Author: Daniel Clausen
Date: 12:53:29 01/04/00
Go up one level in this thread
Hi
On January 04, 2000 at 12:25:25, Gian-Carlo Pascutto wrote:
>On January 03, 2000 at 18:42:31, Daniel Clausen wrote:
>
>>On January 03, 2000 at 18:24:21, Gian-Carlo Pascutto wrote:
>>
>>>On January 03, 2000 at 16:52:21, Daniel Clausen wrote:
>>>
>>>[snip]
>>>> #q=19291 [1%] // #(quiescense-nodes#
>>>
>>>Wow...does *my* chessprogram have a problem or is this very few ?
>>
>>Nope, it was a copy/paste error on my side. Actually the 19291 nodes are 11%
>>of all nodes, not 1%.
>
>Hrm...I'm usually seeing +-50% quiescent nodes...I always thought this was
>a more or less normal value.
I can't test against other engines, since the Crafty version I use doesn't show
the
number of q-nodes and Hiarcs also doesn't. :(
Note that the numbers above were for the initial position though. And my
q-search only includes captures. (no checks, or other extensions in qsearch)
For the position (7 plies)
2rr2k1/1b3ppp/pb2p3/1p2P3/1P2BPnq/P1N3P1/1B2Q2P/R4R1K/ b - -
I get:
#nodes=2'809'852
#q=545'239 [19%]
(Of course without finding the winning move Rxc3)
Hm.. it's more likely that I count q-nodes also different than other people
though. :))
int search(int ply, int maxPly, ...)
{
[snip]
nodes++;
[snip]
if(ply==maxPly)
return qsearch(ply, maxPly, ...);
[snip]
}
int qsearch(int ply, int maxPly, ...)
{
[snip]
if(ply>maxPly) nodes++, qNodes++;
[snip]
qsearch(ply+1, maxPly, ...);
[snip]
}
>Try searching with a window of (-INF, -INF+1). You should get a quite high
>percentage now. I don't think there's any sense in trying to improve it ;)
Lol. Not what I had in mind though. I meant that if the number of fail-highs
increases (not only after 1st move), this should lead to a smaller tree. (since
all moves after the fail-high can be skipped) Of course it's still better to
fail-
high after the 1st move, but I'd rather fail-high after the 5th than not at all.
Kind regards,
-sargon
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.