About Project

Games are suitable and interesting domains for artificial intelligence research since they include well-defined set of rules and specific goals. However, on the other hand, the game of poker has same natural additional characteristics, making it strategically very complex. Poker is an imperfect information game which also requires calculations over stochastic game states. All this features make poker an excellent domain for investigating problems of decision making under conditions of uncertainty.

Poker Playing Agent project has been started in September 2014 with the intention of seeking effective solutions to famous computer poker problem in result of dynamic enthusiasm of four young computer engineering candidates who wants to explore the limits of computer intelligence.

The aim of the project is to bring a competitive agent to computer poker literature. Competitive in this domain can be described as capability of making rational moves to maximise the incomes, ability of generating real time responses during the game and capacity of winning against at least an average human player. Our agent is able to logically analyse the conditions and make reasonable decisions based on many different dimensions of imperfect information games. Also, it learns from the past experiences to enhance the intensity of rational decisions like in the real life. One another feature that boosts the success of our agent is opponent modelling. In the poker game, opponent creates the hidden information and passing this obstacle with an efficient algorithm enables our poker playing agent to act according to whom it is currently playing with. To briefly describe our agent, it is composed of two sub-components: Decision and Learning part. Two continuously feed each other and create a self-learning agent.

To realize all these features mentioned above, Bayesian Poker methodology is adapted during the design and the implementetaion of Poker Playing Agent project. A detailed bayesian network structure has been created to encapsulate game state information and players' characteristics. Network consists of 10 nodes, 6 for opponent modelling and 4 for hand strength observations. This network is being updated every once game turn changes and gives us a win probability of the agent to decide its next most profitable action. Then we are utilizing the agent's wn posibility with its opponent's behaviour and randomness.

We, A4 team, believe that after a groundbreaking solution is found to this problem, there will be motivation and ready to use knowledge to explore daily life problems including hidden information such as weather forecast more deeply and this will be a chance to enhance the quality of people’s life experiences. With this belief, we investigated the literature to learn about current state- of-the-art methodologies, thought about how we can improve the current solutions and hoped that we created a basement for the forthcoming researchers of the field.

An example game of our agent can be seen in the following video.

About Poker Academy

Poker Academy was launched in December 2003 as a commercial Poker Training package. This simulation package contains various famous Poker Agents, important amount developed by Computer Poker Research Group of University of Alberta.

Poker Academy provides a Java based API, called Meerkat, which allows developers to adapt their own custom agents into the system. This creates an excellent simulation and testing environment for poker playing agents. You can easily play against your agent in a quality and handy graphical user interface or you can have your agent play against famous bots in the literature. The program also keeps track of all the hands played and can display comprehensive graphs and analysis of the player statistics.

Unfortunately, Poker Academy is not available since the creative company does not exists anymore. Current availability is being supported by Poker Genius, which offers a similar tool which has almost all features of Poker Academy to computer poker world. Site has not been maintained lately but more information about Poker Academy can be found in here.

We would like to thank Ömer Ekmekçi since he introduced us this system and provided some necessary packages for the development.

Screenshots of the Poker Academy can be found in Screenshots section.