University homepage Suomenkielinen versio puuttuu Inte på svenska In english
University of Helsinki Department of Computer Science
 

Department of Computer Science

582329 AI for Games (3 cr) Autumn 2009

Final Notes

Hi Tomi,

Today I went into town to buy some presents, but I still had time in the evening to look at the competition once more.

I checked all the 22 teams once again, and discovered a few "cheating" statements here and there, which I removed. I also solved some endless loops in a few teams' codes, but not much. The new competition module is enclosed; perhaps you would like to place it online next to the sheets.

After doing this, I ran some more quick competitions to see if we really selected the best team as winner.

First, I pitted each team against SAURON, the winner of yesterday, for 5 minutes. The results were that SAURON defeated all of them with a good margin, except Nameless (who was in the final competition yesterday) and Stoneskin (who was disqualified yesterday, but I now used a version from which the offending statements were removed). Neverwin Fights (who was in the final competition yesterday) and Holy Stephens did reasonably well against SAURON, but certainly could not win.

From this I deduced that probably the final round yesterday should have been between SAURON, Nameless, and the cleansed Stoneskin. I then ran this competition: each team, twice against each other team, once as red and once as blue. The results were as follows:

SAURON-Stoneskin: 348-267 and 374-296
SAURON-Nameless: 230-235 and 287-196
Stoneskin-Nameless: 422-101 and 221-276

What I conclude is the following: SAURON wins three of the four fights that it is in, and the one it does not win (against Nameless) ends very close (only 5 points difference). SAURON thus seems to be outperforming both Stoneskin and Nameless.

Stoneskin wins once against Nameless, and loses once. Its victory is devastating, its loss clear but reasonable. Thus it seems to me that Stoneskin is stronger than Nameless, but Nameless does have the abililty to defeat Stoneskin. The question remains why they give such different results in these matches. I think that the explanation is that they both try to create a situation in which they have a clear advantage, and then maintain that situation. The team that manages to create that situation first, which seems partly chance, defeats the other team with a wide margin. I am not certain that this is actually what happens, but it could very well be.

After this I tried a few more combinations. I found that all three teams which I now matched against each other could defeat Chuck Norris, the best team from Maastricht University, but that Chuck Norris was able to defeat all the 19 other teams.

Finally, I decided yesterday that the total of the scores of the final matches would determine the winner. The total scores are:
SAURON: 1239
Stoneskin: 1206
Nameless: 808

Therefore SAURON is confirmed in its win from yesterday. However, all three teams deserve hearty congratulations for having created such strong AI.

If you want to use the text above for the website or something like that, be my guest.

It was fun, and I hope to return sometime in the future to do something like this again.

See you, Pieter.