![[in]](/mwf/v12/flags/in.png "India")
A blitz game is played between two computers with same configuration.
the opening book used is rybka 3 book in fritz 11 interface.
how should one rybka should be configured so that it can beat the other rybka with default settings.
plz help.
thank you in advance.
the opening book used is rybka 3 book in fritz 11 interface.
how should one rybka should be configured so that it can beat the other rybka with default settings.
plz help.
thank you in advance.
Sounds impossible. For just ONE normal game, there will never be a guarantee anyway. With two basically identical systems, you cannot expect more than 50% in the long run (except with some statistical luck).
I don't know if small adjustments have been found by users, which may improve the default settings. In general (for Rybka and for most engines) the default settings are the strongest, or at least they are intended to be strongest.
I would only make sure that no configuration mistakes are made. Typical mistakes would be: too large hash size, no book loaded, bad book setting, wrong time control (level) setting... Since there is so much adjustable, it is actually not very easy for a new user to get a complete overview. I can only recommend to look through all menus, and try to find the "normal" values each.
I would use Rybka's defaults, but make sure that the engine uses all cores, for example 4 threads if you have a quad core cpu. In that case, while Rybka calculates you should get ~99% cpu load, which you can check with the task manager or another process tool.
(New users often ask for tips here, but do not tell us their exact computer configuration like cpu and RAM size etc., so it is difficult to give specific recommendations.)
Something which is not automatically available are the Nalimov tablebases to support the endgame calculations. I assume you have got some tbs. with your Fritz 11 (search for a directory like ../TB on the DVD). The F1 documentation will give you information about how to install and use these. You need to put them either on harddisk, or even better, on a flash memory device. I strongly recommend that you do NOT try to download gigabytes of 6-men tables, for now. Just go for all 3/4-men tables to start with which probably are on the Fritz DVD (should be 70 files with ~30 MB), and maybe some 5-piece tables later. The impact on strength is not dramatical, but it could give you "the edge", at least in some games. Useful information related to tablebases:
http://www.horizonchess.com/FAQ/Winboard/egtb.html
There are also alternative opening books available for download, some of which are created by engine competition specialists. For Fritz, you need the CTG file format (but these books consist of THREE *.ct? files, make sure to have all three of them each). If have seen results that some of these may score slightly better than the (default) commercial opening books, at least under some circumstances. But I do not have much detail knowledge about these books intended for "super tuned" competition. For example, the Perfect 15.ctg:
http://www.rybkachess.com/index.php?auswahl=Downloads
Anyway, if your match is really one game only, the simple question if you get White or Black will be more important than these things mentioned above. :-) At least four games with alternating colors would be better, to reduce the luck factor a bit.
I don't know if small adjustments have been found by users, which may improve the default settings. In general (for Rybka and for most engines) the default settings are the strongest, or at least they are intended to be strongest.
I would only make sure that no configuration mistakes are made. Typical mistakes would be: too large hash size, no book loaded, bad book setting, wrong time control (level) setting... Since there is so much adjustable, it is actually not very easy for a new user to get a complete overview. I can only recommend to look through all menus, and try to find the "normal" values each.
I would use Rybka's defaults, but make sure that the engine uses all cores, for example 4 threads if you have a quad core cpu. In that case, while Rybka calculates you should get ~99% cpu load, which you can check with the task manager or another process tool.
(New users often ask for tips here, but do not tell us their exact computer configuration like cpu and RAM size etc., so it is difficult to give specific recommendations.)
Something which is not automatically available are the Nalimov tablebases to support the endgame calculations. I assume you have got some tbs. with your Fritz 11 (search for a directory like ../TB on the DVD). The F1 documentation will give you information about how to install and use these. You need to put them either on harddisk, or even better, on a flash memory device. I strongly recommend that you do NOT try to download gigabytes of 6-men tables, for now. Just go for all 3/4-men tables to start with which probably are on the Fritz DVD (should be 70 files with ~30 MB), and maybe some 5-piece tables later. The impact on strength is not dramatical, but it could give you "the edge", at least in some games. Useful information related to tablebases:
http://www.horizonchess.com/FAQ/Winboard/egtb.html
There are also alternative opening books available for download, some of which are created by engine competition specialists. For Fritz, you need the CTG file format (but these books consist of THREE *.ct? files, make sure to have all three of them each). If have seen results that some of these may score slightly better than the (default) commercial opening books, at least under some circumstances. But I do not have much detail knowledge about these books intended for "super tuned" competition. For example, the Perfect 15.ctg:
http://www.rybkachess.com/index.php?auswahl=Downloads
Anyway, if your match is really one game only, the simple question if you get White or Black will be more important than these things mentioned above. :-) At least four games with alternating colors would be better, to reduce the luck factor a bit.
![[au]](/mwf/v12/flags/au.png "Australia")
I was wondering if say i had 6 man table base installed, how would this help. Would it only be helpful once there is 6 pieces or would the table base effect the play before there is 6 pieces. for example say there is 9 pieces on the board, will the table base help Rybka avoid an exchange which will lead to a 6man tablebase draw?
It is a bit hard to explain but i hope you got my meaning.
It is a bit hard to explain but i hope you got my meaning.
![[pr]](/mwf/v12/flags/pr.png "Puerto Rico")
If the draw is the best outcome of the game why will rybka negate that option and yes tablebase installed should help analysis and game play but the question is by how much?
Tablebases can come into effect before there are 6 pieces, only. Most engines access tablebases in the search (sometimes adjustable how intensive they do it). With very big tables, there is also a downside: They take memory and slow down the search. That is why fast storage devices like flash memory or SD disks are recommended. In addition to that, I always say maximum is not optimum and I do not even use all 5-piece tbs.
To test this and to see better how it works in practise (which is better than long explanations), you don't need 6-piece tables though.
To test 5-piece tbs. in search:
ONLY 61...Rxc6 draws. Both Kxh1 and Rd5+ lose against best play.
To test 4-piece tbs. in search:
ONLY 1.Bg3 draws. (Of course, this example only makes sense if you don't have this 5-piece table in use.)
For these tests, I recommend multi-pv mode with two or three vars, to see clearly that the next alternatives are evaluated correctly (= much worse).
P.S.: If you have a Fritz GUI and you run an engine which does not access the tbs. in the search (like WinFinder), then by default, the interface will switch to tablebases if a tbs. position is reached in the game. If that happens, the engine will actually not play anymore and the interface will continue from the tables.
To test this and to see better how it works in practise (which is better than long explanations), you don't need 6-piece tables though.
To test 5-piece tbs. in search:
8/8/2P3P1/2r5/8/8/6k1/3K3R b - - 0 61
ONLY 61...Rxc6 draws. Both Kxh1 and Rd5+ lose against best play.
To test 4-piece tbs. in search:
8/k7/8/2R5/8/4q3/8/4B2K w - - 0 1
ONLY 1.Bg3 draws. (Of course, this example only makes sense if you don't have this 5-piece table in use.)
For these tests, I recommend multi-pv mode with two or three vars, to see clearly that the next alternatives are evaluated correctly (= much worse).
P.S.: If you have a Fritz GUI and you run an engine which does not access the tbs. in the search (like WinFinder), then by default, the interface will switch to tablebases if a tbs. position is reached in the game. If that happens, the engine will actually not play anymore and the interface will continue from the tables.
![[mx]](/mwf/v12/flags/mx.png "Mexico")
Against an equal opponent, I'd set Contempt Play to 0 instead of 15. 15 helps against weaker engines but that doesn't apply here.
Any one has the answer to my question?
which book is stronger? perfect 15 or rybka 3........
how to stay in main line of book till the end?
how to stay in main line of book till the end?
> how to stay in main line of book till the end?
See the F1 help for book options, or book settings (F4). Usually, it will not be possible to force a particular single main line, except the book is designed very restrictive, without variety. A main line is probably the line the opponent is prepared against, the best (except you have a good novelty). A book which plays only the same one move each, in any opening position, is very predictable.
My book philosophy for computer chess is exactly the opposite (big variety, and "normal", not "optimized" book setting), but I do not go for top machine room ratings or high level competition. - Again, for ONE game there is no guarantee whatsoever, anyway.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill