Ivan The Terrible

[Titu]

The point of the post was to know what contempt setting you used for Rybka 3.

If I wanted to discuss whether or not Rybka 3 has clones then I would take advice from engine _experts_ like Gerd I, GPC etc.

[Uri Blass]

What is the final conclusion?

I read in another forum a claim that firebird1.1 is better at long time control because it is losing against rybka3 at 1+0 but win at 3+0

I said that I consider 3+0 as very fast time control and the poster claimed that programs can calculate more than 10 plies forward at blitz so it is not a fast time control for chess programs.

It is better if you stick with Firebird1.1 and Rybka3(contempt=0)

Dann Corbit's post was about
IvanHoe 9.63b x64 and I do not know if it is relevant to Firebird1.1.

Uri

[kingliveson]

FireBird (1.1) is stronger than Rybka at 1+0 time control. I have not seen any contradicting results. Using the Intel (popcnt) compile from Norm's ChessLogik website, at 1+0 time control, here are some games using KLOECOA00E97V test suite:

Code: Select all

Rybka vs FireBird 1+0  
                      
1   FireBird 1.1 x64  +18/=27/-5 63.00%   31.5/50
2   Rybka 3 x64       +5/=27/-18 37.00%   18.5/50

-25 positions with switched colors
-ponder off
-no egtb
-128 mb hash
-AMD Phenom II 940 @ 3.6 GHz
-Engines use 4 cores
-Time Control: 1 minute

At this stage of development of both IvanHoe and FireBird, the question is not which is stronger, but which is more stable on Windows environment. Regarding contempt, take a look at this post on the members only premium sub-forum.

[Hood]

Ivan the severe

[Uri Blass]

FireBird (1.1) is stronger than Rybka at 1+0 time control. I have not seen any contradicting results. Using the Intel (popcnt) compile from Norm's ChessLogik website, at 1+0 time control, here are some games using KLOECOA00E97V test suite:

Code: Select all

Rybka vs FireBird 1+0  
                      
1   FireBird 1.1 x64  +18/=27/-5 63.00%   31.5/50
2   Rybka 3 x64       +5/=27/-18 37.00%   18.5/50

-25 positions with switched colors
-ponder off
-no egtb
-128 mb hash
-AMD Phenom II 940 @ 3.6 GHz
-Engines use 4 cores
-Time Control: 1 minute

At this stage of development of both IvanHoe and FireBird, the question is not which is stronger, but which is more stable on Windows environment. Regarding contempt, take a look at this post on the members only premium sub-forum.

contempt=0 is designed to be the best against equal opponents by definition.

contempt=15 was the default only to win against weaker opponents because at the time rybka came out all opponents were weaker.

Larry kaufman:
"My rule of thumb is that the optimum contempt value for R3 is about 1/12 of the rating difference from the opponent."

assuming that rybka is not significantly stronger than Firebird
it is clear based on larry kaufman that contempt=0 is better.

Uri

[kingliveson]

contempt=0 is designed to be the best against equal opponents by definition.

contempt=15 was the default only to win against weaker opponents because at the time rybka came out all opponents were weaker.

Larry kaufman:
"My rule of thumb is that the optimum contempt value for R3 is about 1/12 of the rating difference from the opponent."

assuming that rybka is not significantly stronger than Firebird
it is clear based on larry kaufman that contempt=0 is better.

Uri

GM Kaufman full quote:

My rule of thumb is that the optimum contempt value for R3 is about 1/12 of the rating difference from the opponent. So against N4 I would expect that the optimum contempt value would be somewhere around 8. I would not expect much difference in the score using 15 and using zero, because they are about equally far from what I would call the optimum value.

It is settled (I think)...from here on forward, all testings with Rybka vs IvanHoe/FireBird will use contempt factor = 0.

[Robert Flesher]

Computer chess and Star Wars, some would argue no direct correlation. However, in retrospect, and in regards to Iggorrit, Ivan The Terrible, Robbolito, etc, etc. All new threads should be callled, " The Clone Wars"

[kingliveson]

Long Time Control: Rybka vs IvanHoe 40+2

Elostat: IvanHoe +42
Bayesian: IvanHoe +30

Code: Select all

 
    Program                          Elo    +   -   Games   Score   Av.Op.  Draws

  1 IvanHoe 9.63b x64              :   21   56  53    50    56.0 %    -21   68.0 %
  2 Rybka 3 x64                    :  -21   53  56    50    44.0 %     21   68.0 %

The matches use KLOECOA00E97V test suite.

-25 positions with switched colors
-Ponder off
-No tablebases
-512 MB hash
-AMD Athlon 64x2 3800+ @ 2.0 GHz
-Engines use 2 cores
-Time Control: 40m +2s inc

Additional Settings:
-Rybka Contempt: 0
-IvanHoe Pawn Hash: 64 MB

Code: Select all

 
Individual statistics:

1 IvanHoe 9.63b x64         :   21   50 (+ 11,= 34,-  5), 56.0 %

Rybka 3 x64                   :  50 (+ 11,= 34,-  5), 56.0 %

2 Rybka 3 x64               :  -21   50 (+  5,= 34,- 11), 44.0 %

IvanHoe 9.63b x64             :  50 (+  5,= 34,- 11), 44.0 %

Code: Select all

 
Games        :     50 (finished)

White Wins   :     11 (22.0 %)
Black Wins   :      5 (10.0 %)
Draws        :     34 (68.0 %)
Unfinished   :      0

White Perf.  : 56.0 %
Black Perf.  : 44.0 %

ECO A =      7 Games (14.0 %)
ECO B =     18 Games (36.0 %)
ECO C =     10 Games (20.0 %)
ECO D =      7 Games (14.0 %)
ECO E =      8 Games (16.0 %)

Code: Select all

 
Bayesian Elo Rating:

Rank Name                Elo    +    - games score oppo. draws
   1 IvanHoe 9.63b x64    15   36   35    50   56%   -15   68%
   2 Rybka 3 x64         -15   35   36    50   44%    15   68%

Games

[kingliveson]

Rybka vs IvanHoe 40+2

Code: Select all

1   IvanHoe 9.63b x64  +11/=33/-6 55.00%   27.5/50
2   Rybka 3 x64        +6/=33/-11 45.00%   22.5/50

The matches use KLOECOA00E97V test suite.

-25 positions with switched colors
-ponder off
-no egtb
-512 mb hash
-AMD athlon 64x2 3800+ @ 2.0 GHz
-Engines use 2 cores
-Time Control: 40m +2s inc

Rybka
-Contempt: 0

IvanHoe
-q = 1025
-r = 550
-b = 325
-n = 320
-p = 110
-bp = 50
-Pawn Hash: 64 MB

Games