Checkers News

Welcome to the homepage of CheckerBoard, one of the most powerful checkers programs for the windows platform. This page contains the latest news on CheckerBoard and related programs; use the menu on the left to download the program and to learn more about CheckerBoard, its main engine Cake, opening books and endgame databases etc.

Cake 1.89f is another small improvement in playing strength and fixes a few bugs in the engine. Download Cake 1.89f - unzip the file, and save cake_189f.dll, book.bin and the help file (cake189.htm) in the "engines" subdirectory of your CheckerBoard directory; save egdb64.dll in the CheckerBoard directory; and finally select cake_189f.dll in CheckerBoard to start using it.
This download includes a new opening book of ~150'000 moves, computed automatically by Cake 1.89f in about 3 weeks. It is smaller than the previous big opening book, but since the engine is so much better, it is more accurate. I have written something about the making of Cake 1.89f, and its playing strength.

Cake 1.89d is another improvement of my machine-learned engine, finally extending the evaluation function to a huge number of parameters (about 200'000). A special thank you goes to Ed Gilbert who explained all the ideas related to this type of pattern-based evaluation to me. Download Cake 1.89d - unzip the file, and save cake_189d.dll and the help file in the "engines" subdirectory of your CheckerBoard directory; save egdb64.dll in the CheckerBoard directory; and finally select cake_189d.dll in CheckerBoard to start using it.
I have written something about the making of Cake 1.89d, and its playing strength.

Ed Gilbert has released another improvement to KingsRow,KingsRow 1.19c.

Ed Gilbert published an improved version of Kingsrow in August 2020, KingsRow 1.19b is even more powerful than KingsRow 1.18f (and he has dashed any hopes of mine of being able to catch up to KingsRow with that new version...).

Jon Kreuzer recently updated his checkers program with neural networks, creating GuiNN Checkers 2.04; likely inspired by the huge sucess that neural networks have had in chess programming. The ever amazing Ed Gilbert added his database driver to Jon's program, creating GuiNN Checkers 2.05.

As far as I can see, GuiNN is still slightly weaker than Cake 1.88d, but only by a very small margin. The new versions of KingsRow (1.19b), Cake (1.88d) and GuiNN (2.05) are far better than anything that was around when Ed started using machine learning for checkers a bit more than two years ago, and after a long period with no or only very little improvement in checkers engines, real (and very large) progress has been made - nice!

Cake 1.88d is another improvement on Cake 1.88. Download Cake 1.88d - unzip the file, and save cake_188d.dll and the help file in the "engines" subdirectory of your CheckerBoard directory, and select cake_188d.dll in CheckerBoard to start using it.

Cake 1.88 is a small improvement on Cake 1.87. It's still not in the league of the machine-learned KingsRow, but it's a bit closer than it used to be. It also fixes a bug in the search which could cause the endgame database to malfunction. Download Cake 1.88 - unzip the file, and save cake_188.dll and the help file in the "engines" subdirectory of your CheckerBoard directory.

After making a lot of evaluation improvements in Cake 1.86, I focused a bit more on the search, and optimized that part of Cake. The result is Cake 1.87, which is another clear improvement. It's still not in the league of the machine-learned KingsRow, but it's far closer than it used to be. It also fixes a bug in the book access code that could make Cake 1.86 crash when used with the opening book. Download Cake 1.87 - unzip the file, and save cake_187.dll and the help file in the "engines" subdirectory of your CheckerBoard directory. Alternatively, you can download the entire current CheckerBoard setup at Ed Gilbert's English checkers website. His latest installers include Cake 1.87.

Since Cake 1.86 is far stronger than previous versions of Cake, I thought I would write up how it came to be, and what type of machine learning was used for it. You can read more on the making of Cake 1.86 in case you are interested how this works.

After reviewing Ed Gilbert's amazing machine-learning-based KingsRow 1.18f, I decided to use similar ideas to improve Cake (logistic regression to automatically improve the evaluation function). The result is Cake 1.86, the first new version of Cake since 8 years (!), far stronger than its predecessor 1.85 - but still very clearly inferior to the new KingsRow 1.18f.
Download Cake 1.86 - unzip the file, and save cake_186.dll in the "engines" subdirectory of your CheckerBoard directory.

Ed Gilbert improved his checkers engine KingsRow massively using machine learning techniques. There is an interesting article in Bob Newell's Checker Maven on this. Congratulations to Ed for once again shifting the boundaries of what is possible in checkers programming.

I'm very late reporting this - the source code of CheckerBoard can now be found in Ed Gilbert's GitHub repository. Ed has also cleaned up the code and fixed many bugs - thanks a lot!

A small bugfix for the position search in CheckerBoard. Download the CheckerBoard setup package for 32-bit windows or for 64-bit windows!

I just posted Checkers Tutor 2.1 on the Android market. It is probably the most sophisticated checkers application for the Android platform; in particular, it has levels ranging smoothly from awful to extremely good, so that it will be an appropriate partner for players of all levels. Read more about it on my Checkers Tutor page, and on my checkers programming blog and then go and buy it on the Android market for a lousy dollar (and support your favorite checkers programmer while doing so)!

I finally managed to finish my Java checkers program jCheckers; it has already been tested on Linux and Mac, and seems to deliver on Java's promise of "write once, run anywhere". You can learn more about the program and download it on the jCheckers page.