SF 280415 vs. K 9 & SF 6 vs. K 9, TC 600" + 3"

[majortom]

TC=600"+3"
Hash=128MB
Cores=1
Ponder=off
250 different opening positions (each plays twice with both colours)
GUI=cutechess-cli 0.6.0
CPU=Intel Core-i7 3960x @4100 MHz

[majortom]

After 438 games: -13 ELO against SF 6 & -7 ELO against SF 280415 for Komodo 9.

Code: Select all

  Program                          Elo    +   -   Games   Score   Av.Op.  Draws

1 Stockfish 6 64 POPCNT          : 3013   29  29   220    51.8 %   3000   59.1 %
2 Stockfish 280415 64 POPCNT     : 3007   31  31   218    50.9 %   3000   56.0 %
3 Komodo 9 64-bit                : 3000   21  21   438    48.6 %   3010   57.5 %

Code: Select all

Komodo 9 64-bit                  : 3000  438 (+ 87,=252,- 99), 48.6 %

Stockfish 280415 64 POPCNT             : 218 (+ 46,=122,- 50), 49.1 %
Stockfish 6 64 POPCNT                  : 220 (+ 41,=130,- 49), 48.2 %

[majortom]

After 628 games: -1 ELO against SF 6 & -11 ELO against SF 280415 for Komodo 9.

Code: Select all

  Program                          Elo    +   -   Games   Score   Av.Op.  Draws

1 Stockfish 280415 64 POPCNT     : 3011   25  25   314    51.6 %   3000   58.6 %
2 Stockfish 6 64 POPCNT          : 3001   24  24   314    50.2 %   3000   60.2 %
3 Komodo 9 64-bit                : 3000   17  17   628    49.1 %   3006   59.4 %

Code: Select all

Komodo 9 64-bit                  : 3000  628 (+122,=373,-133), 49.1 %

Stockfish 280415 64 POPCNT             : 314 (+ 60,=184,- 70), 48.4 %
Stockfish 6 64 POPCNT                  : 314 (+ 62,=189,- 63), 49.8 %

[majortom]

After 803 games: +5 ELO against SF 6 & -9 ELO against SF 280415 for Komodo 9.

Code: Select all

  Program                          Elo    +   -   Games   Score   Av.Op.  Draws

1 Stockfish 280415 64 POPCNT     : 3014   22  22   402    51.4 %   3005   57.0 %
2 Komodo 9 64-bit                : 3005   16  16   803    49.6 %   3007   57.7 %
3 Stockfish 6 64 POPCNT          : 3000   22  22   401    49.4 %   3005   58.4 %

Code: Select all

Komodo 9 64-bit                  : 3005  803 (+167,=463,-173), 49.6 %

Stockfish 280415 64 POPCNT             : 402 (+ 81,=229,- 92), 48.6 %
Stockfish 6 64 POPCNT                  : 401 (+ 86,=234,- 81), 50.6 %

[majortom]

Matches are finished: +2 ELO against SF 6 & -13 ELO against SF 280415 for Komodo 9.

Code: Select all

  Program                          Elo    +   -   Games   Score   Av.Op.  Draws

1 Stockfish 280415 64 POPCNT     : 3015   20  20   500    51.9 %   3002   58.2 %
2 Komodo 9 64-bit                : 3002   14  14  1000    49.2 %   3008   57.8 %
3 Stockfish 6 64 POPCNT          : 3000   20  20   500    49.7 %   3002   57.4 %

Code: Select all

Komodo 9 64-bit                   : 3002  1000 (+203,=578,-219), 49.2 %

Stockfish 280415 64 POPCNT                 500 (+ 95,=291,-114), 48.1 %
Stockfish 6 64 POPCNT                      500 (+108,=287,-105), 50.3 %

[shrapnel]

I don't see the point of playing on only 1 Core.
Its unfair to Komodo 9 as it scales better than Stockfish with increasing Cores and Hash.
Also, you should use Drawscore 0 for Komodo 9 as the AUTHORS themselves recommend using Drawscore 0 against Engines close in strength to Komodo 9.

[majortom]

I don't see the point of playing on only 1 Core.
Its unfair to Komodo 9 as it scales better than Stockfish with increasing Cores and Hash.

I don't see the point of playing with many cores.
It's unfair to SF as it scales worse than Komodo 9 with increasing cores and hash.

Also it spends much more the cpu time to provide 500 games with 6 cores (for example).

By the way, this is the 4-core results:
http://www.computerchess.org.uk/ccrl/404/
and the 1-core results:
http://www.inwoba.de/bayes.html

Also, you should use Drawscore 0 for Komodo 9 as the AUTHORS themselves recommend using Drawscore 0 against Engines close in strength to Komodo 9.

Do you think this is the good idea to make the different settings against each Komodo's opponent (if Komodo's authors will recommend that)?
I think it's not fair for other engines - each engine could have a few settings for different situations. I hope the default settings are recommended by authors too, because the default settings should be suitable for most opponents.

[dark_wizzie]

IT's just that, nobody runs chess seriously on only one core. Any mainstream CPU nowadays are 4 cores.

[Ozymandias]

nobody runs chess seriously on only one core

How about IDeA users? Just off the top of my head.

[dark_wizzie]

nobody runs chess seriously on only one core

How about IDeA users? Just off the top of my head.

I don't know? What other things can you think of off the top of your head?

Idea analysis can't be multithreaded? Why not?

I still think most people that run chess engines use four cores.