Komodo 5.1 MP results

[lkaufman]

Here are some early results for Komodo 5.1 MP. All tests on Windows sse4 machines using Fritz 11 interface, equal number of cores for Komodo and opposing engines.
First, it should be noted that the underlying SP engine is slightly stronger than Komodo 5 but clearly weaker than Komodo CCT. This has been verified by multiple tests including independent testing.
Against Critter 1.6a, at 3' + 2" on 4 cores I got a tied score after 69 games, but at 5'+3" Komodo currently leads by six games out of 45. On 12 cores at 5' + 3" I got a 25 to 25 tie.
Against Stockfish 3.0 at 3'+ 2" on 12 cores Komodo won by 27.5 to 22.5.
Against Deep Rybka 4 on 12 cores at 5' + 3" so far Komodo leads 5 to 2.
Against Houdini 3 on 4 cores at 5' + 3" Komodo is 5 games down after 52 games (-33 elo).

So samples are still quite small, but it appears likely that Komodo 5.1 will be the number 2 MP engine. Updates later.

[Nelson Hernandez]

Congrats on this achievement and best wishes to you, Larry!

[lkaufman]

Thanks, Nelson!

Some updated results: 4 core critter match at 5' + 3" finished with Komodo winning by 35.5 to 23.5 (match stopped to begin testing an alternate version so a final choice can be made for the release). Against Deep Rybka 4 at same time limit but 12 core match, Komodo now leads by 13.5 to 8.5. Against Houdini 3 at same time limit on 4 cores Komodo is losing by 29.5 to 39.5.

[geots]

Here are some early results for Komodo 5.1 MP. All tests on Windows sse4 machines using Fritz 11 interface, equal number of cores for Komodo and opposing engines.
First, it should be noted that the underlying SP engine is slightly stronger than Komodo 5 but clearly weaker than Komodo CCT. This has been verified by multiple tests including independent testing.
Against Critter 1.6a, at 3' + 2" on 4 cores I got a tied score after 69 games, but at 5'+3" Komodo currently leads by six games out of 45. On 12 cores at 5' + 3" I got a 25 to 25 tie.
Against Stockfish 3.0 at 3'+ 2" on 12 cores Komodo won by 27.5 to 22.5.
Against Deep Rybka 4 on 12 cores at 5' + 3" so far Komodo leads 5 to 2.
Against Houdini 3 on 4 cores at 5' + 3" Komodo is 5 games down after 52 games (-33 elo).

So samples are still quite small, but it appears likely that Komodo 5.1 will be the number 2 MP engine. Updates later.

Larry, you have completely, entirely, all-encompassed, without a doubt- 100% left me totally lost!


george

[lkaufman]

Sorry, George, but I can't see any part of my post that is not clear. Can you please clarify what you find confusing so I can clarify it?

Larry

[lkaufman]

Updated results:

Four cores each, 5' + 3:

35.5 - 23.5 vs. Critter 1.6a.
38-53 vs. Houdini 3.

So apparently about midway in strength between Critter 1.6a and Houdini 3 on 4 cores.

On 12 cores:

27.5 - 22.5 vs. Stockfish 3 (3'+2")
23.5 - 20.5 vs. Deep Rybka 4 (5' + 3")

[lkaufman]

We have decided to go with a newer beta version for our Monday release, as it appears to be a bit stronger on both SP and MP. Here are my latest results for this new version on 4 cores:
Against Critter 1.6a at 3' + 2": 26.5 -15.5.
Against Critter 1.6a at 5' + 3": 34 - 22.
Against Houdini 3.0 at 5' + 3": 28 -28.

I don't expect we will stay even with Houdini 3, that is surely a bit lucky result. But both the Critter and Houdini results are better than what we get on single core, which is a pleasant surprise to us.

On 12 cores we beat Stockfish 3 by 27 - 23.

[bupalo]

Any update?

[lkaufman]

Update (4 cores each), 5' + 3":

Against Houdini 3, we lost by 35 to 37.

Against Stockfish 3, combining my own tests with Don's (which are 10'), we lead by 114-93.

[beram]

Update (4 cores each), 5' + 3":

Against Houdini 3, we lost by 35 to 37.

Against Stockfish 3, combining my own tests with Don's (which are 10'), we lead by 114-93.

Against Houdini 3 it looks promising, but 55% against Stockfish 3 is not that great. But time will tell

thx for publishing
grts Bram