Status Quo 2016-1 , Stockfish - Komodo

[beram]

Great tournament ...... but why not against Sf7 and rather against a dev version ...... the dev version could have bugs which we don't know so far

thanks again for running this

Great work Clemens !

What would be interesting is that You play same testmatch under diminishing time controle. From 60 to 30 to 15 to 7.30 min for a game
Just to See what happens
Is the 24 core setup the most important or is iT the timecontrol that made Komodo win this match

Greetings Bram

[JJJ]

very similar result to TCEC anyway.

[APassionForCriminalJustic]

very similar result to TCEC anyway.

It is similar, but there is still a big time difference between 60 minutes with no increment to 180 minute\30 second increment games. This is a solid win for Komodo no doubt. I would expect even better results if the time control continues to rise.

[arunsoorya1309]

somehow the scaling of SF is still not that good as that of K9.3 in > 20 cores ........ 24 cores is a lot and not useful for the average person ....... however its a very good performance from k9.3 which was released few months ago versus the latest SF ....... i believe 20 cores will make SF stronger than K

[Leto]

Are we sure this is just a scaling issue? I'm starting to think Komodo is just stronger at longer time controls.

Didn't someone post a thread here recently that showed that when given a very long time to solve positions Komodo is actually even more tactical than the latest Stockfish? That might be part of the reason Komodo seems to win these long time control matches.

[gordonr]

Are we sure this is just a scaling issue? I'm starting to think Komodo is just stronger at longer time controls.

I'm now running a tournament at 4 hours + 30 sec increment on a 8 CPU i7-5960X @ 4.4Ghz. How close to the 'TCEC final' conditions is this in terms of the extra hour offsetting some of the hardware difference?

As well as Stockfish and Komodo, I decided to include a 3rd engine in the tournament as I think it's good to see engine ability against a range of opponents. This 3rd engine is currently Houdini but I may swap it out for Fire, Gull, etc. after so many games and keep rotating this 3rd player. All engines will use 0 contempt.

At this time control I'll never have enough games to be statistically significant but I want to see high quality games and will be interested to see how Stockfish-7-beta-1 and Komodo 9.3 win/lose any games.

Finally, thanks to Clemens for running his excellent tournament. And great broadcasting too.

[ernest]

It is curious that when score = 0.5...

Then you have you conclude, as I did:
... probably those ChessBase GUI error-bars are just based on wrong (and idiotic) theory...
Or programmed wrong !

[S.Taylor]

I get the impression that many of the draws, were not perfectly drawn, but that neither engine had the know-how what to do, to further the struggle, at a certain point in the game, towards the end.

Amongst humans, Karlsen for example, displayed much more ingenuity in ways that computers do not even try to do. If computers DID try to, the results might be even more staggering than what they are, and than what humans can do too.

[cdani]

Are we sure this is just a scaling issue? I'm starting to think Komodo is just stronger at longer time controls.

My view is that at some point, with enough time the best engines solve probably 99% of tactical tricks, so what rules the results in this situation is the knowledge.

So I will say that chess is a tactical game in the leaves, and a positional game in the core. I vote that with more computer power we will see a big raise in the importance of the evaluation function.

[Laskos]

Are we sure this is just a scaling issue? I'm starting to think Komodo is just stronger at longer time controls.

My view is that at some point, with enough time the best engines solve probably 99% of tactical tricks, so what rules the results in this situation is the knowledge.

So I will say that chess is a tactical game in the leaves, and a positional game in the core. I vote that with more computer power we will see a big raise in the importance of the evaluation function.

Interesting Dani, I am pretty much of the same opinion. The decisiveness of tactics in (computer) chess seems to decrease with strength. There seem to be less and less tactical shots the higher is the level of play. TCEC level chess seems very different from blitz of 10 years ago engines. Yes, I guess the importance of evaluation will increase in the future.