Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Computer Chess / How good is the Cerebellum book with Stockfish?
1 2 Previous Next  
Parent - - By Ozymandias (****) [es] Date 2016-08-21 14:31

>I forced it to play the French Defence.


People "testing" this book, are doing the strangest things ever.
Parent - - By shrapnel (***) [in] Date 2016-08-22 15:25
Hi Ozy
         Maybe it appears strange to you because I'm not really a Tester, but a Gamer.
I see Books and Engines from a Gamer's competitive point of view ; am not really comfortable with the rather abstract Tester's POV, which to tell the truth appears rather strange to ME !
    I once read an Article which compared Chess to War.
Victory goes to the warrior who is able to surprise his opponent with unexpected actions ; same applies to chess to a large extent.
So, when  Venator says that " Defences in my book are Ruy Lopez, Sicilian and Caro-Kann" you will understand that his Statement seems rather tame and ridiculous to a Gamer's mindset.
   For example, just to take this very case, no matter how good his Book is in the above Defences, there is absolutely no way an opponent can force me to play any of these Defences as Black !
To 1. e4 I can always reply 1...e6 and to 2.d4 I will reply 2....d5 and Venator can say "Bye Bye " to his dear Ruy Lopez, Sicilian and Caro- Kann !
And anyone who underestimates the French Defence, does so at his own peril.
It is perfectly playable as Black, provided one is using a fast i7 like mine, or a dual-xeon.
I use any standard Book for just 10-12 moves and then let my PC takeover.
This way, I avoid playing the rather over-analyzed Ruy Lopez, Sicilian and Caro-Kann.
The best part is my opponent also comes out of Book soon, which is just what I want !
And as I have around 2 minutes per move on C2P, I rarely face any problems from anyone, mo matter how good his Book is.
Hope you see my POV .
Regards
Parent - - By Ozymandias (****) [es] Date 2016-08-22 19:26
It doesn't matter if you consider yourself a tester or not, if you're testing books, you should adhere to the principle of letting them play, without restrictions. Imposing a particular defence makes sense for engines, not for books. You can play a ton of games, and extract those played in the French, as long as both books chose it. Anything else can be considered tampering.
In this particular case, if one book doesn't play the French with black, the most you'll get, is to see how it handles the white pieces against it, if the other book has the French in its repertoire. You should be content with that.
Parent - - By shrapnel (***) [in] Date 2016-08-23 05:36
Sorry, but it DOES matter a lot whether I consider  myself a Tester or a Gamer. The two are very different from each other.
Let me explain with the Combat equivalent of Chess which I mentioned earlier.
Suppose my opponent is an Expert Swordsman.
Now, as a Tester, I too should pick up a sword and fight him. If I lose, too bad.
Now, as a Gamer/Warrior/True Chess Player my only objective is to WIN, by ANY means.
(As any chess player worth his salt will tell you, Checkmate is all that matters ; how you get there doesn't really matter.
It doesn't matter how prettily you play, if you lose in the end.)
I may choose to fight him with a sword, if I think I can beat him ; but if I'm not sure, there is nothing stopping me from using a Bow and Arrow and shooting him from a distance OR Spearing him if I'm good with Spears OR taking a Battleaxe to him if I'm good with Battleaxes.
The aim is to have my opponent dead and me standing, after the fight.
     Similarly, when my  computer chess opponent, who has a weak or average computer, says he has a very good Book for Sicilian, Ruy Lopez or Caro-Kann, it would be very foolish of me to oblige him by playing those Defenses as Black !
Rather, it is only common sense, as a Gamer, that I would choose a Defense that will get my opponent all hot and bothered and uncomfortable, forcing him to come out of Book soon, thereby increasing my chances for a win, which is ALL that matters. This is in conformity with the true philosophy of Chess and War.
Of course, if I was a Tester, I would also use a Book which is strong in Sicilian, Ruy Lopez or Caro-Kann and HOPE that my Book is stronger.
Hope you get my Drift.
Parent - - By Ozymandias (****) [es] Date 2016-08-23 07:11
I don't think I do, is there or isn't there a difference, between a chess game and a test chess game? For me, there is, if you don't acknowledge it, there's nothing I can tell you that will help clarifying the matter.
Parent - By shrapnel (***) [in] Date 2016-08-24 15:38
Indeed, let us agree to disagree and leave it at that, as I see that I have been wasting my breath, so to speak.
Parent - By Antares (****) [de] Date 2018-03-08 16:50
I'm missing this Shrapnel-guy... :grin:
Parent - By Peter Grayson (****) [gb] Date 2016-08-21 10:55

> I ran Brainfish/Cerebellum on 1 computer against Asmfish/Noomen.ctg on my 2nd computer, in a French Defence.


Quote from the Brainfish Website ...

"Brain Fish can of course handle also moves or openings which are never played by the book, for example 1. ..e6. " Therefore your argument is at odds with the very same opening book you argue is stronger. Why should the Noomen book be expected to play the line when the Cerebellum author(s) expressly seek to avoid it too?

In choosing to force the book to play a sub-optimal line then a sub-optimal performance must be expected too.

PeterG
Parent - - By Fritz1987 [us] Date 2018-03-07 21:33
Any chance you still have this book? Been looking for it for a while.
Parent - - By Eelco de Groot (***) Date 2018-03-08 10:33 Edited 2018-03-08 15:43
I suppose you mean the Vesely book? I think I still have it on my old computer. But it is probably too large to post it here. I have no idea what the maximum size of file is that I can upload. Maybe Felix knows that. I suppose Peter and others still have it. Probably too large for e-mail as well.

Or did you mean Jeroen's book?

Update: Sorry, I'm afraid Vesely's book is not on my computer. I may have removed it from harddisk to create some space, not thinking this one would live longer. Unfortunately the computer I moved it to crashed and I don't know if the harddisk still lives. So I don't seem to have it, unless I can revive that disk some day :fat:
Parent - - By Fritz1987 [us] Date 2018-03-08 16:47
Thank you for your efforts! I have done a lot research and I think I found it! In the progress of downloading now.
Parent - By Antares (****) [de] Date 2018-03-08 16:50
AFAIK its just an "optimized" (=pirated) and very overloaded (imported every game known to mankind) copy of Mr. Noomen's Rybka3-book anyways...
Parent - - By shrapnel (***) [in] Date 2016-08-10 06:23
Hi Peter
           I think you are underestimating Cerebellum.
I have found it a great help in my online Matches against even strong players using the latest Commercial Book.
You may have found a weak line, but NONE of my many opponents have found a chink in my Armour (Cerebellum).
In any case, Cerebellum is being updated regularly and I doubt if the weakness you found, exists any longer.
Cerebellum/Brainfish is a real BOON to online players who otherwise have to waste a lot of time updating and searching for the strongest ctg Books.
Provided the Rates are not too high , I wouldn't mind paying a subscription as it is really useful and convenient.
Parent - By Ozymandias (****) [es] Date 2016-08-10 10:36

>I have found it a great help in my online Matches against even strong players using the latest Commercial Book.


They also say that Sicilian King 3 wins most tours at IC, and yet, how many Engine Masters do they have under their belt? How did the players using Cerebellum, fared in the last one?
Parent - By Peter Grayson (****) [gb] Date 2016-08-10 11:42
In terms of implementation the Cerebellum concept is some way behind the HIARCS engine's internal book that does tune for ongoing results. Neither system influences the engine during analysis but in the Brainfish case, without a .ctg book it offers no more than the equivalent Stockfish + LP engine would give.

I think the Brainfish + Cerebellum_Light combination is OK, especially for those who do not have the time to adjust their own database and move mark their book, but it does not offer the fexibility such as the Chessbase .ctg and database system offers. That could change if the Cerebellum software is released to enable the end user to construct their own book from their own database. How time consuming that may be could be another matter. I am not familiar with the Aquarium system.

Also be aware there are some short-comings in tuning the book for the Stockfish engine. It raises the question: for which engine release is it tuned? Small code changes with time can significantly impact on the direction the engine will take and of course there are still some holes in its opening knowledge.

Personally I preferred to use my own games database and build and move mark my own book rather than rely on someone else's construction. The point of this exercise was to understand the "New Concept" but in reality it has been done before and offers no more than can be achieved by other more user accessible systems for the enthusiast who prefer to be in control.

If you are obtaining benefit from the Cerebellum_Light book that is good but I looked at it from the perspective of how it compares with tools available for the enthusiast. That is my assessment of it. I would not be surprised if the full package, not just the_Light book, is released commercially soon and whether or not the present supplier of the book will continue to offer a prebuilt book remains to be seen.

PeterG
- - By Peter Grayson (****) [gb] Date 2016-08-08 15:21
Here is the problem with the Stockfish (Brainfish) + Cerebellum_Light book combination. Although described as a new concept the externally tweaking of the book may come under that category but the combination has all the flaws contained in the lack of learning by result inflexibility.

Consequently if opponent finds a bad line the engine + book combination gets stuck in a losing mode that was the case in the match. The Brainfish + Vesely book played a winning line in the Sicilan poisoned pawn to which the Cerebellum_light book had no answer with the consequence of a bad run of losses for the Cerebellum with no way to get out of the bad line ...
Parent - - By MarshallArts (***) [us] Date 2016-08-09 00:27
The problem I see is that there's no fluid, on-the-fly book learning,
such as "oh, this line is really bad, let me not play it".

:roll:
Parent - By Peter Grayson (****) [gb] Date 2016-08-09 00:46

> The problem I see is that there's no fluid, on-the-fly book learning,
> such as "oh, this line is really bad, let me not play it".


Yes that does indeed look to be the major drawback.
Parent - By Venator (Silver) [nl] Date 2016-08-22 14:45
Hi Peter,

I agree with your conclusion. Not only the Cerebellum book repeats losing lines, but IMO it also plays suboptimal lines in which SF believes in a white advantage, which simply isn't there. The Sicilian Sveshnikov given elsewhere in this thread is just one example (I have seen more).

Hence, it is necessary to have a strong player in the project, one that can overrule SF decisions, one that can choose the lines that have better chances to play for a win. Nevertheless I think it is a very interesting project. In checkers they use the same method, all good books are completely computer generated.

Kind regards,

Jeroen
Up Topic The Rybka Lounge / Computer Chess / How good is the Cerebellum book with Stockfish?
1 2 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill