Article View: gnu.chess
Article #5269Re: Gnuchess PL62 plays inconsistent (bug)
From: Tom Vijlbrief
Date: Wed, 24 Nov 1993 11:54
Date: Wed, 24 Nov 1993 11:54
104 lines
4482 bytes
4482 bytes
In gnu.chess you write:
>Tom Vijlbrief (tom@izf.tno.nl) wrote:
>: I noticed when running some of the Bratco-Kopec tests that
>: GNUCHESS shows some strange behaviour.
>: The idea is to turn off the hashtable and let it play a move:
>----- much text deleted ------
>: =====================================
>: Note that at the first move computation ply 2 shows a tree count of 4431 .
>: After the undo the tree count at ply 2 is 4437 !!!
>: Does somebody has an explanation for this. I think that somewhere
>: a variable is uninitialised !!!
>: This is more important than it seems at first sight
>: because after the undo, gnuchess sometimes plays a much better move !!!
>The difference is due to the static cache which is not cleared following
>the undo. If you were to add a ZeroTTable () in SelectMove () before the
>iterative deepening begins, you will find that the inconsistency will
>disappear. (I've tested this of course).
Too bad, I already added this myself and it makes a difference but I
got still differing node counts ! 6437 and 6337. Just trie undo-ing
the second or third move and recompute. The first move is always ok now.
Anyway, this does not explain why there is a difference in the node count.
It should only make a difference in speed, not in the node count !
It means that pruning is different when scores are taken from the
evaluation cache instead of recomputed. I can also get the difference
when I disable the cache (don't def -DCACHE)....
Any other hints ????
Tom
Your move is? go
Stage: 3
Move# 3 Target= 0 Clock: 0
2. 334 0 6437 g4d1 c8c1 d1c1 d7f5
3& 305 0 10100 g4d1 d7f5 h2h4 c8c1 d1c1
3& 329 0 10766 f7g7 d7g7 c1c8 a8c8 g4c8 g7f8 c8f8 g8f8
g5h7 f8g7
3& 364 0 12736 c1c8 d7c8 g4c8 a8c8 f7f6
3. 364 0 15163 c1c8 d7c8 g4c8 a8c8 f7f6
4- 196 1 18006 c1c8 d7c8 f7h7 c8g4 f2f3 g4f3 g2f3
4+ 487 1 23752 f7g7 g8h8 g7h7 h8g8 h7g7
4& 418 2 36067 f7g7 d7g7 g4e6 g8h8 g5f7 g7f7 c1c8 a8c8
e6f7
4. 418 4 93025 f7g7 d7g7 g4e6 g8h8 g5f7 g7f7 c1c8 a8c8
e6f7
5& 419 5 98609 f7g7 d7g7 g4e6 g8h8 g5f7 g7f7 c1c8 a8c8
e6f7
5. 419 7 156764 f7g7 d7g7 g4e6 g8h8 g5f7 g7f7 c1c8 a8c8
e6f7
^C 6. 419 9 192159 f7g7 d7g7 g4e6 g8h8 g5f7 g7f7 c1c8 a8c8
e6f7
8. ... f7g7
Gen 671129 Node 192159 Tree 479 Eval 8479 Rate 21351 EC 13124/7846 RS hi 0 lo 1
Hin/Hout/Tcol/Coll/Fin/Fout = 0/0/0/0/0/0
My move is: f7g7
Your move is? undo
White 0:00 Black 0:00
R - R - - - K -
P P - Q - r - P
- - - - - P P -
- - - P - - n -
- - - - - - q -
- - - - - - - -
p p - - - p p p
- - r - - - k -
Your move is? go
Stage: 3
Move# 3 Target= 0 Clock: 0
2. 334 0 6337 g4d1 c8c1 d1c1 d7f5
3& 305 0 10174 g4d1 d7f5 h2h4 c8c1 d1c1
3& 329 0 10840 f7g7 d7g7 c1c8 a8c8 g4c8 g7f8 c8f8 g8f8
g5h7 f8g7
3& 364 0 12179 c1c8 d7c8 g4c8 a8c8 f7f6
3. 364 0 14928 c1c8 d7c8 g4c8 a8c8 f7f6
4- 196 1 18155 c1c8 d7c8 f7h7 c8g4 f2f3 g4f3 g2f3
4+ 487 1 23905 f7g7 g8h8 g7h7 h8g8 h7g7
4& 418 2 36220 f7g7 d7g7 g4e6 g8h8 g5f7 g7f7 c1c8 a8c8
e6f7
^C 4. 364 3 66230 f7g7 d7g7 g4e6 g8h8 g5f7 g7f7 c1c8 a8c8
e6f7
9. ... f7g7
Gen 208752 Node 66230 Tree 423 Eval 3845 Rate 22076 EC 14697/8879 RS hi 0 lo 1
Hin/Hout/Tcol/Coll/Fin/Fout = 0/0/0/0/0/0
My move is: f7g7
Technical Details
Message-ID:
<CGzvrE.5z2@izf.tno.nl>Path:
arc1-ca.pugleaf.net!gmd.de!xlink.net!math.fu-berlin.de!netmbx.de!Germany.EU.net!EU.net!sun4nl!izfcs!tomReferences:
<CGy6FF.JE0@izf.tno.nl> <2cud45$p46@nuscc.nus.sg>Show Full Headers
Complete Article Headers
X-Google-Language: ENGLISH,ASCII-7-bit X-Google-Thread: 10d71d,a71519a005c13ba7 X-Google-Attributes: gid10d71d,public X-Google-ArrivalTime: 1993-11-24 07:45:14 PST Newsgroups: gnu.chess Path: arc1-ca.pugleaf.net!gmd.de!xlink.net!math.fu-berlin.de!netmbx.de!Germany.EU.net!EU.net!sun4nl!izfcs!tom From: tom@izf.tno.nl (Tom Vijlbrief) Subject: Re: Gnuchess PL62 plays inconsistent (bug) Message-ID: <CGzvrE.5z2@izf.tno.nl> Organization: TNO Institute for Perception References: <CGy6FF.JE0@izf.tno.nl> <2cud45$p46@nuscc.nus.sg> Date: Wed, 24 Nov 1993 11:54:50 GMT Lines: 106