TRACE v1.29 and Eeyore v1.39 released (n/t)

[Leo Dijksman]

Geschrieben von:/Posted by: Leo Dijksman at 08 June 2004 07:22:00:




WBEC Ridderkerk homepage.

[Norm Pollock]

Geschrieben von:/Posted by: Norm Pollock at 08 June 2004 19:06:25:
Als Antwort auf:/In reply to: TRACE v1.29 and Eeyore v1.39 released (n/t) geschrieben von:/posted by: Leo Dijksman at 08 June 2004 07:22:00:

I found an egtb problem with the new eeyore 139. It is supposed to implement tablebases. The log file (below) says that it found the 5-man tablebases. However when faced with KR v KRB, eeyore did not invoke the tablebases (which were available in the 2nd folder listed).
log file:
---------------------------------------------------------
SearchRepetionHashTableMemory = 128Kb (slots = 16384)
SearchHashTableMemory = 120703Kb (slots = 4577810)
PawnStructureHashTableMemory = 10240Kb (slots = 1048576)
Total hash memory = 131071Kb
message:KPKBitbase create : elapsed time = 0.078
message:EGTBPath = f:\winboard\tb;f:\nalimov5m
message:EGTBCacheSize = 8388608 bytes
message:5 piece tablebase files found

[Roger Brown]

Geschrieben von:/Posted by: Roger Brown at 08 June 2004 22:31:59:
Als Antwort auf:/In reply to: Re: TRACE v1.29 and Eeyore v1.39 released (n/t) geschrieben von:/posted by: Norm Pollock at 08 June 2004 19:06:25:

I found an egtb problem with the new eeyore 139. It is supposed to implement tablebases. The log file (below) says that it found the 5-man tablebases. However when faced with KR v KRB, eeyore did not invoke the tablebases (which were available in the 2nd folder listed).

Hello Norm,
An experiment:
Try referring only to the 5 man egtb's. Set up a position which is a tablebase mate with 5 pieces and check if the engine returns a mate score instantly - or at least without much calculation.
The second part of the experiment could be to transfer all the lesser egtb's to the 5 man egtb folder and see if the engine responds correctly.
Finally, it may just be an egtb problem as mentioned in your post.


Later.

[Norm Pollock]

Geschrieben von:/Posted by: Norm Pollock at 09 June 2004 01:14:47:
Als Antwort auf:/In reply to: Re: TRACE v1.29 and Eeyore v1.39 released (n/t) geschrieben von:/posted by: Roger Brown at 08 June 2004 22:31:59:

I found an egtb problem with the new eeyore 139. It is supposed to implement tablebases. The log file (below) says that it found the 5-man tablebases. However when faced with KR v KRB, eeyore did not invoke the tablebases (which were available in the 2nd folder listed).

Hello Norm,
An experiment:
Try referring only to the 5 man egtb's. Set up a position which is a tablebase mate with 5 pieces and check if the engine returns a mate score instantly - or at least without much calculation.
The second part of the experiment could be to transfer all the lesser egtb's to the 5 man egtb folder and see if the engine responds correctly.
Finally, it may just be an egtb problem as mentioned in your post.


Later.

I already saw that it calculates with 5 pieces on the board instead of using the nalimovs.
And mess up my hard drive and many configs? No!

It probably is the fact that I use 2 folders for the nalimov files instead of one folder. One folder is for 3-4s, and the other for some 5s. I prefer 2 folders in order to stay organized and for cases where the engine may only be able to do the 3-4s.
What I will try later is just testing with the 3-4s, and remove the folder reference for the 5s from the config file.
s-cheers

[Anton Maydell]

Geschrieben von:/Posted by: Anton Maydell at 10. June 2004 01:24:
Als Antwort auf:/In reply to: Re: TRACE v1.29 and Eeyore v1.39 released (n/t) geschrieben von:/posted by: Norm Pollock at 08 June 2004 19:06:25:

For check your tablebases try famous Philidor position:
3k4/5R2/3K4/3B4/8/8/4r3/8 w - - 0 1

However when faced with KR v KRB, eeyore did not invoke the tablebases
(which were available in the 2nd folder listed).

I think that Eeyore tryed win in draw position.
With a poor computer or human opponent, it is a good idea.
Without EGTG it is very difficult play without terrible mistakes in some theoretic draw positions like KQPKQ.

Practical example in rook vs knight ending:
[Event "Test group-2250 (5 min / 40 moves)"]
[Site "SDCHESS"]
[Date "2004.06.07"]
[Round "6.1"]
[White "Eeyore 1.39"]
[Black "Bestia 0.8"]
[Result "1-0"]
[SetUp "1"]
[FEN "3R4/8/8/5k2/8/1K2n3/3r4/8 w - - 0 120"]
[PlyCount "85"]
[EventDate "2004.??.??"]
120. Rxd2 Ng4 121. Re2 Nf6 122. Kc4 Ng8 123. Kd5 Nh6 124. Rf2+ Kg6 125. Ke4 Ng8
126. Ke5 Ne7 127. Rd2 Ng8 128. Ke6 Nh6 129. Rg2+ Kh7 130. Rg5 Ng8 131. Rg1 Nh6
132. Rg2 Ng8 133. Kf7 Nh6+ 134. Kf6 Ng8+ 135. Kf7 Nh6+ 136. Kf6 Ng8+ 137. Ke6
Nh6 138. Rg1 Ng8 139. Rg4 Nh6 140. Rg1 Ng8 141. Rg3 Nh6 142. Kf6 Ng8+ 143. Ke6
Nh6 144. Kf6 Ng8+ 145. Kf7 Nh6+ 146. Ke6 Ng8 147. Rg4 Nh6 148. Rg3 Ng8 149. Rg1
Nh6 150. Rg2 Ng8 151. Kd5 Ne7+ 152. Ke5 Ng6+ 153. Kf6 Nf4 154. Ra2 Kg8 155. Ra5
Kh7 156. Ra4 Ne2 157. Rh4+ Kg8 158. Kg6 Kf8 159. Rc4 Ng3 160. Rc1 Ke7 161. Rg1
Kd6 162. Rxg3 {Bestia resigns} 1-0