I wrote a suicide checkers program - Suicidal Cake - in early 2005 and continued to improve it until some time in 2006, after which I found no incentive to develop it further - this checkers variant is not very popular, and I didn't publish my engine for CheckerBoard at the time. Around christmas 2009 I decided to publish Suicidal Cake at last, because I didn't want my work to rot on my harddisk. A lot of the development history of Suicidal Cake is documented in my computer checkers blog. This page sums up the matches I played with Suicidal Cake and links to pages for each match.
The first match: Suicidal Cake 1.11 - Roshi47 (March 2006)
The full match report on Suicidal Cake 1.11 - Roshi47
Shortly after writing Suicidal Cake, I played a few games with it against some well-known suicide checkers players like George Miller and Michael Taktikos. Suicidal Cake won every single game it ever played against them. I more or less stopped working on it after that, since there was no incentive to develop it further. However, onJanuary 18 2006, I received the following e-mail from Jonathan Schaeffer (author of Chinook):
I have a couple of students who built a suicide checkers program as a course project. They have tinkered around with it since then and are curious to know just how good it is. Obviously it kills all humans. I understand you have a strong suicide checkers program. Would you be willing to have your program play a match with them? Nothing at stake, except to satisfy some student's curiosity.
Suicidal Cake 1.11 won this match against Roshi 47 with 4-0 although it only had a 7-piece endgame database compared to the Canadian program's 8-piece database. My program had a 200'000 move opening book. After the match, I received the following e-mail from Jonathan Schaeffer:
To: Martin Fierz
From: Jonathan Schaeffer
Subject: Suicide Checkers
Date: Wed, 8 Mar 2006 16:48:02 -0700
X-Mailer: Apple Mail (2.746.2)
Congratulations! Experience trumps youth :) I think both
Mike and Frano are stunned at the result.
Now, that is something to be proud of :-)
The second match: Suicidal Cake 1.11 - SuicideKallisto 0-4 (April 2006)
The full match report on Suicidal Cake 1.11 - SuicideKallisto
Shortly after the match against Roshi47, I was challenged by a Russian programmer named Igor Korshunov. He had written a program called SuicideKallisto, and wanted to play against my program. I accepted his challenge with a lot of confidence - after all, I had just beaten two students of the Chinook research group with their 8-piece database. What could go wrong? Well, just about everything - Suicidal Cake was beaten fair and square, and its book and larger endgame databases never seemed to play a part. I'm not the kind of guy who gives up easily, so I went back to the keyboard, and tried to fix some evaluation issues I had spotted in this match. I also noticed that I hadn't enabled the lopsided parts of the 7-piece database (5-2 and 6-1 pieces), and fixed that bug. After some testing I had a new version of Suicidal Cake, and recomputed a small opening book (only ~20'000 moves, but less buggy) and challenged Igor again.
The third match: Suicidal Cake 1.13c - SuicideKallisto 1.5-2.5 (June 2006)
The full match report on Suicidal Cake 1.13c - SuicideKallisto
The third match was not such a clear affair - SuicideKallisto won its first two games, but then Suicidal Cake managed to win one, and the last game was drawn.
After the match I made some small improvements to Suicidal Cake's evaluation function, and computed the 8-piece endgame database. I then computed an opening book of about 50'000 moves, but never played another match with Suicidal Cake. Development stopped somewhere towards the end of 2006. At the end of 2009, I decided to make Suicidal Cake publicly available, and included it in the CheckerBoard distribution. It comes with the 4-piece endgame database, and no opening book. The 50'000-move opening book can be downloaded on the download page -- December 26, 2009