is one of today's strongest Othello programs. Besides the powerful hardware on which it is running and the efficient implementation of standard game-tree searching techniques, the program's considerable playing strength is mainly due to several new approaches for the construction of evaluation features, their combination, selective search, and learning from previous games which I have investigated in my Ph.D. thesis and considerably improved while working at NECI.

Download Logistello's book skeleton (i.e. (self-)played games) for studying openings, training evaluation parameters, or just improving your bot's book. Games are stored as ASCII move sequences annotated with the final game result in black's view. All ~37K lines are at least 24-ply WLD correct. The file is gzip'ed. Click here if you prefer the WTHOR format.

Logistello's source code is now available. Last update: November/4/2002: Stephane Nicolet found a bug in the MPC code - fixed.
Logistello's X-window GUI source code.

README file:

This is the Logistello source directory which is licensed under the GPL. I worked on it from 1992 till 1998. As a result, it's not a pretty sight: a zillion tools for handling different game file formats, computing statistics, solving logistic and linear regression systems, book learning, running tournaments etc. Many of them are now obsolete. Some of the code is hard to read - some comments and variable names are even in German (Brett=board,Zug=move,Wert=value). I hardly can tell which file is actually used anymore. The makefiles can be a starting point for exploring the source code. The current Logistello version is generated by "make ../oplayl". I have not included the training data, nor the opening book, nor the pattern and Multi-ProbCut tables. Although the program compiles, it cannot be used to play Othello out of the box. All the software for creating missing tables is included, however. What's missing is the 17 million training positions - which are not that hard to generate nowadays.

The purpose of this release is to present the implementation details of a strong game-playing program. My hope is that it is useful for designing coming generations of even stronger programs.

- Michael Buro, November 2002.

Program overview

Development History

Recent Improvements

Tournament Results (last updated: January/28/98)

Murakami vs. Logistello &
Workshop on Game-Tree Search

(last update: August/20/97 Match Report!)

Princeton I: Logistello wins 22 games in a row

last modified on ; you are visitor # since Dec/18/1995
to homepage of LOG's creator