Rybka on "fire"

[jix]

Time Controls: 3/2
Hardware: Q9550 @ 4.4
Book: Zeeman 3.4
TB: None

Deep Rybka 4 x64 - 48.0 - 52.059.5 - 40.556.0 - 44.0 163.5/300
Houdini 1.01 x64 4_CPU - 52.0 - 48.053.5 - 46.550.5 - 49.5 156.0/300
Fire 1.31 x64 40.5 - 59.546.5 - 53.562.5 - 37.5 149.5/300
Stockfish 1.7.1 JA 64bit 44.0 - 56.049.5 - 50.537.5 - 62.5 131.0/300

After running numerous tournaments to find the right piece values and settings I think have finally found a winning combination for Rybka. The scary thing is the score would have been even more in Rybka's favor if not for games where it lost on time in a drawn position--oddly this really only occured with Houdini.

If anyone is interested in looking over the games I will upload them.

[Albert Silver]

Time Controls: 3/2
Hardware: Q9550 @ 4.4
Book: Zeeman 3.4
TB: None

Deep Rybka 4 x64 - 48.0 - 52.059.5 - 40.556.0 - 44.0 163.5/300
Houdini 1.01 x64 4_CPU - 52.0 - 48.053.5 - 46.550.5 - 49.5 156.0/300
Fire 1.31 x64 40.5 - 59.546.5 - 53.562.5 - 37.5 149.5/300
Stockfish 1.7.1 JA 64bit 44.0 - 56.049.5 - 50.537.5 - 62.5 131.0/300

After running numerous tournaments to find the right piece values and settings I think have finally found a winning combination for Rybka. The scary thing is the score would have been even more in Rybka's favor if not for games where it lost on time in a drawn position--oddly this really only occured with Houdini.

If anyone is interested in looking over the games I will upload them.

I couldn't really read the individual match-up scores or conditions. As to losing on time with an increment, I have never seen this happen, so I am intrigued.

What were your settings, and how many time losses did you suffer?

[kingliveson]

Code: Select all

                            1	          2	          3	          4                                           	
1 Deep Rybka 4 x64         ***           48.0 - 52.0   59.5 - 40.5   56.0 - 44.0   163.5/300
2 Houdini 1.01 x64 4_CPU   52.0 - 48.0   ***           53.5 - 46.5   50.5 - 49.5   156.0/300
3 Fire 1.31 x64            40.5 - 59.5   46.5 - 53.5   ***           62.5 - 37.5   149.5/300
4 Stockfish 1.7.1 JA 64bit 44.0 - 56.0   49.5 - 50.5   37.5 - 62.5   ***           131.0/300

Hope it makes for easier read, if I haven't corrupted the data.

[Albert Silver]

Code: Select all

                            1	          2	          3	          4                                           	
1 Deep Rybka 4 x64         ***           48.0 - 52.0   59.5 - 40.5   56.0 - 44.0   163.5/300
2 Houdini 1.01 x64 4_CPU   52.0 - 48.0   ***           53.5 - 46.5   50.5 - 49.5   156.0/300
3 Fire 1.31 x64            40.5 - 59.5   46.5 - 53.5   ***           62.5 - 37.5   149.5/300
4 Stockfish 1.7.1 JA 64bit 44.0 - 56.0   49.5 - 50.5   37.5 - 62.5   ***           131.0/300

Hope it makes for easier read, if I haven't corrupted the data.

Thanks. Odd scores: Houdini barely beating Stockfish, and even DR4's 56% against it is very low IMHO.

[M ANSARI]

Unfortunately R4 is losing on time quite often, even with increment. This only happens in very fast games and I still can't really figure it out because in some setups (even sudden death) R4 doesn't get any time losses. Most likely there is something wrong in the interaction between engine and GUI. Another thing is that R4 is probably the most underrated engine on big hardware. It is a real monster on 8 cores and more.

[notyetagm]

Code: Select all

                            1	          2	          3	          4                                           	
1 Deep Rybka 4 x64         ***           48.0 - 52.0   59.5 - 40.5   56.0 - 44.0   163.5/300
2 Houdini 1.01 x64 4_CPU   52.0 - 48.0   ***           53.5 - 46.5   50.5 - 49.5   156.0/300
3 Fire 1.31 x64            40.5 - 59.5   46.5 - 53.5   ***           62.5 - 37.5   149.5/300
4 Stockfish 1.7.1 JA 64bit 44.0 - 56.0   49.5 - 50.5   37.5 - 62.5   ***           131.0/300

Hope it makes for easier read, if I haven't corrupted the data.

Thanks. Odd scores: Houdini barely beating Stockfish, and even DR4's 56% against it is very low IMHO.

The same result I see over and over gain: Rybka 4 is the strongest engine, Houdini is slightly weaker.

[beram]

Code: Select all

                            1	          2	          3	          4                                           	
1 Deep Rybka 4 x64         ***           48.0 - 52.0   59.5 - 40.5   56.0 - 44.0   163.5/300
2 Houdini 1.01 x64 4_CPU   52.0 - 48.0   ***           53.5 - 46.5   50.5 - 49.5   156.0/300
3 Fire 1.31 x64            40.5 - 59.5   46.5 - 53.5   ***           62.5 - 37.5   149.5/300
4 Stockfish 1.7.1 JA 64bit 44.0 - 56.0   49.5 - 50.5   37.5 - 62.5   ***           131.0/300

Hope it makes for easier read, if I haven't corrupted the data.

Thanks. Odd scores: Houdini barely beating Stockfish, and even DR4's 56% against it is very low IMHO.

The same result I see over and over gain: Rybka 4 is the strongest engine, Houdini is slightly weaker.

My observation to. Houdini yet seems to be second best program after Rybka4.

[ernest]

My observation to. Houdini yet seems to be second best program after Rybka4.

I am surprised: in all your other posts you claim Houdini is better than Rybka 4...

[beram]

My observation to. Houdini yet seems to be second best program after Rybka4.

I am surprised: in all your other posts you claim Houdini is better than Rybka 4...

I gave good Houdini vs R4 results of own testing at 32 bit.
At 64 bit - at blitz - i gave about even for what I did test and what I read f tests of others.

This is what got tonight on my new AMD x6 with the "repared" Houdini 1.01 8CPU:

AMD Phenom II X6 1090T 3200 Mhz, (Fritzmark 23,1)

1 Houdini DEVEL x64 8_CPU +11/=30/-9 52.00% 26.0/50
2 Deep Rybka 4 x64 +9/=30/-11 48.00% 24.0/50

and this what Pavel Hase gets at very long timecontrol:
http://www.talkchess.com/forum/viewtopi ... 31&t=34876
Rybka 4x64 - Houdini 1.01x64 24.5:23.5 (8:7=33, +- 7 ELO)
Books both: PH-TestBook2010.ctg (min_use=3, max_depth=25)
AMD Athlon64 3200+, 256MB hash, 90m+30s, GUI DeepRybka3, 3-5TB (+several 6), pondering off.

So very close these engines are so 1 and 2 or even

Kind regards Bram