Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Question re: 6-piece Nalimov tablebases
- - By herakleitos (***) Date 2007-08-19 22:56
One of the many nice things about Rybka is its support of the 6-piece Nalimov tablebases. Since I use engines entirely for annotation and analysis, I wouldn't think of spending money on an engine that didn't consult the 6-piece bases.

Of the engines out there, it seems only Rybka, Fritz, and Hiarcs use the 6-piece bases. Why is this? I've read that it's not difficult for a programmer to get the code needed to implement the bases. Why haven't the other programmers taken advantage of this?

Vas, if you read this, do you have any thoughts on this?

Bob
Parent - By Arrière Pensée (Gold) Date 2007-08-19 23:11
Hi herakleitos,

Well, that is refreshing to hear!
Parent - By turbojuice1122 (Gold) [us] Date 2007-08-19 23:27
Most of the good engines support them--Shredder, Junior, Spike, some of the Fruits, etc.
Parent - By Banned for Life (Gold) Date 2007-08-20 00:39
Using Nalimov's tablebases requires using his code too (as Vas once mentioned, either 1 or 0 people actually understand this code :-) ). This code is available for inspection as part of the Crafty source. People who want to use it should get Nalimov's permission first.

Alan
Parent - - By Vasik Rajlich (Silver) [hu] Date 2007-08-20 20:29
Nalimov's code automatically supports 6-man TBs. Some people have a problem with Nalimov's license and refuse to use it.

Vas
Parent - By herakleitos (***) Date 2007-08-20 22:16
This is the answer I was looking for. Thanks.
Parent - By MoKy (**) [eu] Date 2007-08-22 15:03
I thought how to reduce amount of EGTB without loss of informational value. Here are two ways:
1. Simple pruning  method stores positions with DTM=ar only (a¢N, r is radius)
2. This method selects important positions with value>min. First in position(i) retract move(j). If move(j) is the best, then position(i).value += abs(shortest_DTM-next_shortest_DTM)
I believe we can compress 4-5-6 pieces on one CD.
Mario
Parent - - By Uri Blass (*****) [il] Date 2007-08-22 17:06
I also use engines for analysis but I do not see the tablebases as extremely important.

for knowing if 6 piece tablebases position is drawn or won I can look at the internet and I do not need tablebases and if the root position is not tablebases positions there are other things except tablebases that are important and it is not clear that engine that support tablebases is better.

Uri
Parent - - By Yoav Dothan (**) [il] Date 2007-08-23 13:59
I do not think that you are correct. Just imagine a 7 piece ending , where the program reaches a 6-piece and knows at once what is the result of that ending, it will produce a better analysis of the 7-pieces.
Parent - - By Uri Blass (*****) [il] Date 2007-08-23 14:08
This is right but on the other hand the program may search less deeply because of probing tablebases and miss a win because of not seeing deep enough a line when no pieces are traded.

Uri
Parent - - By M ANSARI (*****) [kw] Date 2007-08-23 19:27
The best thing is to have a passive extra machine with full EGTB's that is observing the main engine and will only get involved if it sees a big evaluation jump.  This way the main engine will work without EGTB handicap yet still play perfect 6 EGTB endings ... the best of both worlds.
Parent - By Uri Blass (*****) [il] Date 2007-08-23 20:33
I was talking about analysis and not about playing.

In analysis if I get a 6 piece tablebases position I can simply look at the internet for the right result and right move.

Uri
Up Topic Rybka Support & Discussion / Rybka Discussion / Question re: 6-piece Nalimov tablebases

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill