Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Here will be posted results of Rybka 4 vs Stockfish 1.8 !
- - By UltimateChess (**) [cz] Date 2010-07-17 17:55
It seems that there is a controversy about strengths of this two programs. I myself am confused. By this thread we can GAIN MORE COLLECTIVE KNOWLEDGE and also Rybka operators will have easy time to find our results and can assimilate data we created for the benefit of Rybka.
Try not to comment matches here, use other threads. Try not to comment anything here :-)

If you performed a match then:

Post your results here.

Post all relevant settings of the match so everybody can check and repeat your experiment if necessary.

If you can do statistical analysis, good, if you are not sure dont write here any statistics just data from match.

Hope this idea works. I am looking for your matches.
Parent - - By UltimateChess (**) [cz] Date 2010-07-18 04:53
Match - Rybka 4 vs Stockfish 1.8

Number of games: 116

Rybka won:   36
Rybka draw:  44
Rybka lost:    36
Result- DRAW

Tablebases: 4 pieces for both
Opening book: Random move 3 ply, learning off, each opening played by both
Engine parameters: default for both
Computer: 1,6Ghz, single core
Time for game: 5min + 1s

All games can be downloaded here.
http://www.4shared.com/file/HaPOtcZF/4min1s_R4-vsStckf18.html

From this amount of games I cannot conclude even slightest advantage of R4. Very interesting result. Often R4 didnt manage to win with advantages.
More matches to come... :-))
Parent - By Wayne Lowrance (***) Date 2010-07-18 05:16
very interesting. Have you follow R4-SF1.8 at long time control. Match results can be viewed at "talk chess"
Parent - By titanium cranium (***) [us] Date 2010-07-18 13:00
If we could plot ELO over time, wouldn't we see Stockfish's rating increasing at a higher rate than Rybka, suggesting that it may pass R4 soon?
Parent - By Athinira (**) [dk] Date 2010-07-18 14:45
In my experiences, Stockfish (even the older versions) perform very well in games with short time controls/limited processing power. Also the limited opening book doesn't help your testing IMO.

Rybka 4 is also getting known for being one of the engines that scales the best on multiple cores. At least single core matches doesn't favor it, thats for certain.
Parent - By UltimateChess (**) [cz] Date 2010-07-19 06:46
Match ended

Games 114

Rybka wins    38
Rybka draws  46
Rybka loses   30

Rybka is WINNER

Time for game: 4min
Opening book: Fritz11, learning off, opening played by both
Engine settings: default for both
Tablebases: 4 pieces
Hash: 128MB for both
Computer: 1,6Ghz single

Games can be downloaded here:
http://www.4shared.com/file/8karSSmg/4min_R4-vsStckf18.html

Comment: From the 114 games I cannot conclude that Rybka is stronger on classical alpha level 5% although R4 won the match. Those who critisise - post your own match so we can see your results. My opinion is not important, nor yours - the only good answer is to perform match and post your data here. With resources I have I try to do best I can. See you tomorrow and post your matches :-)
Parent - - By UltimateChess (**) [cz] Date 2010-07-20 05:30
We saw that Rybka has problems in crazy openings. So I run the match one more time with different time settings.

Games 64

Rybka wins     18
Rybka draws   23
Rybka loses    23

Rybka LOST

Time settings: 3min + 3s
Opening book: Random move 3 ply
all other settings are same

This crazy book can be downloaded here:
http://www.4shared.com/file/WMaHUoIM/Random_Move_Chess_Opening_Book.html

And all games from this match can be downloaded here:
http://www.4shared.com/file/FPd4SZf2/3min3s_R4-vsStckf18.html

This clear loss is very unexpected when we think that we deal with top chess entity. There is something that Stockfish can easily orient itself in really crazy positions - not Rybka (as I thought). I would really think that Rybka should win under any circumstances when all other things being equal for her and her opponent. The crazyness of 1.f3 g5 2.h4 really gets handled better by Stockfish - but I am not 100% sure.
Although for practical game this may not be important because we want Rybka to  be good in "standard" chess....
So in next match I will return to classical Fritz11 book and will give longer time controls. So see you :wink:
Parent - By Quapsel (****) [de] Date 2010-07-20 06:26

> This clear loss is very unexpected when we think that we deal with top chess entity.


We wouldn't have expected this. Sure.

But we should be aware also:
Normaly R4 is less than 40 ELO stronger than SF18. This would fit to a 55:45 result.
Then the likelyhood for R4 to get 29.5 Points or less is 0,044.
Not much!
But if you or other people do a lot of short testrows in this way you'll see such results sometimes. And such results are astonishing and discussed in public then.
But there musn't be another reason than simple miisfortune for R4.
0,044 is a good deal bigger than 0. ;-)

Quap
Parent - By hal9000 (**) [no] Date 2010-07-20 10:24 Edited 2010-07-20 10:31
Is there any controversy? The results I've seen shows that SF 1.8 is only marginally better than 1.71 at blitz levels, still leaving it behind Rybka 4.

Rybka 4 is destroying Stockfish 1.8 in this match at tournament level:

http://home.halden.net/mordor/match/broadcast.php

(At the time of writing, Rybka 4 is leading 26-16)

I personally find these results (on a fast multicore CPU) to be far more interesting than those based on blitz games with only one (weak) CPU/core. Since most people use (faster) multicore CPUs nowadays, I also think it's more realistic to use more cores.

Besides, Rybka's default blitz/bullet time controls don't seem to be optimal (as indicated by the TC3100150 results). This seems to be less of an issue at longer time controls, and this (in addition to more cores) may explain why Rybka is clearly winning the tournament level match against SF 1.8.

EDIT: I forgot one thing: It seems to me that you're not running the 64 bit version of Rybka, is this correct? Which CPU are you using? (Atom?)
Parent - - By UltimateChess (**) [cz] Date 2010-07-21 11:41
Hello!

I took some advice seriously. I wouldnt say that Rybka plays weak in short games - I would say that Stockfish is very good in blitz games. Again I used Fritz11 opening book and then listen to advice that Rybka plays better if given more time. Here are the results.

Games 39

Rybka wins   16
Rybka draws 12
Rybka loses  11

Rybka won clearly.

Time for game increased to 10min
Opening: Fritz11
Hash 128
all other setting are default and standard

All games can be inspected here:
http://www.4shared.com/file/dCXCNs-K/10min_R4-vsStckf18.html

This match really proved that with time increase Rybka is able to play better game.
Parent - By hal9000 (**) [no] Date 2010-07-24 13:57
In the tournament level match mentioned above (Rybka 4 vs Stockfish 1.8), Rybka 4 won 29-19 (scroll down to bottom of page). Rybka was winning one game which was cancelled because Stockfish crashed (the game was replayed and this time it ended in a draw). However, this didn't change the outcome of the match, and SF would've lost clearly anyway.
Parent - By UltimateChess (**) [cz] Date 2010-07-26 05:22
Check out this spectacular win against Stockfish!

I did a little mach at long time control ( amazing 30 min for each!)

And really the longer time control the better. Rybka crushed Stockfuck like a bug. Even in very bad opening positions.

The match data are here:

http://www.4shared.com/file/gnFUY7V3/30min_R4-vsStckf18.html

See you!
Up Topic Rybka Support & Discussion / Rybka Discussion / Here will be posted results of Rybka 4 vs Stockfish 1.8 !

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill