Saturday, May 5, 2012
Chess Engine Grand Tournament (CEGT) is a well-respected ratings organization. Founded in 2005, the group tests chess engines at 40 moves/4 minutes, 40 moves/20 minutes, and 40 moves/120 minutes (ponder off), as well as 40 moves/20 minutes (ponder on). Also, it tests 32-bit and 64-bit versions of engines, as well as single cpu and multiple cpu versions of engines.
All members use a benchmark (a compile of Crafty by Bryan Hoffman) to determine tournament time controls (as played on their own computers) in order to approximate the effective engine thinking time for the reference computer. All engines use common books or starting positions.4 men and 5 men endgame tablebases and bitbases are used.
The CEGT's rating lists are very relevant. The lists cover a wide range of conditions, and many newer engines have been tested. Statistically significant results are generated fairly quickly, especially for higher rated engines.
The precision of the CEGT's results are affected by the use of multiple, non-identical computers and various books/starting positions. The focus on testing many different engines under various conditions limit the number of games that can be conducted for a particular engine, thus lowering the accuracy of its estimated rating.
at 12:14 AM