tablebases

emschorsch

I've become interested recently in the progress being made on tablebases. I know 7 piece tablebases are largely completed but they are simply too large to store on an average computer. Does anybody have information on when 8 piece tablebases will be produced or who is working on it.

Liburkin

Emanuel,

As you say, tablebases have grown in size to the point where they are no longer easy to distribute, either on DVD or any other media. There is a chess site that allows free access to their six-man tablebases. It's here: http://www.k4it.de/index.php?topic=egtb&lang=en

There is also a link to a program there called Freezer. I have it and have found it useful for analyzing my endgames. The author has a doctoral thesis in which he suggests that there is probably a better way than to develop tablebases for many pieces. His paper can be found here: http://www.minet.uni-jena.de/preprints/bleicher_04/FREEZER_.PDF

Scarblac

Have the 7-piece ones been completed? I thought only some of them had been done.

Wikipedia says the 5-piece ones take ~ 7GB, 6-piece ~ 1.2TB, so presumably the 7-pieces tablebases take ~~ 200 TB in total. Have they really been calculated?

AMcHarg

200 TB storage on an average comp is just around the corner anyway so that won't be an 'issue' for long.

emschorsch

200TB is not around the corner. I know that it isn't outrageous to own a 2-4 TB harddrive, but 200 is definetly far away, at least for the average computer.

goldendog

"Another drawback is that tablebases require a lot of memory to store the many thousands of positions. The Nalimov tablebases, which use state-of-the-art compression techniques, require 7.05 GB of hard disk space for all five-piece endings. The six-piece endings require approximately 1.2 terabytes.[28][29] Nalimov seven-piece tablebases require more hard drive storage capacity and RAM to operate than will be practical to use for the foreseeable future. Bitbases, however, take much less space. Shredderbases, for example, used by the Shredder program, compress all three, four and five piece bases into 157 MB. This is a mere fraction of the 7.05 GB that the Nalimov tablebases require.[30]"

So says wikipedia. Any cognoscenti know about Shredderbases and if we can expect such compression for the six- and seven-man bases anytime soon?

kingwangthegreat

Shredderbases have been completed for the 6 man tables already. http://www.shredderchess.com/online-chess/online-databases/endgame-database.html

 

It will give values for almost all 6 piece endgames. The only exception I think are 5 vs 1 endgames, which are trivial.

TheGrobe
Fezzik wrote:

...

Since it is now "common knowledge" and universally available for free, is there any chance chess.com will allow the use of this dB?

...


Engines are also universally available for free.  Like engines, though, no way should these be allowed during play here.

TheGrobe

No, I didn't misinterpret, I just disagree with the idea that they should be permissible.  Being the stored results of exhaustive computational analysis, they are effectively equivalent to engines in terms of the type of assistance they represent -- in fact, they are worse as they are infallible (or, as you put it, equally strong for everyone).  Why not just have games automatically adjudicated based on a table base as soon as they reach six remaining pieces?

RainbowRising

No Fezzik, cheating is not allowed here, whatever you may think.

TheGrobe

I'd prefer to actually play out my endgames.  I'm here to learn.

Now, I can see this as a user optional setting, and provided both users have turned on their preference for tablebase adjudication then by all means apply it, but I personally would not turn it on.

Loomis
TheGrobe wrote:

Being the stored results of exhaustive computational analysis, they are effectively equivalent to engines in terms of the type of assistance they represent


Much of what is available in books is also the stored result of computational analysis and effectively equivalent to engines, but is not against the rules. I am permitted to use an opening book, which the author may have relied heavily on computer analysis to produce. But I'm not permitted to use the endgame tablebases.

 

I don't have any problem with the rules as they are, I just don't see that they are really as logical as you put it.

Loomis
TheGrobe wrote:

I'd prefer to actually play out my endgames.  I'm here to learn.


Some people are here to learn openings, but the use of books is allowed.

TheGrobe

Well, opening databases diverge whereas endgame tablebases converge.  Moreover, they are exhaustive and, if both players are using them, they might as well be used to adjudicate the game as soon as they can be employed.  Opening databases do not have this problem --  you make choices about what lines you enter as you employ them, and at a certain point in the game you must leave the book and fend for yourself.  At no point, once and endgame tablebase is employed, are you subsequently requried to actually think for yourself.  You are always told what the optimal move is.

Loomis

Ok, but this is a completely new argument, different from the "tablebases are equivalent to engine use" argument from before. :-)

The use of books is allowed including any printed material. Correct me if I'm wrong, but I think endgame manuals are allowed. For example, if I find myself with KRP v. KR, I could check my endgame book and see if the position is there. If it is, I likely have step by step instructions on how to win (or draw). So, it seems that in some cases it is permitted to have outside help that converges to a win and no longer requires the player to think for himself.

SirakD

Looking forward to the 40 pieces db!

TheGrobe

I don't see it as being a different argument.  An endgame manual will help you find the right idea, not the optimal move.  A tablebase doesn't tell you why the move presented is the best one (mind you, neither does an opening database), it just presents you with the definitive answer (which an opening database does not -- it presents you with options).  As such, blindly following one is very much akin to engaging an engine.

Loomis

I meant that your argument about converging vs. diverging is completely different from your argument about it being equivalent to engine use.

 

In many cases an endgame manual will help you find the optimal move, no thinking for yourself required.

SirakD
Fezzik wrote:
SirakD wrote:

Looking forward to the 40 pieces db!


 What game has 40 pieces?


Oooops, i wanted to say 32. I was confused, cause of you have 20 possibel moves in the start Position :-)

emschorsch

so checking back in a year and a half later, anyone know of any progress towards compression of 7 piece tablebases into a usable form?