[ExI] AI poker

[John Clark]

On Sun, Jan 29, 2017 at 11:04 AM, Dave Sill <sparge at gmail.com> wrote:

There is no "correct" decision for every hand because all of the players
> have incomplete knowledge.
>

​That's what makes it so impressive, the real world is not Chess, Jeopardy
or even GO, decisions must be ​
made ​and in the real world you very seldom have complete knowledge.



> ​>​
> These competitions consist of playing thousands of hands, and the
> advantage goes to the players best able to calculate the probabilities of
> the cards
>

​If it were just a question of probabilities computers would have beaten
humans at Poker in the 1950s, but only now is that becoming true. ​

​Their recent success is all the more impressive because to the best of my
knowledge none of the Poker programs have code to detect facial ticks or
other tells in their human opponents, at least not yet.  And they win even
without that.


> ​>​
> and their opponent's behavior.
>

​That's the problem. I can calculate odds too, so if you always play the
odds and never bluff but I occasionally and unpredictably bluff then I will
always know what you're going to do but you won't know what I'm going to
do. ​So I win in the long term.


> ​> ​
> Clearly, the AI has the edge in calculating the card probabilities. It
> also has the advantage of perfect memory of every hand played and every
> time an opponent bluffed.
>

​Not necessarily. ​If I have a weak hand and bluff, and you get cold feed
and fall for my bluff and fold you'll never know if I was bluffing or not.
And what about you, when should you bluff? If you've figured out my
bluffing strategy you can't use that yourself because I'll recognize it,
you have to figure out your own strategy. clearly there should be an
element of randomness in making your bluffing decisions but how much? Too
little and I'll be able to predict when you're bluffing, too much and
you'll lose money making with too many illogical bets. What is the sweet
spot for any given hand? AIs are starting to figure that out.

  John K Clark
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20170129/427b50b7/attachment.html>