Next Komodo vs GM match - 2 pawns

[lkaufman]

I think Komodo use 4 seconds per move at Erenburg match

4* 1.000.000 kn = 4.000.000 (kn/move of Komodo at Erenburg match )

With 4.000.000 kn/move Komodo easily beat (3.5-0.5) a 2591 elo chess player (with 3 opening book, black pieces, no tablebases, low contempt)

I guess how many kn/move is needed for Komodo to play a 2600/2700 elo human level ( strong opening book, white/black pieces, tablebase on)

If we know this number ( kn/move) we easily know the human elo of Komodo with strong hardwares

Not really. The problem is that although we know pretty accurately the elo value of doubling the thinking time for engine vs. engine rating lists, we don't really know that number for engine vs. human matches. I think 3/4 of the engine vs engine number has been a pretty good guide over the past 25 years, but it may be way off for levels around 3000.

[mehmet karaman]

I think Komodo use 4 seconds per move at Erenburg match

4* 1.000.000 kn = 4.000.000 (kn/move of Komodo at Erenburg match )

With 4.000.000 kn/move Komodo easily beat (3.5-0.5) a 2591 elo chess player (with 3 opening book, black pieces, no tablebases, low contempt)

I guess how many kn/move is needed for Komodo to play a 2600/2700 elo human level ( strong opening book, white/black pieces, tablebase on)

If we know this number ( kn/move) we easily know the human elo of Komodo with strong hardwares

The nodes/s of Komodo at Erenburg match is 1.36 milion
5.5 million nodes/move is enought to reach a 2930 elo.
If Komodo has a good opening book and higher contemp it easily beat Erenburg with a 4-0 score

Time odds match is a good inducator to learn the human rating of Komodo

[Frank Brenner]

I dont like handicap games where the GM gets some more material very much, because when watching to these games i observe, that the engine (Komodo) is playing very defensive, very drawish:
When there is a chance to simplify the position (with at least a +0,01 better evaluation) komodo executes this simplification.


The contempt does not help in this situation.


I like Kai's idea of of allowing the GM to take back as many moves as he want.
For example: Komodo plays with full strength with 15 Seconds/move
The GM has 3 hours time to finally win the game, while he can take back as many moves as he want.
Of course the GM is not allowed to look at the evaluation.

[lkaufman]

I dont like handicap games where the GM gets some more material very much, because when watching to these games i observe, that the engine (Komodo) is playing very defensive, very drawish:
When there is a chance to simplify the position (with at least a +0,01 better evaluation) komodo executes this simplification.


The contempt does not help in this situation.


I like Kai's idea of of allowing the GM to take back as many moves as he want.
For example: Komodo plays with full strength with 15 Seconds/move
The GM has 3 hours time to finally win the game, while he can take back as many moves as he want.
Of course the GM is not allowed to look at the evaluation.

I think we'll eventually try something like this at least once. We really don't have much basis to evaluate how big an advantage these takebacks are. I'll see if some GMs are interested.

[dusando]

The most interesting match to me was the one with the 30x time handicap. It would be even closer to normal chess, if the handicap was done via just slowing down the PC 30 times instead of giving it a different timecontrol. Ponder on, as the GM would also be able to think longer. And, it would be easier to operate. There are simple tools to throttle CPU usage, I've just installed something called BES 1.6.2 (http://mion.faireal.net/BES/) and am running Stockfish slowed down by 99% on my sublaptop, showing some 2000 nodes per second. This thing is still quite strong, but already beatable by many humans.

[lkaufman]

The most interesting match to me was the one with the 30x time handicap. It would be even closer to normal chess, if the handicap was done via just slowing down the PC 30 times instead of giving it a different timecontrol. Ponder on, as the GM would also be able to think longer. And, it would be easier to operate. There are simple tools to throttle CPU usage, I've just installed something called BES 1.6.2 (http://mion.faireal.net/BES/) and am running Stockfish slowed down by 99% on my sublaptop, showing some 2000 nodes per second. This thing is still quite strong, but already beatable by many humans.

If we did this it would be much more tiring on the GM as the games would be twice as long. Given that it is not practical for us to play only one game per day, I think we did it the right way.

[dusando]

The most interesting match to me was the one with the 30x time handicap. It would be even closer to normal chess, if the handicap was done via just slowing down the PC 30 times instead of giving it a different timecontrol. Ponder on, as the GM would also be able to think longer. And, it would be easier to operate. There are simple tools to throttle CPU usage, I've just installed something called BES 1.6.2 (http://mion.faireal.net/BES/) and am running Stockfish slowed down by 99% on my sublaptop, showing some 2000 nodes per second. This thing is still quite strong, but already beatable by many humans.

If we did this it would be much more tiring on the GM as the games would be twice as long. Given that it is not practical for us to play only one game per day, I think we did it the right way.

Yes, the timecontrol should be faster, then. I'd like to see how the top engines perform on a very slow hardware (simulated by the CPU throttler) with all other conditions being equal. Perhaps a GM strength human opponent is not really needed at 1% of an Intel Atom speed.

[lkaufman]

Commentary on the games will be by GM Simon Williams on chess.com.

I had Komodo analyze each handicap out to 35 plies, and it seems that e2 + f2 is the worst for Komodo, even worse than the c2 + f2 with which we lost badly to GM Petr Neuman. The other three come out pretty close to each other.

[Vinvin]

...
So this time (July 15 and 16, noon and 5 pm Eastern time each day, games on chess.com) ...

Is it "15 and 16" or "16 and 17" ?
https://fr.chess.com/tv

[lkaufman]

...
So this time (July 15 and 16, noon and 5 pm Eastern time each day, games on chess.com) ...

Is it "15 and 16" or "16 and 17" ?
https://fr.chess.com/tv

OOPS! It's 16 and 17. I must have relied on my memory, which is not perfect at age 68. Sorry.