Next Komodo vs human match

[lkaufman]

This Tuesday, Sep 10 (Wednesday didn't work this week), at 1 pm Eastern time, Komodo will have a rematch with FM Larry Gilden at knight odds on chess.com (normal chess, not 960). Gilden beat Komodo 3 to 0 a couple years ago with knight odds at 45' + 15", but this will be at the World Rapid Championship time limit (15' + 10"), the same one I have been playing Komodo at in 960 knight odds. Larry Gilden isn't exactly a kid, now age 77, but you would be mistaken to assume he can't play decent chess anymore. Just a few weeks ago he won clear first place and $2,500 in the Premiere (under 2250) section of the Washington International, and currently plays, teaches, and studies chess full time. Sixty years ago he was U.S. Junior champion, was ranked in the top 12 overall in the U.S. in the early 1960s, was top scorer in the Student Olympiad in 1967, and played in the U.S. Championship in the 1970s. There is currently only one American player older than Gilden and rated above him (IM Anthony Saidy, who would surely have gotten the GM title by more modern standards). His USCF rating is about a hundred elo below my own, but playing normal chess knight odds (alternating b1,g1, four games) is much more favorable for the human than the 960 knight odds that I have been playing, so I would think his chances against Komodo are pretty good. We'll use Komodo MCTS with a Contempt setting this time. My long-term goal is to see Komodo eventually win or draw a match giving knight odds at this World Rapid TC to a 2500+ FIDE GM. Although Gilden is a long way from this level, this should at least shed some light on whether that goal is at all plausible or just a pipe dream.

[Marcus9]

I'm very excited for this match. Seeing Komodo someday that will give a knight odd vs GM at rapid TC is also my dream

[lkaufman]

I'm very excited for this match. Seeing Komodo someday that will give a knight odd vs GM at rapid TC is also my dream

I should also mention that Gilden plays quite regularly in both blitz and rapid tournaments (as well as standard), and in his younger days was a blitz sparring partner for Bobby Fischer and also played blitz with many of the top Soviet players of the time, so I don't think the rapid time limit will be too hard for him. Some older GMs have even claimed that rapid is the most favorable time limit for older players, slow enough that reflex speed and vision don't matter, fast enough to minimize the fatigue factor. I'm not convinced of this myself, at least in my own case I'm relatively stronger with more time, but I can't say that they are wrong in general. In the 1970's we contested a 20 game rapid match, which I won by a single game.

[carldaman]

I'm very excited for this match. Seeing Komodo someday that will give a knight odd vs GM at rapid TC is also my dream

I should also mention that Gilden plays quite regularly in both blitz and rapid tournaments (as well as standard), and in his younger days was a blitz sparring partner for Bobby Fischer and also played blitz with many of the top Soviet players of the time, so I don't think the rapid time limit will be too hard for him. Some older GMs have even claimed that rapid is the most favorable time limit for older players, slow enough that reflex speed and vision don't matter, fast enough to minimize the fatigue factor. I'm not convinced of this myself, at least in my own case I'm relatively stronger with more time, but I can't say that they are wrong in general. In the 1970's we contested a 20 game rapid match, which I won by a single game.

I concur with the older GMs' view, from a club player's perspective, especially when one starts with 20 minutes or so - there is just enough time to think and also not enough to get overly tired.

[lkaufman]

starting shortly

[JJJ]

Is lczero better than Komodo at knight odds ? of course I m talking about a 10x net.

[lkaufman]

First game was a spectacular sacrificial win by Komodo.

[lkaufman]

Is lczero better than Komodo at knight odds ? of course I m talking about a 10x net.

Lc0 11248 (or 58) might be better, as it was trained without resigning, but later nets are awful at knight odds.

[lkaufman]

Second game had to be restarted due to a wrong move entry (Gilden is playing on a real chessboard, I'm inputting his moves on chess.com), but within a few moves he had to resign. He could have traded queens earlier and perhaps a draw was a likely result if he had done so.

[lkaufman]

Third game was also won by Komodo. Gilden got the queens off at the price of two pawns, lost a third, and couldn't hold the endgame with knight for three pawns.