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


There were four functioning entries submitted by the deadline, plus one that I was unable to run.
Congratulations go to the winning team behind ICE Pambush , with a new screen capture software agent Ms Pac-Man world record of 30,010

This run was performed live during the Ms Pac-Man competition session at the conference, and was exciting to watch, with the entry surviving many seemingly impossible situations either based on it's implicit knowledge of ghost behaviour, or perhaps more likely, based on sheer dumb luck!  The complete software for ICE Pambush 3 is available here, thanks to Ruck and team.

The four functioning entries are listed below with links to their descriptions (PDF format).  The paper describing the Robles entry was published in the proceedings of IEEE CIG 2009 (and can therefore be cited as such), the other descriptions were submitted as PDFs to accompany the software and can only be cited by their URLs.

Entry / Author(s)


ICE Pambush 3
Hiroshi Matsumoto, Takashi Ashida, Yuta Ozasa, Takashi Maruyama, and Ruck Thawonmas

Ritsumeikan University, Kyoto, Japan

L. Kelly, L. Dicken, J. Levine

University of Strathclyde, Glasgow, UK

David Robles

University of Essex, Colchester, UK

Ting-Chi Li, Yu-Chieh Lin, and Tsung-Che Chiang

National Taiwan Normal University, Taiwan


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 3, by Hiroshi Matsumoto, Takashi Ashida, Yuta Ozasa, Takashi Maruyama, and Ruck Thawonmas of the Intelligent Computer Entertainment Laboratory, Department of Human and Computer Intelligence, Ritsumeikan University

The team improved significantly on their IEEE CEC 2009 entry.  Like its predecessor (ICE Pambush 2) ICE Pambush 3 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.  It also seems to have a knack for sending the ghosts the wrong way. 

Some obvious improvements to make: it seems to have a fairly greedy short term strategy, and in many cases single food pills are left behind in the pursuit of more direct rewards, which makes life difficult towards the end of the level when these must be consumed in the absence of the relative security provided by the energiser pills.


ICE Pambush 3 can be seen in action on Youtube.


Two versions of the Ms. Pac Man competition will be run for IEEE CIG 2010 (and possibly other conferences in 2010 TBC), the screen-capture version (described here), and also modified version where people can enter either a Pac-Man agent or a team of ghosts - the Ms Pac-Man versus Ghost Team Competition.


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 and in IEEE CIG 2009.

[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