Author: Robert Hyatt
Date: 18:27:14 06/09/04
Go up one level in this thread
On June 09, 2004 at 18:01:26, Dan Honeycutt wrote:
>Others might enjoy this example of stupidity. I'm in QSearch looking for
>captures to search that will get me to alpha. pm1 is the start of my moves and
>pm2 is the next start position (one beyond the end of my moves). My code:
>
>need_to_win = 0;
>if (stand_pat < alpha) need_to_win = alpha - stand_pat;
>while (pm1 < pm2) {
> if (value(pm1->victim) < need_to_win) {
> //no hope for this move
> pm2--;
> *pm1 = *pm2;
> continue;
> }
> if (value(pm1->victim) - value(pm1->attackor) >= need_to_win) {
> //sure winner
> pm1++;
> continue;
> }
> //no answer from MVV/LVA - try SEE
> if (SEE(pm1) >= need_to_win) pm1++;
> else pm2--;
>}
>
>I go to the trouble to do the static exchange evaluation - if the move is no
>good I keep it and throw away all the rest!
>
>Dan H.
That is not unexpected. I entered a human event with a similar type of bug. I
actually reversed min and max so that in one odd case the program chose the
worst possible move. Turned out to be a capture that the human thought about
for 45 minutes and then declined to take back. He lost. He asked after the
game, "if I had taken would I have gotten mated?" I answered, "no, you would
have won, it was a bug." :)
Another game against GM Stuart Rachaels in Mobile Alabama with my brand new 1984
parallel search based on Monty Newborn's "PVS idea". 4 cpus. Program played
Bxp+. And announced a mate. Stuart captured the bishop "out of mate". Bug was
that after the check, black had 3 legal moves. I had 4 cpus. Three searched,
one said "I am mated" and a bug caused that result to be remembered. :)
It happens...
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.