Ms Pac-Man Competition
IEEE CEC 2009 Results
Simon M. Lucas


There were five functioning entries submitted by the deadline, but only four of these used used a screen-capture method (only screen-capture based systems are allowed in the competition).

Entry / Author(s)


ICE Pambush 2, Hiroshi Matsumoto, Chota Tokuyama, and RuckThawonmas

Ritsumeikan University, Kyoto, Japan

Max Chan

City University of Hong Kong

USM, Fitzgerald, Kemeraitis and Bates Congdon

University of Southern Maine

Ho Ho Kwong

City University of Hong Kong

Tómas Guđmundsson

University of Iceland


The results for the screen-capture methods are shown below with the high-score of each entry shown in bold.

The winning entry is the one that achieved the highest score given ten runs each.  The averages are also shown for interest, and it so happens that sorting by high score or by average leads to the same ranking in this case.

The winning entry was ICE Pambush 2, by Hiroshi Matsumoto, Chota Tokuyama, and Ruck Thawonmas of the Intelligent Computer Entertainment Laboratory, Department of Human and Computer Intelligent, Ritsumeikan University.  The team improved significantly on their WCCI 2008 entry.  ICE Pambush 2 does a very good job of extracting the game objects from the screen capture, and as the name suggests, does a good job of luring the ghosts to the power pills and them ambushing them.

Ruck Thawonmas has kindly made the code for ICE Pambush 2 available for others to build on (zip file here).

The second place entry from Max Chan is also worth a mention - this performed nearly as well as ICE Pambush 2, and also made a competent job of extracting the game objects.  The University of Southern Maine team provided a tweaked version of their WCCI 2008 entry and improved slightly on their WCCI 2008 high score.

The Entry by Tómas Guđmundsson interfaced directly to a machine emulator running the original Ms Pac-Man ROM code.  This has the advantage of perfect game-state information with minimal perceptual and actuator delay.  The screen capture methods take a significant number of milliseconds to perceive the state of the game, and also to send a key event to the game window to control the agent.  Tomas's entry achieved a score of 39,410 but there were also some peculiarities noticed with the emulator e.g. for much of the time the game was deterministic, and occasionally the ghosts disappeared completely.  A deterministic game can be solved by following set routes and therefore obviates the need for intelligent behaviour.  While this could not qualify as a regular entry, the Evolution Strategy approach used by Tómas seems very interesting, and we look forward to a future entry from him using a screen-capture method.


ICE Pambush 2 can be seen in action on Youtube.


Two versions of the Ms. Pac Man competition will be run for IEEE CIG 2009: the screen-capture version (described here), and also modified version where people can enter either a Pac-Man agent or a team of ghosts.  More details coming soon.


When you have a technique that works well, you may want to write it up as a paper for the IEEE Transactions on Computational Intelligence and AI in Games.


We encourage entrants to try evolutionary and machine learning methods.  The papers below might provide a useful starting point.

Also see the Pac Man papers in IEEE CIG 2008.

[1] Simon M. Lucas, Evolving a Neural Network Location Evaluator to Play Ms. Pac-Man, IEEE Symposium on Computational Intelligence and Games (2005), pages: 203 -- 210 [pdf]

[2] Szita and A. Lorincz, Learning to Play Using Low-Complexity Rule-Based Policies: Illustrations through Ms. Pac-Man, Journal of Artificial Intelligence Research (2007), Volume 30, pages 659-684 [pdf].


end of page