5042.pdf

Anyway, the raw (unsimplified) data is also provided here, this may be easier to verify.

You may find this interesting:
http://www.husng.com/content/will-ti...video-pack-2-0