Another result from CEGT (http://husvankempen.de/nunn/phpBB2/viewtopic.php?t=976):
Hiarcs 12 SP - Rybka 2.3.2a SP 34,5-65,5.
That makes an Elo difference of 111 points.
Hiarcs 12 SP - Rybka 2.3.2a SP 34,5-65,5.
That makes an Elo difference of 111 points.
>Hiarcs 12 SP - Rybka 2.3.2a SP 34,5-65,5.
Yes but with some range of error.
See for example in the page you give:
H.K had as a result:
Hiarcs 12.0 4CPU vs Deep Shredder 11 x64 4CPU ( 50) 32.0-18.0 perf= 3046 (H.K.)
But G.S had:
Hiarcs 12.0 4CPU vs Deep Shredder 11 x64 4CPU (100) 50.0-50.0 perf= 2947 (G.S.)
If we see the first result of 50 games some can say that "Oh Hiarcs 12 seems to be much better than Shredder 11.".
But if he sees the second result first he would probably say: "Oh it seems Hiarcs 12 is about equal with Shredder 11.".
Go figure! :-)
If we see these 2 results separately, then we will definitely be confused.
So more games should be played to make a more definite conclusion....
-------------------------------------
So if we see the:
Hiarcs 12.0 4CPU vs Rybka 2.3.2a x64 4CPU ( 50) 18.0-32.0 perf= 2990 (H.K.)
Then we should not draw immediate conclusions that Rybka 2.3.2a x64 4 CPU is stronger than Hiarcs 12 4 CPU, because after 100 more games xcore maybe 50-50 or even better for Hiarcs 12.
I mean we should play at least the number of games required, until the range of errors don't overlap.
Scores CEGT sofar:
2 CPU: H12 - R232a 29,5:70,5
1 CPU: H12 - R232a 34,5:65,5
That is 200 games and still no sign of a H12 recovery :-)
2 CPU: H12 - R232a 29,5:70,5
1 CPU: H12 - R232a 34,5:65,5
That is 200 games and still no sign of a H12 recovery :-)
That is a :-( and not :-)
I hoped Hiarcs 12 to reach Rybka's 2.3.2a strength. I knew it was difficult, i just hoped.
I hoped Hiarcs 12 to reach Rybka's 2.3.2a strength. I knew it was difficult, i just hoped.
Nah, don't worry! If Mark keeps on improving Hiarcs, there will be a day that Hiarcs is stronger than Rybka 2.3.2a :-)
The completed match result has been posted with my comments and I have added a link to download the games for those interested.
http://www.hiarcs.net/forums/viewtopic.php?t=1133
PeterG
http://www.hiarcs.net/forums/viewtopic.php?t=1133
PeterG
really ?
i found this in the internet:
Motor Punkte Ry
1: Hiarcs12MPUCI [t4,tc002,nobook] 92,0/182 0=0=01010=11==01=0010==10=11010==0===1==1=0===11110101==0==11110001=101=0110==1111110101===00=0==1100=000000=1=1=0=1==1011=0=11==11=100=110==0==000=1=101=000111==01=1====0=1000==0==0
2: Rybka v2.3.mp.x64 90,0/182 ······················································································································································································
182 Partien gespielt / Turnier beendet
Name des Turniers: test
Ort/ Land: SCW, Germany
Spielstufe: Blitz 3/2
games + positions here
http://schachcomputerwelt.foren-city.de/topic,151,-hiarcs12-spielstile-kleines-experiment.html
intel core2quad. 4 cores each. opening=91 positions, game repeated with exchanged colors
i found this in the internet:
Motor Punkte Ry
1: Hiarcs12MPUCI [t4,tc002,nobook] 92,0/182 0=0=01010=11==01=0010==10=11010==0===1==1=0===11110101==0==11110001=101=0110==1111110101===00=0==1100=000000=1=1=0=1==1011=0=11==11=100=110==0==000=1=101=000111==01=1====0=1000==0==0
2: Rybka v2.3.mp.x64 90,0/182 ······················································································································································································
182 Partien gespielt / Turnier beendet
Name des Turniers: test
Ort/ Land: SCW, Germany
Spielstufe: Blitz 3/2
games + positions here
http://schachcomputerwelt.foren-city.de/topic,151,-hiarcs12-spielstile-kleines-experiment.html
intel core2quad. 4 cores each. opening=91 positions, game repeated with exchanged colors
Is tc002 some sort of "Private" personality for HIARCS 12?
yes the text says it is not the default style.
You can follow it in English here http://hiarcs.net/forums/viewtopic.php?t=1166
I also made a small test of Hiarcs 12 mp against Rybka 2.3.2a mp (Q6600, 256MB hash, my personal book against Hiarcs´s own book, 15m+10s) result was 1,5:8,5 and it looks that Hiarcs is still significantly weaker in comparison to Rybka especially in middlegame. I would like to know how much will Hiarcs benefits from self-learning...
Hiarcs is very weak, I consider it worse than Toga/Fruit. i would be happy to receive a test version from the Hiarcs people to see if my opinions are still valid.
> i would be happy to receive a test version from the Hiarcs people to see if my opinions are still valid.
So what you want is a HIARCS for free? I don't think that claiming that the engine is weak will get you one. I know it's your opinion, but mine is that HIARCS is very strong.
The book makes a huge difference with H12.
What I can't work out is whether H12 + book is strong because it out-books everything else and just arrives at better positions than they do (in which case I'd have a fairly low opinion of the engine) or whether it just gets the program into positions it plays well (in which case I'd have a higher view of the engine).
What I can't work out is whether H12 + book is strong because it out-books everything else and just arrives at better positions than they do (in which case I'd have a fairly low opinion of the engine) or whether it just gets the program into positions it plays well (in which case I'd have a higher view of the engine).
> i would be happy to receive a test version from the Hiarcs people to see if my opinions are still valid.
LOL, so Hiarcs is weak but you'd still be happy to get it for free. :)
>LOL, so Hiarcs is weak but you'd still be happy to get it for free.
Yip that's part of the diddle.
As you asked so nicely you can download Hiarcs source code here http://www.hiarcs.com/1981_source.htm
>As you asked so nicely you can download Hiarcs source code here http://www.hiarcs.com/1981_source.htm
I would have liked to have test-driven the latest version not a Dad's Army version. Out of curiosity how many played the freestyle with Hiarcs and of those how many finalists do we have with Hiarcs?
With the 1981 version I have given you the source to I expect not many played.
If no one can trust Hiarcs in freestyle then it shows it is nothing more than an expensive toy.
As you have enjoyed the 1981 source code so much please feel free to download the 1983 source code. Then perhaps you will play a few games against it to see if you can beat it? My bet is that you will lose.
As for Freestyle nobody knows what engine/engines people are using in the final. I am sure Rybka is widely used. Several of the finalists own Hiarcs. Which engines have you bought?
As for Freestyle nobody knows what engine/engines people are using in the final. I am sure Rybka is widely used. Several of the finalists own Hiarcs. Which engines have you bought?
You can be 100% sure that people in the latest Freestyle have used Hiarcs 12 or Hiarcs Paderborn for at least some analysis. You can be pretty sure that at least one finalist has also done this.
Can anyone here download Hiarcs 12 or is this just an offer to Terminator? I didnt check to see if it is free or
pay to download?
pay to download?
Yeah I have to break tradition and stick up for Harvey / Hiarcs here. It is absolutely ridiculous to say a 2900+ elo program is weak.
Even when I had a big moan about their website it was never because the program was weak - what does that make all Grandmasters??
H12 is, probably, the best at attacking the king in certain positions. It also has the best included opening book.
Even when I had a big moan about their website it was never because the program was weak - what does that make all Grandmasters??
H12 is, probably, the best at attacking the king in certain positions. It also has the best included opening book.
> H12 is, probably, the best at attacking the king in certain positions. It also has the best included opening book.
And don't forget that with HIARCS 12 you get HIARCS Paderborn 2007, another very strong and interesting engine, for free.
Yeah I've been thinking of your comments recently. I've begun testing on Paderborn - my early results seem to show that it is quicker tactically. I haven't had time to observe style yet, but hopefully in a week I should have more results.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill
