Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Hiarcs vs Rybka
- - By Gryphus (**) Date 2007-09-11 16:51
Hiarcs 11.1 mp UCI vs Rybka 232a mp 32bit

I turned Hiarcs vs Rybka 10 games, 4mim without books.
7+  2=  1- for Hiarcs.
Why?
Parent - By Vempele (Silver) [fi] Date 2007-09-11 17:16 Edited 2007-09-11 17:20
Statistical variance. Could be a bunch of other reasons, but you left out pretty much all the relevant information:

Ponder?
Number of CPUs?
How are we supposed to deduce why it happened if you don't even post the games?

Obviously, you could have just searched for similar topics - there are several.
Parent - By Hetman (*****) Date 2007-09-11 17:51
Hiarcs was using antichess strategy :-) :-).
Rgds
Hetman
Parent - By Aquaplex [us] Date 2007-09-11 18:27
I ran into a similar problem.  For me it wasn't that rybka was losing, it was that it was losing because of time.  It would take forever to make moves even with a book.  So what i did was turn down the process % usage from 100 to something like 50 and decreased hash size.  That worked.  Check to make sure u're not losing games on time.  I ran rybka 2.3.2a mp 32bit vs. hiarcs 11.1 and rybka won plenty.
Parent - By Harvey Williamson (*****) Date 2007-09-11 18:30
It is possible that as you played without books that the same game repeated several times. Are you sure Hiarcs played without a book - did you untick own book in the Hiarcs parameters?

But I like the result :-)

Harvey
Parent - - By Gryphus (**) Date 2007-09-11 19:45 Edited 2007-09-11 19:53
My observation is that in the first 10 games, there is a good advantage for Hiarcs that decreases in the following games.
I already tested three times, in three match's of 30 games.
In the first 10 games, Hiarcs wins, 7 + 2 = 1 -. In 30 games Rybka wins, 9 + 9 = 12 -.
Only a statistical coincidence?
Already with the books, Rybka wins 90% of the games always.
AMD core duo 3800+; hash 128MB; engines parameterns deatful.
Parent - - By Vempele (Silver) [fi] Date 2007-09-12 05:08

> Only a statistical coincidence?


Or something changed and accidentally fixed the problem.

> Already with the books, Rybka wins 90% of the games always.


There's definitely something wrong. Rybka is nowhere near strong enough to score that well.
Parent - - By Gaмßito (****) [cr] Date 2007-09-12 05:53 Edited 2007-09-12 05:57
Rybka is nowhere near strong enough to score that well.

Of course that it can.

Look here: http://www.husvankempen.de/nunn/Replay/hiarcs112rybka232a.htm

Rybka just, demolished, squashed, exploded and broke Hiarcs 11 in a slow match of 50 games! :-)
It wins by the HUGE margin of: 40.5 - 9.5!

When I saw this result, I could not believe it.  Honestly, it is one of the most surprising results that I have seen in my life, since I have memory.

Thank you for this monster Vas!

Regards,
Gambito.
Parent - By Gaмßito (****) [cr] Date 2007-09-12 06:02
2.3.2a also won against Hiarcs 11.2 with a similar result.  An overwhelming victory! 
 
Look here: http://www.husvankempen.de/nunn/Replay/table.htm

Regards,
Gambito.
Parent - - By Vempele (Silver) [fi] Date 2007-09-12 08:12

>Rybka just, demolished, squashed, exploded and broke Hiarcs 11 in a slow match of 50 games! :-)


It wins by the HUGE margin of: 40.5 - 9.5!

Rybka won 34 of those games, that's only 68%.

>2.3.2a also won against Hiarcs 11.2 with a similar result.  An overwhelming victory!


Rybka won only 29 games (58%). That's utterly ABYSMAL compared to winning 90% (and most likely drawing the other 5).

In other words, these results prove my point: Rybka doesn't always win more than 90% of her games against Hiarcs. She'd have to be around 700 Elo stronger than Hiarcs to even do it regularly.
Parent - By Gaмßito (****) [cr] Date 2007-09-12 08:35
Yes, The beatings that Rybka gave to Hiarcs, was not from 90%. You are right Vempele.

90% is already too much. But, anyway 68% and 58% is really a brilliant score. A very wonderful result in my opinion.

Regards,
Gambito.
Parent - By Gryphus (**) Date 2007-09-12 11:11
Maybe Vampele you can be right.
The book that use in Rybka is HS Rybka that is very strong goes.
Hiarcs plays with the book Hiarcs 10 .ctg (it is not the original book of the version 11.1)
The one that I found strange is that the first 10 games there are releases advantage of Hiarcs (without books) and this result almost changes for the inverse the measure that the amount of games increases. Would you have some theory for this?
Thank you.
Parent - - By InspectorGadget (****) [za] Date 2007-09-12 05:38
I once tried Fritz 10 vs Rybka232a without books, anyway, it was not a lot of games. Fritz got the advantage in the early stages of the game. It seemed to me that Rybka does not open the game very well without a book. That is my oppinion, you are welcome to differ. :-)
Parent - - By masomusic (**) [nl] Date 2007-09-12 18:56
...with or without a book both Hiarcs 11.x and Fritz 10 get killed before the endgame most of the time. It's obvious as Hiarcs 11 (the worst upgrade for Hiarcs in history so far) and Fritz are kinda very tactic engines and Rybka rather exploits her "chessknowledge".
Parent - By Harvey Williamson (*****) Date 2007-09-12 19:04
Is Hiarcs 11 worse than Hiarcs 10? There have been several releases of many engines which have shown very small increases. If you mean Hiarcs 11.1 to 11.2 then it was only advertised as 8 CPU support and some bug fixes for Vista and was free.
Up Topic Rybka Support & Discussion / Rybka Discussion / Hiarcs vs Rybka

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill