Komodo 9.0

[TShackel]

Hi all,

I just wanted to say that I'm done with 80 games of testing at longer time control games (90' + 20" increment...avg of 3.5 hours per game utillizing 4 cpus on my haswell i7) and Komodo 9.0 is beating Stockfish April 12th development version on the rating list. Stockfish beat komodo twice, but komodo beat stockfish 5 times. That propelled its rating upward. Komodo got beat by gull once, but Komodo has fought back with several unanswered wins to be the clear winner in the match. So I've been having very good tournament results for Komodo 9.0.

Komodo sees tactics and plans that stockfish was oblivious to. Very rarely is stockfish proven wrong about a tactic. So it's weird to see this new komodo engine say to stockfish "you didn't play my anticipated move, so now I'm going to crush you" and that's virtually every time. Komodo ruthlessly capitalized on stokcfish's mistakes. And its weird for me to now think that stockfish can make some mistakes. Because for a while, when competing with Komodo 8.0 it appeared to be the clear leader and seemed that stockfish approached near perfection. No one can prove stockfish wrong in that case. Well I'm now learning stockfish is not invincable and does have evaluation imperfections and tactical oversights. I think this will be good for stockfish to face a strong competitor like Komodo 9.0 to point out the weak points of stockfish's eval and search.

I've also seen Komodo 9.0 sacrifice pawns, and pieces several times which was very interesting. It has very good king safety values. And I will post a game soon where evaluations were all over the place with Komodo and Stockfish because both king's were not in great shape and there were some pretty wild lines.

Has anyone else had good results with Komodo 9.0?

Sincerely,

Tim.

[shrapnel]

Interesting results indeed !
I have always felt that Komodo was a better Engine than Stockfish, but not always proven by results.
What was the Drawscore used in your Matches ?
Any other changes from Default Settings ?
Regards

[TShackel]

Interesting results indeed !
I have always felt that Komodo was a better Engine than Stockfish, but not always proven by results.
What was the Drawscore used in your Matches ?
Any other changes from Default Settings ?
Regards

Everything was defaults and -7 drawscore. I left it at -7 because houdini was involved in the tournament as well and it is a bit weaker than the big two.

I'm kind of excited about the results so far. But the sample size is only 80 games so far since it is a long time control. And more games would be required. CEGT usually posts their long time control engine results when they have at least 300+ games. So I have a way to go.

Sincerely,

Tim.

[shrapnel]

Hmm...Default Score -7, with Stockfish in the field, is really amazing !
You must be playing some REALLY long Matches indeed !

[JJJ]

So Komodo 9 still the strongest at very LTC. It's nice, can't wait for TCEC.

[Dann Corbit]

So Komodo 9 still the strongest at very LTC. It's nice, can't wait for TCEC.

Stockfish is a moving target.
It will be stronger tomorrow that it is today.
And stronger the day after that.
A commercial software really can't have daily releases.

So the Komodo team will have a difficult time to keep up.

However, I must admit that they have performed admirably so far.

[JJJ]

Komodo is also a moving target. It is improving almost everyday to quote Larry Kaufman. That is more than a year, maybe two that Komodo is catching Stockfish. So, they don't make less progress.

[Dann Corbit]

Komodo is also a moving target. It is improving almost everyday to quote Larry Kaufman. That is more than a year, maybe two that Komodo is catching Stockfish. So, they don't make less progress.

They make more progress, less progress, or the same progress.
I guess that it will be very hard to sustain the same progress because the Stockfish team has very large resources. But not impossible if they work hard.

But Stockfish can make releases as often as they like and commercial products usually don't do that.

I suppose that Larry and Mark could time a release of Komodo for just before the TCEC, which would mitigate this difference.

[JJJ]

They already say they will update Komodo more often. So I think they can still catch up for the moment, even being fewer on it. As long they work as much as they do right now.

[Dann Corbit]

I do expect that these two will be the dominant engines unless the Gull author finds time to improve it.

He seems to have dropped steam on the project for some reason.