Pairwise Analysis of Chess Engine Move Selections Revisited

Discussion of chess software programming and technical issues.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
Adam Hair
Posts: 3201
Joined: Wed May 06, 2009 8:31 pm
Location: Fuquay-Varina, North Carolina

Pairwise Analysis of Chess Engine Move Selections Revisited

Post by Adam Hair » Sun Mar 04, 2012 8:13 pm

I have redone the study of the similarity in chess engine move selections that I first presented last year. This time around, I have tweaked the amount of time each engine had to think for each position (as found in the similarity tool of Don Dailey's) and included additional engines.

I would like to state that my motivation for this study is to initiate a discussion on how engines written by different authors could have a high level of move selection similarity. In particular, is it possible for one engine to match moves at a high percentage with another engine without an excessive amount of copying, whether it be literal or non-literal copying? In my studies, I have found most instances of high similarity involves an open source program, which leads me to be these are cases of engines derived from the open source engine. However, I have found two instances where an engine showed a high level of similarity with a closed source engine (though the closed source engine had an open source relative in both cases).

Ed Schröder is graciously hosting my newest round of comparisons at the Rebel website. Look on the bottom right for the link labelled "detecting clones". The name of that link comes from the possible usage of the similarity tool, in conjuction with other tests, to serve as a filter to catch possible clones. I list some precautions for this sort of usage of the similarity tool in the report found at that link.

User avatar
Rebel
Posts: 4705
Joined: Thu Aug 18, 2011 10:04 am

Re: Pairwise Analysis of Chess Engine Move Selections Revisi

Post by Rebel » Mon Mar 05, 2012 10:10 am

Thank you Adam for your kind words but most of all for your extensive study and work. It was not hard to recognize quality and quality needs a place and I feel honored to provide it.

As for content, it made me sad to see our hobby is so polluted, on the other hand it simply has amazed me that a simple software concept can reveal most of the clones. Kudo's to Don Dailey!

Don Dailey's similarity utility can serve as a tool for tournament directors to investigate participants BEFORE a tournament starts and this would save a lot of stress afterwards. Basically SIM03 has the potential to deliver us from the clones.

Each tournament director (upon his own interpretation) may decide a percentage what is acceptable and what is not.

What do you all think ?

Post Reply