a question

[Uri Blass]

If Ippolit is 40-50 elo point stronger than Rybka 3, without "Rybka´s tables", how much could it be stronger WITH Rybka´3 tables ?

Any clue?

Curious SL

Maybe it cannot be stronger with the tables and one of the advantages of it is having something else instead of the tables.

Uri

Do you think Rybka gains nothing with its tables?

I think that it may be possible to replace the tables by a better alternative.
It is not the same as saying that rybka gains nothing from them because if you do not think about the better alternative rybka may be weaker without the tables.

Uri

[solis]

This is kind of crazy but Ippolit just destroyed Rybka 3. I will do more games and also repeat match with Naum 4.
Engine Score Ip Ry S-B
1: Ippolit 7.0/10 ·········· =11=1=10=1 21.00
2: Rybka 3 3.0/10 =00=0=01=0 ·········· 21.00

10 games played / Tournament is finished
Name of the tournament: Arena tournament
Site/ Country: XPS-3GHZ, United States
Level: Tournament Game in 5 Minutes
Hardware: Intel(R) Pentium(R) 4 CPU 3.00GHzDual 2992 MHz with 2,048 MB Memory
Operating system: Microsoft Windows XP Professional Service Pack 3 (Build 2600)

[Leto]

Hi Slobodan,
For some reason this programs works only on this computer that uses 1CPU.On the others it crashes even they are 1CPU computers.On Dual Core and Quad also no luck but this match was done just for fun although the result is kind of scary.
Miodrag

Hi Miodrag,

Why don´t you play a match of 20 games between Rybka 3 - 1cpu and Ippolit. I suspect that the difference beetween the two engines is much higher than 40 elo points.

SL

Played two matches of 10 games each at 2 minute blitz with the Perfect 8.32 book, Rybka 3 won the first match 5.5 - 4.5, and the second match 6-4. So far I'm not seeing evidence that Ippolit is stronger than Rybka 3. I'll run some more 10 game matches.

Boy do I feel stupid, I forgot to change the thread count so those two matches Rybka 3 was actually running on 4 CPUs! Ugh I'll do it over.

[Leto]

First played a 20 game match, 2 minute blitz with the perfect 8.32 book, Ippolit won 12 to 8, then I played a 60 game match with same conditions and Ippolit won 31 to 29. Total score after 80 games is 43 to 37, so Ippolit scored 54% against Rybka 3.

I'll do a final 20 games to bring the total to 100, then I'll do testing with increased time controls.

[mclane]

but why. if this an illegal clone of a commercial engine than testing makes IMO not much sense. maybe we should better concentrate on legal engines and to see how they make progress.

[Volker Pittlik]

If Ippolit is 40-50 elo point stronger than Rybka 3...

It isn't:

Code: Select all

   1 Rybka 2.2n2 32-bit   230   25   24   771   81%   -35   20% 
   2 Stockfish 1.5.1      205   86   80    55   68%    61    5% 
   3 Stockfish 1.5        197   54   54    90   45%   230   43% 
   4 IPPOLIT 0.080a       178   55   52   155   69%     3    5% 

And even if it were, it would be a piece of crap, crashing in 25% of the games.

amen

[Sylwy]

Versions b & c are better compiled.

[Volker Pittlik]

Versions b & c are better compiled.

Better compiled than what? Please note: I compiled it myself and I know how to do it. That program is crashing WAY to much and is therefore junk.

[Sylwy]

OK !
But I played with b version a lot of games-at various time controls-without any chrash. I noted also very different results ( on 32 bit OS ) between Intel & Athlon CPU. On Athlon the results of IPP aren't great. Rybka 3 is much better here.

[M ANSARI]

I think they don't have the evaluation tables because they can't get them to work properly or they have yet to disassemble them. All the matches I have seen reports on have been at very fast time controls. Most likely Rybka 3 has protection code that slows it down a little, and also with such a small executable it is bound to perform well at fast time controls. I would guess that contempt also is affecting performance.

Do the test at 5 0 or 10 0 and put the contempt factor of Rybka at "0" then try the test. I would be very surprised if Rybka 3 came out on the losing end.