Author: Scott Gasch
Date: 02:02:15 08/17/99
Go up one level in this thread
Man, I'm embarassed to even post this but it relates to a question I have so:
1) val= 0 delta=65536 score= 225 n= 814
0.02 sec :-- e4h4
2) val= 225 delta= 250 score= 193 n= 2787
0.08 sec :-- e4h4 a5c3
3) val= 193 delta= 125 score= 187 n= 26441
0.48 sec :-- h5g3 g6f6 e4d5
4) val= 187 delta= 62 score= 186 n= 57185
0.97 sec :-- h5g3 g6f6 e4d5 d7f5
5) val= 186 delta= 50 score= 155 n= 166644
2.56 sec :-- h5g3 g6f6 e4d3 d7g4 g3e4
6) val= 155 delta= 50 score= 151 n= 431692
6.37 sec :-- h5g3 g6f6 e4d3 d7c6 g3e4 c6e4
7) val= 151 delta= 50 score= 149 n=1329501
18.58 sec :-- h5g3 g6f6 e4d3 d7g4 g3e4 f6g6 e1e3
8) val= 149 delta= 50 score= 149 n=3746592
51.89 sec :-- h5g3 g6f6 e4d3 d7g4 g3e4 f6g6 e1e3 g4f3
9) val= 149 delta= 50 score= 146 n=13013019
176.69 sec :-- h5g3 g6f6 e4d3 d7g4 g3e4 f6g6 e1e3 h8g8 g2g3
As you can see my idiot program does not see this move even after 9 ply. But
that's not the question -- I attribute this lack of vision to the fact that I
have check extensions turned off because they are too slow.
My question is this... compare the numbers below for "screamer" -- at ply 5 he
has searched a total of about 50k nodes. My program is already up at 166k.
Likewise on down the tree...
I am doing an ab search with null move and futility pruning in the qeval
routine. Moves on the PV are ordered first in the generator... other moves get
bonuses for being captures (MVV-LVA), promotions, killers, or history bonus.
So either there is 1) another way to prune the tree that I am not doing or 2) a
less than optimal move ordering resulting in a vastly larger tree. I have heard
of razoring and read a journal article on it but it seems to make my program
play badly. Yet, I am at a loss for what else to do to improve the move
ordering (internal iterative deepening?) and cannot explain the width of the
tree at deeper plys. Here are the stats:
internal nodes..................................1598135
leaf nodes......................................546
qeval nodes.....................................38988701 (32268522 cuts)
depth nodes.....................................0
-------
total nodes.....................................42450029
time (sec)......................................600.33
rate (nodes/sec)................................70711
positions added to hash.........................1403917
hash lookup hits................................194218 (4914 exact)
hash lookup misses..............................3266934
hash efficiency percentage......................5.61%
null move successes.............................1862647
null move failures..............................1216553
null move success rate..........................60.49%
pawn hash hits..................................36814950
pawn hash misses................................2172693
pawn hash success rate..........................94.43%
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.