10% of rybka2.3.2a 32 bit-Movei00_8_438 57-43

[Uri Blass]

I wonder how many programs can beat 10% of 32 bits of the fish.

Note that I expect better result at faster time control and time control of the match was 40/40 on PIV 3.19 ghz under arena when rybka was set to 10% strength.

Tablebases were not used and hash was set to 64 mbytes per engine.

I wonder if there are programs that can perform better against 10% of the fish at longer time control(note that very fast time control is not recommended to check it because arena steal time from engines and I think that it may be significant in time control that is faster than 10 minutes/40 moves(when rybka get less than 1 minute/40 moves)

Possible candidates are Deep Sjeng or Naum when you can test Naum against 10% of Rybka64 bit or even against 20% of her because
10% of rybka 32 bit is probably too weak for naum to score something in the enviroment of 50% in these conditions.

Uri

[Dr.Wael Deeb]

I don't like this type of testing,it's definitely misleading....

[pichy]

I don't like this type of testing,it's definitely misleading....

How can it be misleading when the heading specifically states 10% of rybka NOT 100%

[Michael Sherwin]

I wonder how many programs can beat 10% of 32 bits of the fish.

Note that I expect better result at faster time control and time control of the match was 40/40 on PIV 3.19 ghz under arena when rybka was set to 10% strength.

Tablebases were not used and hash was set to 64 mbytes per engine.

I wonder if there are programs that can perform better against 10% of the fish at longer time control(note that very fast time control is not recommended to check it because arena steal time from engines and I think that it may be significant in time control that is faster than 10 minutes/40 moves(when rybka get less than 1 minute/40 moves)

Possible candidates are Deep Sjeng or Naum when you can test Naum against 10% of Rybka64 bit or even against 20% of her because
10% of rybka 32 bit is probably too weak for naum to score something in the enviroment of 50% in these conditions.

Uri

I was curious, so, I ran the latest RomiChess beta against the Rybka free beta at 10 % in a small 20 game match. Romi scored +3 -13 =4. 40/40 was used.

[Uri Blass]

Thanks for sharing your results.
It seems that the gap between movei and romichess is bigger than the gap between rybka beta and rybka2.3.2a

I tried the same match at faster time control(40/10) and movei lost 51-49

36-34 for rybka and 30 draws.

There were more draws at 40/40(final result at that time control was 35-21 and 44 draws)

Note that rybka2.3.2a is about 150 elo better than the beta version of rybka.

From the CEGT 40/20:
12 Rybka 2.3.2a w32 1CPU 2975
62 Rybka 1.0 Beta w32 2821

From the CCRL 40/40:

Rybka 2.3.2a 32-bit 3018
Rybka 1.0 32-bit 2883

Uri

[Uri Blass]

I decided to play a match between movei and rybka32 bit.

This time I told the 32 bit to use 30% strength and movei used 100% strength.

Result after 47 games is 23-10 for rybka 32 bit with 14 draws.
I am surprised because I thought that 2.3.2a is more than 3 times better than beta 32 bits and the result suggest that it is not the case

I suspect that there is a problem with the 32 bit version of rybka because rybka lost on time some games(it is not supposed to happen because rybka was supposed to use only 30% of the time that it got)

I wonder if the beta version did not cause other problems like stealing cpu time from movei because results seem too good for rybka beta.

Uri

[Michael Sherwin]

The changes to Rybka from the 32 bit beta to the current version may benifit Rybka overall like you thought, however, the changes may not be that good against Movei.