Komodo vs Smerdon: Update
A quick update on the details for the match:
- I’ve managed to convince two great guys to stream commentary of the match. They are Evgenij Miroshnichenko, who you may already know as arguably the world’s best chess commentator, and Andras Toth, a Hungarian IM living in Australia. They are doing this for free, which is pretty amazing.
- That means the easiest way to watch the match is to log on to Evgenij’s stream at www.twitch.tv/mironiusvulgaris , starting with the first three games on April 10 at 8:00 pm New York time (check your local day/time here), and then three more games the next day (check your local day/time here). It may also be streamed on chess.com, but they are not supporting the event so it’s safer to stick with Evgenij’s stream.
- Technical details about my opponent: Komodo 13 MCTS is the latest version of the world computer champion, featuring Monte Carlo tree search algorithms. The computer on which it runs is a 3970x AMD Threadripper with 32 cores, 64 GB memory, and an RTX 2080 ti GPU. I don’t know what any of this means, but Larry tells me that the total cost is US $7,000.
- I’ve been sourcing predictions from the chess community for the subsequent article, and it seems that most people are convinced that it’s going to be a massacre. They just can’t decide in which direction! Strong GMs like Emil Sutovsky, Matthew Sadler, Maksim Chigaev, Twan Burg, Evgeny Postny, Kiril Georgiev, Konstantin Maslak and Kevin Goh Wei Ming, along with correspondence and computer expert Nikolaos Ntirlis, predict a crushing win for Komodo (Sadler, Chigaev and Ntirlis expect that I won’t even score a single draw). But there are some on the side of humanity: GMs Nigel Short, Jon Levitt and Vadim Malakhatko, along with Evgenij, predict that I will come out victorious. Maxim Dlugy is one of the few grandmasters ‘on the fence’, predicting a 3-3 tie. Perhaps most of note, Larry Kaufman (Komodo’s creator) estimates that I will win 4-2.
Very interesting to read about the predictions before the match. Glad you proved most people wrong!
I have heard that earlier nets of LC0 were especially good at odds chess, possibly because LC0 hadn’t worked out the value of a knight by that stage! I find that quite interesting. I practiced against the more sophisticated recent versions of LC0 but these were much weaker in odds chess than other engines (Stockfish was only marginally better; Komodo was the best, even on a weak machine).
Always interested to check out new opening books by computer experts! 🙂
Nice to see a prediction come true. Looks like after the first game, you worked out how to handle this thing. Games 3,4, and 6 seemed one-sided; the extra N told straightforwardly. In Games 2 and 5, the only question was whether the engine could swindle a draw; you never looked in danger of losing. Game 5 required some fine technique, that march of your K along the enemy back rank was most impressive. Very well done!
Keep going like this, and you should get 3-0 tomorrow.
Go Dr Smerdon!
Great, I am lucky u play against engine, no one wants 🙂
Next time maybe u can try Lc0 with small net T70 which is like human play style. Btw I am opening book maker, I am focused on Top book, I can send to you if you response. My engines and books result :
https://lh3.googleusercontent.com/-CnWnGqXeTTE/XoZaBYghTqI/AAAAAAABi2Q/VKU9WPI6njoD06a8tuzC40N8DXO_se0OgCK8BGAsYHg/s0/2020-04-02.png
Good luck! 🙂