Author: Larry Griffiths
Date: 17:53:11 07/24/00
Go up one level in this thread
On July 24, 2000 at 19:48:37, Dann Corbit wrote:
>
>Check something that happens only very rarely for every move verses return error
>status when rare thing happens. Suppose, naively, that the cost of both
>operations is the same, and that 1/100 moves is illegal.
>
>Cost of method 1:
>brute cost + 100 * test cost
>
>Cost of method 2:
>brute cost + 1/100 * test cost
>
>Which one looks cheaper?
With ALPHA/BETA Cutoffs occuring the next ply might never get to test to see if
it captured a king. If a cutoff occurred on the next ply and a king capture was
not found, then an incheck for the opponents king must still occur to insure
that the move was really valid. This is interesting as a king check would only
have to be done once at the next ply. I only have brute-force move generation
in place at this time, but it would be interesting to see what happens using
this method since brute-force will always find a move that captures a king.
I ran with and without my incheck code and the results seem to indicate that I
have a lot of room to play with. Here are those results by the way...
WITH INCHECK...
Elapsed Time= 121.05 seconds
Ply ----CAPTURES--- -----MOVES----- -----TOTAL----- ----Invalid----
1 3 47 50 1
2 152 1,973 2,125 297
3 6,768 89,294 96,062 12,750
4 314,667 3,885,858 4,200,525 573,784
5 14,297,742 177,121,633 191,419,375 25,863,251
_______________ _______________ _______________ _______________
14,619,332 181,098,805 195,718,137 26,450,083
Captures/Moves Per Second= 1,616,877
WITHOUT INCHECK...
Elapsed Time= 148.78 seconds
Ply ----CAPTURES--- -----MOVES----- -----TOTAL----- ----Invalid----
1 3 48 51 0
2 166 2,306 2,472 0
3 8,599 117,795 126,394 0
4 457,128 5,802,155 6,259,283 0
5 23,970,850 298,958,050 322,928,900 0
_______________ _______________ _______________ _______________
24,436,746 304,880,354 329,317,100 0
Captures/Moves Per Second= 2,213,435
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.