Komodo vs GM Smerdon knight odds match
Posted: Mon Mar 16, 2020 6:17 am
The first non-blitz match in history in which a GM will take knight odds has been scheduled! On the evenings of April 10 and 11 (starting at 8 p.m. New York time) we have scheduled a six game match between Komodo (regular, MCTS, or NN to be decided) and Grandmaster David Smerdon, FIDE 2508 Standard, 2562 Rapid, at the FIDE Rapid time control used in the World Rapid Championship (15' + 10"), on chess.com. David, born 1984, is a professor in Australia, and is the author of a new chess book, The Complete Chess Swindler, so he should presumably know something about how to avoid being swindled when a piece ahead! The match is being played specially for the ICGA Journal, and I'll be doing an article on it there with contribution by David. He will receive prize money of $100 per win and $25 per draw so he'll be motivated to play for the win while still having motivation to bail out to a draw once the win is gone. As with past knight odds matches, Komodo will play White alternating between b1 and g1 odds. It will have a small opening book to provide variety, which I'll share with him before the match. Hardware will be my new 32 core Threadripper 3970x. Komodo will use a sizable Contempt setting, but will still be playing for a draw unless/until Smerdon loses much of his advantage. The match is not part of a show on chess.com, but I expect to comment live during the games.
Readers may recall that my son IM Raymond Kaufman (FIDE 2281) played a similar four game match with Komodo (MCTS) in January, in which Komodo won the first two games, while the last two were time-forfeited by Komodo's refusal to move in tablebase drawn positions due to a problem that was fixed right after the match. If we count those forfeits as draws, and include Rapid games at knight odds against Jeremy Kane and FM Larry Gilden (including both shorter and longer rapid time limit games), Komodo's total score was 13.5 to 3.5 against an average FIDE rating of 2121 for a performance of 2355. So based on Smerdon's Rapid rating, Komodo's expected score would be 23.3%, so if Komodo draws half the games and loses the other half, that would be better than expected based on this calculation. But there are many unknowns.
What do you think? Can Komodo at least hold half the games?
Readers may recall that my son IM Raymond Kaufman (FIDE 2281) played a similar four game match with Komodo (MCTS) in January, in which Komodo won the first two games, while the last two were time-forfeited by Komodo's refusal to move in tablebase drawn positions due to a problem that was fixed right after the match. If we count those forfeits as draws, and include Rapid games at knight odds against Jeremy Kane and FM Larry Gilden (including both shorter and longer rapid time limit games), Komodo's total score was 13.5 to 3.5 against an average FIDE rating of 2121 for a performance of 2355. So based on Smerdon's Rapid rating, Komodo's expected score would be 23.3%, so if Komodo draws half the games and loses the other half, that would be better than expected based on this calculation. But there are many unknowns.
What do you think? Can Komodo at least hold half the games?