Some Komodo 9 results

[JJJ]

Nice !
So, wich time control is needed to Komodo 9 to be better than Stockfish 6 or stockfish Dev ?
Would you bet on 5+3 ? Maybe more against Stockfish Dev ?

[lkaufman]

Nice !
So, wich time control is needed to Komodo 9 to be better than Stockfish 6 or stockfish Dev ?
Would you bet on 5+3 ? Maybe more against Stockfish Dev ?

I have some new results now to add. On 1 cpu at 2' + 1" against the latest SF (April 18) K9 lost by 1043 to 957 which is 15 elo. Since on the same test K9 won by 6 elo against SF6, the indicated gain for SF is 21 elo since SF6, which agrees pretty well with the claimed 25.
On 8 cores K9 is leading against SF6 at 1' + .5" by a whopping 32 elo after almost 500 games. It seems that K9 benefits much more than SF6 going from 4 cores to 8.
On 4 cores K9 lost very narrowly to SF6 at 2' + 1", 495 to 505, which is an insignificant minus 3.5 elo.
So against SF6, even at very fast time controls like 2' + 1" K9 has a slight overall lead if we combine 2 and 4 core results, and is also plus on one core and way ahead on 8. So I would bet on k9 against SF6 at any level not less than 2' + 1" and on any number of cores.
Against April 18 SF, k9 is an underdog (undercanine?) at blitz levels. I'm starting a single-core run tonite at ten times the previous level, so 20' + 10", to see if this is enough time for Komodo to come out ahead. Naturally it will take a while to get a statistically meaningful number of games.

[JJJ]

Thanks Larry, very nice to hear for me. I have 4 CPU to use. So when I ll buy Komodo 9, I might try some test. I m curious to see how Komodo 9 play without opening book. Wich move he prefers. I prefer the way Komodo play than Stockfish, because it's fits me better.

[lkaufman]

Update on the 20 min. plus 10 second single-core match vs Stockfish April 18. After 412 games score is 224.5 - 187.5 which is plus 31 elo, well beyond error margin now.

[JJJ]

And just on 1CPU...

[shrapnel]

It seems that K9 benefits much more than SF6 going from 4 cores to 8.

Now THAT'S definitely interesting for ME !....will soon find out if that's true or not.

[APassionForCriminalJustic]

It seems that K9 benefits much more than SF6 going from 4 cores to 8.

Now THAT'S definitely interesting for ME !....will soon find out if that's true or not.

Of course Komodo scales better as the core count increases. Definitely on my 20 core it is a beast; it certainly scales better than SF6 by one million miles (but SF 6's SMP really was not good anyways). Let me know what you find Shrapnel.

Komodo will without question be the engine to beat in the up and coming TCEC tournament. Stockfish development has some definite work to do despite all of the recent multiple-threading patches that have dully been added.

[JJJ]

Stockfish Dev Scales a lot better than Stockfish 6. So let us know if Komodo 9 scales better than Stockfish Dev instead of Stockfish 6.

[APassionForCriminalJustic]

Stockfish Dev Scales a lot better than Stockfish 6. So let us know if Komodo 9 scales better than Stockfish Dev instead of Stockfish 6.

Good, fair point.

[Modern Times]

Thanks Larry, very nice to hear for me. I have 4 CPU to use. So when I ll buy Komodo 9, I might try some test. I m curious to see how Komodo 9 play without opening book. Wich move he prefers. I prefer the way Komodo play than Stockfish, because it's fits me better.

Try chess960, Komodo plays it now