Another Fascinating Game

This forum is for general discussions and questions, including Collectors Corner and anything to do with Computer chess.

Moderators: Harvey Williamson, Steve B, Watchman

Forum rules
This textbox is used to restore diagrams posted with the fen tag before the upgrade.
User avatar
Harvey Williamson
Site Admin
Posts: 6079
Joined: Sun Jul 29, 2007 6:57 am
Location: Media City, UK
Contact:

Another Fascinating Game

Post by Harvey Williamson »

Hi All,

Following on from the fascinating game http://hiarcs.net/forums/viewtopic.php?t=102

Our latest version Hiarcs 11.55e just scored a very nice win on Playchess. Hiarcs was playing on my 8x2.66 intel machine. However Zap operated by Suj was on his 32 core machine. He tells me it is the first time Zappa has ever lost on this machine.

Best Wishes,

Harvey

[Event "90m + 30s, rated"]
[Site "?"]
[Date "2007.08.12"]
[Round "?"]
[White "KingOfEngines, Zappa Zanzibar X64"]
[Black "HarveyHiarcsX, HIARCS 11.55e MP"]
[Result "0-1"]
[WhiteElo "2644"]
[BlackElo "2605"]
[PlyCount "84"]
[EventDate "2007.08.12"]

1. e4 {0} c5 {B/0 0} 2. Nf3 {0} Nc6 {B/0 0} 3. d4 {0} cxd4 {B/0 0} 4. Nxd4 {0}
Nf6 {B/0 0} 5. Nc3 {0} e5 {B/0 0} 6. Ndb5 {0} d6 {-0.01/1 0} 7. Nd5 {0} Nxd5 {
-0.01/1 0} 8. exd5 {0} Nb8 {-0.01/1 0} 9. c4 {0} Be7 {-0.01/1 0} 10. Bd3 {0} a6
{-0.01/1 0} 11. Nc3 {0} O-O {-0.01/1 0} 12. f3 {0} Bh4+ {0.14/19 338} 13. g3 {
664} Bg5 {0.06/21 0} 14. Qc2 {(Bxg5) 375} Bxc1 {-0.17/19 262} 15. Rxc1 {333} f5
{KingOfEngines reconnected. -0.23/20 0} 16. O-O {311} a5 {-0.22/21 0} 17. c5 {
(Qf2) 277} Na6 {-0.17/19 405} 18. Bxa6 {161} Rxa6 {-0.22/19 0} 19. Rfe1 {646}
Bd7 {-0.19/19 0} 20. a4 {(Kg2) 158} f4 {-0.69/18 227} 21. g4 {(Ne4) 121} h5 {
-0.67/18 152} 22. Ne4 {42} hxg4 {-0.64/17 94} 23. Nxd6 {0} Qf6 {-0.39/18 686}
24. fxg4 {0} Bxg4 {-0.36/17 35} 25. Qg2 {80} Bd7 {-0.83/18 467} 26. Rc2 {150}
Bxa4 {-1.21/17 120} 27. Rf2 {214} Rxd6 {-0.71/18 172} 28. cxd6 {0} Bd7 {
-1.03/19 160} 29. Qe4 {0} Re8 {-1.03/18 112} 30. Kh1 {(Rg2) 100} Qxd6 {
-1.13/18 204} 31. Rfe2 {(Rg1) 113} g5 {-1.53/17 101} 32. Ra1 {(Kg1) 91} a4 {
-1.57/17 159} 33. Rae1 {(Rg1) 102} Kg7 {-1.72/18 155} 34. Kg1 {297} g4 {
-1.75/18 0} 35. Rc2 {(Qd3) 189} Re7 {-2.00/17 134} 36. Rf2 {(Rd2) 0} Kf6 {
-2.10/17 108} 37. Qc4 {(Qe3) 121} Kg5 {-2.29/17 143} 38. Ref1 {145} b5 {
-2.49/18 0} 39. Qe2 {73} b4 {-2.41/17 89} 40. Qc2 {(Qc4) 169} Qxd5 {-3.01/16 88
} 41. Rd1 {119} Qc6 {-3.42/17 0} 42. Qd2 {(Qxc6) 0} Qb6 {KingOfEngines,Zappa
Zanzibar X64 resigns (Lag: Av=1.52s, max=15.1s) -3.57/17 120} 0-1
Last edited by Harvey Williamson on Sun Aug 12, 2007 4:55 pm, edited 2 times in total.
User avatar
mackgra
Member
Posts: 93
Joined: Sat Aug 04, 2007 9:18 pm
Location: Sheffield, England

Post by mackgra »

Hi everyone,

I followed this game real time and yet again Hiarcs understood the unbalanced position much better then Zap.

As i commented to Harvey at the time Hiarcs' Bishop and pawn (and 2nd pawn a few moves later) would stand an excellent chance against zaps extra rook. Looking at the piece combination for white the two rooks and queen had a certain piece redundency about them.
I also mentioned that as long as Hiarcs didn't allow an exchange of rooks (thus trading off its last rook) it had excellent chances as was later proved.

One extra bit of info i have found out is that in exchange up games the person with the exchange generally benefits from exchanging queens. I wonder if anyone would like to check the databases for me!

Cheers,

Graham
User avatar
Ted Summers
Member
Posts: 269
Joined: Tue Jul 31, 2007 2:20 pm
Location: Marietta, GA
Contact:

Knowledge is Power

Post by Ted Summers »

mackgra wrote:Hi everyone,

I followed this game real time and yet again Hiarcs understood the unbalanced position much better then Zap.

As i commented to Harvey at the time Hiarcs' Bishop and pawn (and 2nd pawn a few moves later) would stand an excellent chance against zaps extra rook. Looking at the piece combination for white the two rooks and queen had a certain piece redundency about them.
I also mentioned that as long as Hiarcs didn't allow an exchange of rooks (thus trading off its last rook) it had excellent chances as was later proved.

One extra bit of info i have found out is that in exchange up games the person with the exchange generally benefits from exchanging queens. I wonder if anyone would like to check the databases for me!

Cheers,

Graham
Thanks Harvey and hi Graham,

I thought about upgrading to the MP version of Hiarcs but with results like these from the beta version, I think I will wait for the new MP version of Hiarcs 12. Clearly I see Hiarcs as the only program with a chance to over take Rybka. Interesting that both progams are knowledge based and have been from the start. Goes a long way to proving the statement that "Knowledge is Power".
User avatar
mackgra
Member
Posts: 93
Joined: Sat Aug 04, 2007 9:18 pm
Location: Sheffield, England

Post by mackgra »

Hi Ted,

You're absolutely right its the evaluation of any programm that is key here.
The more knowledge that is put into Hiarcs (assuming it is correct knowledge) the stronger and stronger it will get.

It is interesting that even some GM's say rules and guidelines don't matter only calculation does, but even calculation has to be based on some form of evaluation.

"Evaluation trumps calculation" Kasparov 2007, How life imitates chess
David_W
Member
Posts: 4
Joined: Wed Aug 01, 2007 12:02 am

Post by David_W »

My understanding is that both Hiarcs and Rybka are heavily knowledge-based, whereas Fritz, Junior and Shredder are more like bean counters (which might, in part, explain their very high nodes-per-second counts). ChessBase really likes to overemphasize Fritz's mega-high node count whenever it takes part in ChessBase-sponsored comp-human or comp-comp matches, because they think it matters when in fact we know that the NPS isn't really so important.

Still, it's interesting that Chess Tiger author, Christophe Theron, has repeatedly accused Vas Rajlich of deliberately fiddling Rybka's NPS count (which seems to be the lowest in the business!) in order to make everyone think that Rybka must be processing tons and tons of chess knowledge.

David.
User avatar
cornetmike
Member
Posts: 30
Joined: Fri Aug 10, 2007 4:22 am
Location: Hemet, California
Contact:

Knowledge is Power

Post by cornetmike »

My favorite program to play against is Chess System Tal II, also knowledge based. It averages 8 to 12 thousand NPS and plays a far more interesting game than the "bean counters." And with a an ELO of 2300, it's quite hard to beat... :D

regards
Mikey
“They must find it difficult – those who take authority as the truth, rather than truth as the authority.” - Gerald Massey, Egyptologist
Uri Blass
Member
Posts: 82
Joined: Sun Aug 12, 2007 1:40 pm

Post by Uri Blass »

David_W wrote:My understanding is that both Hiarcs and Rybka are heavily knowledge-based, whereas Fritz, Junior and Shredder are more like bean counters (which might, in part, explain their very high nodes-per-second counts). ChessBase really likes to overemphasize Fritz's mega-high node count whenever it takes part in ChessBase-sponsored comp-human or comp-comp matches, because they think it matters when in fact we know that the NPS isn't really so important.

Still, it's interesting that Chess Tiger author, Christophe Theron, has repeatedly accused Vas Rajlich of deliberately fiddling Rybka's NPS count (which seems to be the lowest in the business!) in order to make everyone think that Rybka must be processing tons and tons of chess knowledge.

David.
Some notes:

1)My understanding is that nodes per second is meaningless because it is possible to count nodes in different ways(for example you can count all moves that you make in the tree and you can count only positions that you do full evaluation of them when positions when you do cutoff based on hash tables and read evaluation from the hashtables are not counted)

2)I think that christophe was right and at least if you compare rybka beta with strelka1.8UCI that was designed to behave almost in the same way as rybka you can see that they show almost the same behaviour except the fact that strelka report higher number of nodes and bigger depthes.

3)I think that search knowledge(knowledge which lines to search is important) and knowledge is not only about evaluation function.

Uri
User avatar
Ted Summers
Member
Posts: 269
Joined: Tue Jul 31, 2007 2:20 pm
Location: Marietta, GA
Contact:

Post by Ted Summers »

Uri Blass wrote:3)I think that search knowledge(knowledge which lines to search is important) and knowledge is not only about evaluation function.
Uri
Uri, I understand your point here a little, but can you define search knowledge in a little more detail. Ahh, never mind I think I get what you mean.

Thanks
Ernst
Member
Posts: 9
Joined: Tue Jul 31, 2007 7:05 pm

Re: Another Fascinating Game

Post by Ernst »

Harvey Williamson wrote:Hi All,

Following on from the fascinating game http://hiarcs.net/forums/viewtopic.php?t=102

Our latest version Hiarcs 11.55e just scored a very nice win on Playchess. Hiarcs was playing on my 8x2.66 intel machine. However Zap operated by Suj was on his 32 core machine. He tells me it is the first time Zappa has ever lost on this machine.

Best Wishes,

Harvey
Harvey, could you please post the opponents's machine info at the beginning of the game? It is my believe that this 32 core story is bullocks.

Ernst.
User avatar
Harvey Williamson
Site Admin
Posts: 6079
Joined: Sun Jul 29, 2007 6:57 am
Location: Media City, UK
Contact:

Re: Another Fascinating Game

Post by Harvey Williamson »

Ernst wrote:
Harvey Williamson wrote:Hi All,

Following on from the fascinating game http://hiarcs.net/forums/viewtopic.php?t=102

Our latest version Hiarcs 11.55e just scored a very nice win on Playchess. Hiarcs was playing on my 8x2.66 intel machine. However Zap operated by Suj was on his 32 core machine. He tells me it is the first time Zappa has ever lost on this machine.

Best Wishes,

Harvey

Harvey, could you please post the opponents's machine info at the beginning of the game? It is my believe that this 32 core story is bullocks.

Ernst.

Hi Ernst,

It maybe bullocks but the evaluations or machine details are not kibitzed in these games. As(so I am told) the gui sits on 1 machine while the engine plays on the server.

Best Wishes,

Harvey
BigBen

Post by BigBen »

Hi,
If Ernst goes and reads about Freestyle 6 and checks out the Mission Control entry from Suj, his questions and doubts will be answered :)

Regards
Ernst
Member
Posts: 9
Joined: Tue Jul 31, 2007 7:05 pm

Post by Ernst »

I know the remote story and read the Freestyle report, and yet there is no prove of this machine, only his word.

Mission Control has a computer rating of 2657 and a and a computer slow rating of 2491.

He doesn't show his eval, what does he have to hide?

Ernst.
BigBen

Post by BigBen »

Hi,
Suj does not always use the 32 core machine, he also has other nicks on the playchess server. It maybe the centaur results are most relevant.

It maybe that the way he has the remote machine setup that no expected moves\evaluations are sent.

I will send Suj a message with a link to this thread so he has the oppotunity to answer

Regards
Tony
suj
Member
Posts: 21
Joined: Wed Aug 15, 2007 5:47 pm

Post by suj »

Try using a rybka2.ctg book and see what rating one will get-Rybka 232a is anyway crap above 4 cores.Apart from that I dont play for elo and i dont care about it.

As for this-I would possibly say this is my first loss on 32 way on zappa.I dont care if anyone thinks if am using a laptop or any machine its for them to believe what I use and why would I accept a loss on a 32 way if I used say a quad or an octa.

As for hiarcs team-Kudos on the great game I loved the loss and for sure it has come a long way.

I would also like to thank Harvey and Mark for the very efficient support to all customers.

Thanks

Suj
User avatar
Harvey Williamson
Site Admin
Posts: 6079
Joined: Sun Jul 29, 2007 6:57 am
Location: Media City, UK
Contact:

Post by Harvey Williamson »

Hi Suj,

Welcome to the Hiarcs forum :)

Harvey
Post Reply