CSc 180 -- Fall 2007
2007 Tournament Results - "Lemming Chess"
The 6th Annual AI Strategy Game Contest took place Nov 5-7 2007.
The atmosphere was lively, and we had many spectators. The game this
year was called "Lemming Chess", which is similar to the old game of
"giveaway" chess, but on a smaller-sized board. Special thanks to
Dean Macari for taking the time to drop by and lend his support.
Also a big thank you to the ACM for raising student interest,
which led to a large number of student spectators. Thanks also to
the student volunteers who helped out and acted as proctors:
James Boheman, Brandon Urban, and Marcus Watstein, for helping the
tournament run so smoothly.
The competition was extremely close, and the final decision was not
known until the very last match was played. In the end, the clear
winner was Andrew Redden's "Judgment". In an extremely close 2nd
(after tiebreaks) was Hao Le's "KnowHao" over the 3rd place (and
cleverly-named) "Pwynage!!1" by Derek Goering. Judgment was written
in C++, using minimax with alpha-beta pruning, iterative deepening
and transposition tables with Zobrist hashing, to search between 12
and 22 plies deep in the 5 seconds allotted per move.
KnowHao, which searched less deep but had a more sophisticated
evaluation function, was one game away from being the first program
written in Java to win the competition, when it inexplicably threw
away a win (which it itself saw!) against the instructor's program,
"Fuzzy Wuzzy". Speaking of which, although Fuzzy Wuzzy made the
final round of 4, the top six programs were all so close that some
additional tiebreaking games were played. Fuzzy dropped to 6th
place - the lowest finish for the instructor in the six years that
the competition has been held.
Complete results are shown below, including results from prelims,
quarterfinals, semifinals, finals, and consolation rounds:
1. Judgment (Redden) 15-5 (F:4-2 S:3-1 Q:3-1 P:5-1 )
2. KnowHao (Le) 15-5 (F:3-3 S:3-1 Q:3-1 P:6-0 )
3. Pwnage!!1 (Goering) 12-8 (F:3-3 S:2-2 Q:2-2 P:5-1 )
4. XTremeRend (Burks) 8-6 ( S:1-3 Q:3-1 P:4-2 )
5. Bobby Fischer (Yee) 9-5 ( S:1-3 Q:3-1 P:5-1 )
6. Fuzzy Wuzzy (Gordon) 6-8 (F:2-4 S:2-2 Q:2-2 )
7. Old Tea Chess (Yeh) 6-4 ( Q:1-3 P:5-1 )
8. Maudite (Usher) 6-4 ( Q:1-3 P:5-1 )
9. Trixie (Lockwood) 4-6 ( Q:0-4 P:4-2 )
10. Int.Imbicile (Rodriguez) 7-3 ( P:3-3 c:4-0)
11. Prowler (Germino) 5-7 ( P:2-4 c:3-3)
12. Mindfield (Kalugin) 4-6 ( P:0-6 c:4-0)
13. CLemmingsRun (Lamb) 5-7 ( P:2-4 c:3-3)
14. Panda (Martin) 2-4 ( P:2-4 )
15. Peshka (Razdobarin) 2-8 ( P:0-6 c:2-2)
16. Munchya (Ambroff) 0-10 ( P:0-6 c:0-4)
- Surface (Zhao) 0-6 ( P:0-6 c:0-2)
As always, there were numerous interesting bugs and other twists
that happened during the tournament. One of the most promising
programs, Kyle Ambroff's "Munchya", used bitmaps to search 20 plies
deep from the opening position, but was not fully debugged and so
was unable to make it through entire games without erring badly.
Alex Rodriguez's "Intelligent Imbicile" was another strong program,
but Alex forgot to compile it with debug turned off, slowing it
down and reducing its search depth. Even Fuzzy Wuzzy fell victim
to an untimely bug when playing against Bobby Fischer... Fuzzy
had a win, saw it, then strangely played a losing move instead.
It could be said that this year, more than ever, debugging was
the key to victory.
The game itself proved a unique challenge. Even by the end of the
tournament, it was not clear that we had learned any clear strategies
for playing the game, besides simply looking ahead as far as possible.
We don't even know whether it is an advantage to move first or second.
So for now, the game is still far from being solved.