Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / The perfect opening for Rybka
- - By Udav18 (**) Date 2007-06-19 12:57
Hallo,I started a few days ago a discussion about the perfect opening for Rybka.What was the main point of this discussion?
well,the discussion digresst in something much bigger.Maybe it created the idea ,how to improve chessengines,how to see immediately where their weaknesses and their strengths are.
It might be that this short discussion may develop in something real big!
notctiferius made a table,where he put some openings,and the results of a few engines in this openings. http://home.interact.se/~w100107/openings_newer.htm
In such kind of tables ,if they would be improved,by collecting many informations from chessfriends all over the world,I think there might be hiding the truth about many engines.
Also Vas himself tried it :
" Actually, I've done a bit of work on this myself. It's easy - just take a ton of games where Rybka is white, import them into some tree format (fe. ctg), and browse the tree, digging deeper into branches with bad scores for white. Then, repeat with black.

I don't remember really getting any meaningful conclusions out of it.

Vas
"

Well, my opinion to that is,that a tree doesnt really show the main points!
I think you have only to look at this factors:
Engine,Opening,Result,Games,Number of moves that were made.
Well I tried it myself at home.And collected a few informations.
It helped me to find out the roughly weaknesses and strenghts of Fritz10.
E.g. what openings this programm should play and what not,so I created a tree for it and started a match between Rybka2.3.2 and Fritz 10(53 games /4 min ).Result: 28-25 for Rybka2.3.2
Fritz lost,but compared to the other matches i have made so far,Fritz improved really amazingly.For example  Rybka crushed Fritz 10 with 16-4 ! and 23-7!

So ,who wants to improve or create such tables may help maybe to understand engines much better.

Ed
Parent - - By noctiferus (***) Date 2007-06-19 14:49
"notctiferius made a table"
Hi, Udav,
sorry not to have done it myself, I simply found it looking around in internet
Parent - - By Udav18 (**) Date 2007-06-19 15:04
Oh sorry,:-) I missunderstood you,but anyway thanks!!
Parent - - By Udav18 (**) Date 2007-06-20 17:38
Amazing!!!!!I made a new tree for Fritz10,with openings Fritz can play really well and started a match with 50 games Blitz (5min).
The Result :26-24 for Fritz10!!!!!!!!    Fritz10 - Rybka2.3.2  +15 =21 -14
Really unbelievable!
Parent - - By Udav18 (**) Date 2007-06-20 17:43
At first i was happy that Fritz10 was able to hold the match in balance ,the first 9 games endet up with a draw,but than Rybka2.3.2 won 3 games in a row,so I thought game over.
But than Fritz10 started to win more and more games and only in game 50 Fritz10 decided the game with a draw and the better score!
Parent - - By Rybka_King88 (*) Date 2007-06-20 18:23
So your saying to use Fritz10.ctg!?!?            :-o
Parent - By Fulcrum2000 (****) Date 2007-06-20 18:36
No, he made a new book himself.
Parent - By turbojuice1122 (Gold) Date 2007-06-20 19:24
To make these conditions "fair", you would need to do one or both of two things: make a book for Rybka with openings that Rybka plays really well and/or make a book for Rybka with openings that Fritz 10 plays poorly.  If you do both, then the book for Fritz 10 should also include openings that Rybka plays poorly.
Parent - - By ok373 (**) Date 2007-06-20 20:10
Yes,you are right. Rybka plays very bad aggressive opening and very well positional ones.
There for it will be a very good idea to create a special book for Rybka.
Work on it a few days and tell us if you increased Rybka's performance by creating a special book for it.
Parent - - By Udav18 (**) Date 2007-06-20 20:46
Well,I will try!But the main point was not to make conditions equal,and the match between fritz10 and Rybka 2.3.2 wasnt the main idea!The main point of the results is,the results just confirmed my expectations about such tables.The book was also not the main idea.The importent thing is,that i quickly could find out flaccidities and strengths of an engine only using such tables.So as I said,when many chessfriends would help to develop such tables,they might help to improve engines,because an engine-proggrammer as Vas wouldnt need to waste much time to find out the truth about his engine and he wouldnt ask himself where the problem at his engine is,but rather how to eliminate this problem.
In my opinion thats the main point.But of course such tables could also be used for creating books.
Parent - - By billyraybar (***) Date 2007-06-21 04:52
The abosolute worst as black has to be the Scandanavian regardless of variation.  Then again, I don't see many world class players play it much anymore.  Maybe there is good reason for that.
Parent - - By Kapaun (****) Date 2007-06-21 08:12 Upvotes 1
I don't know much about Scandinavian, but I'm a long time wondering now why Black would dare to play Dutch...
Parent - - By Alkelele (****) Date 2007-06-21 08:41
Not to mention 1.d4. It is a complete mystery to me why even GMs keep blundering with that move. Already Fischer told us that 1.e4 is best.
Parent - - By ok373 (**) Date 2007-06-21 10:23
no,1.e4 is not for sure the best move.
1.d4 is a good move. You sould fear from the Grinfeld and the King's Indian(both are very strong).
Against 1.e4 you have the Caro-kann,e5 and the best variation in chess the Najdorf...

They are about equal.
Parent - - By Udav18 (**) Date 2007-06-21 15:12
Hallo I finished a interessting book for Rybka and tested it.It was really hard to find good openings for Rybka,much difficult than for Fritz 10,but I tried it .E.g. I wanted that rybka does play allways 1.c4 and play in every Opening not the best move!,but rather a move which leads to a closed position,with many stratigical ideas and a solid advantage.
I started a match between fritz10(with the book,helped him to beat rybka)and rybka2.3.2 ; 40 games(5min) and the result was 23-17 for Rybka.So the point of this match was again,that I was able to improve Rybka by creating a book using only such tables!
It was really funny to see that both engines tried to find the best ways to play their book moves,every engine wanted to allure the other in its "own opening" ,i could often see that the engine,which could manage to do that has won.Of course the Engine who had white was able to do that in most of the cases,but fritz lost 3 games in its "own opening".
Parent - By Uly (Gold) Date 2007-06-22 23:47
Hey, very interesting, could you repeat the test but now with this enhanced Rybka against normal Fritz?

I'm interested in knowing if Rybka can beat Fritz 50-0 by using this method (Or, I want to compare it with her previous result of 39-11).
Up Topic Rybka Support & Discussion / Rybka Discussion / The perfect opening for Rybka

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill