How Poker Artificial Intelligence Has Evolved Over The Years




Poker has been used as a social game for fun or even for some money at the casinos. Some people have perfected the game so much that they actually make a living from it.

Poker has evolved over the years with the incorporation of AI which dates back to 1984.
This was the first ever Poker AI which was a basic Poker AI software called Orac that was created by Mike Caro competed in the World Series of Poker.

Poker Sites UK gives a visual representation of this evolution, however,  this article will give an in-depth overview of AI in poker along with insights from the research to date.

1984

The first ever Poker AI software called Orac was developed by Mike Caro who had spent two years developing the software.
It competed with humans during the world series of poker and though not as developed as today’s AI, it was able to beat the poker guru Doyle Brunson in one match.

1997

Loki, a Poker AI software was introduced by The University of Alberta Computer Research Group. The Loki Poker AI software was designed to play a full table of nine players and it focused on the Limit Hold’em technique instead of the No-limit variation.
Though expecting an advancement, Loki’s poker skills lie below the average human poker player.

2003

The transition from old models started becoming more pronounced in 2003 having Poker AI developers transition away from Chess Model. This is because researchers began to shift their focus away from chess methodology that had been used in the previous years. One of the key role players included developer Michael Bowling who was the driving force behind the research team at the University of Alberta.

2005

The World Series of Poker Robots was held at Golden Palace which hosted six different entries.
A Pokerprobot owned by Holton Givens merged victorious. However, the Pokerprobot narrowly defeated Poki-X which is an advanced version of Loki from the University of Alberta.It also went head to head with human pro Phil Laak who won during the warm-up. It is believed that Pokerprobot will advance over the years.

2006

Annual Computer Poker Competition which was once known as ACM officially begins having various universities competing for the various awards. Heavyweights that took part in the competition included teams from the University of Alberta and Carnegie Mellon University.

2008

Polarism, a Poker AI bot was put through an informal test during the gaming life expo which it competed against six humans. It was able to beat humans by three wins but also recorded two losses and one tie. During this year, the bot was upgraded from its predecessor and was set to challenge human pole pro Phil Laak but lost<

2009

Similarity Assessment Reasoning for Texas Hold’em via Recall of Experience commonly known as Sartre was introduced by the University of Auckland.This was one of the rare successful poker AI that would go on to compete in the AI world and performed well in competitions over the years.The interesting thing about Sartre is that it never came from the Heavyweights, Carnegie Mellon or the University of Alberta.

2012

Over the years,  bots were not good enough to compete against humans in online poker but with development in AI software, programmers saw potential in this lucrative market to develop advanced Poker AI bots that would challenge their human counterparts. However, most gaming sites like PokerStars and Full Tilt do not allow bots to play, and once an account is detected as a bot it is banned from the site. Some casinos like Bellagio casino in Las Vegas installed a $2/$4 Limit Hold’em bot that everyone could play against.

2015

Cepheus, an AI software from the University of Alberta was able to solve and master Limit Hold’em by playing itself for two months.However, Cepheus’s  No-Limit version remains unsolved and some scientists believe it may remain the same over the years due to the unpredictable nature of the game. Cepheus and other similar AIs are mainly programmed for use with other applications such as helping governments by negotiating tactics or improving security strategies It can also be used to help doctors modify treatments for their patients.

2016

Carnegie Mellon and Tuomas Sandholm collaboratively release Claudio, their own super-smart poker AI.This was a move to outshine Cepheus from the University of Alberta.Claudio was much advanced as it played No-Limit Hold’em which is hard to master but this was not good enough for Claudio as it was defeated by human poker pros during a 13-day challenge at Rivers Casino.

2017

DeepStack AI a new algorithm that was developed by the University of Alberta shines as the first AI to beat professional players in No-Limit Hold’em. It employs deep neural networks to emulate human intuition and learn on the go.It became a notable poker player because it won without training from expert poker players. However, Carnegie Mellon and Tuomas Sandholm challenge this study from the University of Alberta and wish for DeepStack to go head to head with great AI’s like libratus. This is because, in January 2017, the most decisive blow in the history of human vs. poker AI was witnessed as Libratus stole the show.

So you think you have what it takes to be poker champ? Well, why don’t you give libratus a challenge?