Author: Anthony Cozzie
Date: 05:25:03 08/12/03
Go up one level in this thread
On August 12, 2003 at 06:23:26, macaroni wrote:
>Hi,
>in an alpha-beta search, with hash tables, null moving, history heuristic,
>killers, pvs search, quiescent search, and itterative deepening.
>about how many nodes would one expect after a 6 ply search from the opening
>position, I get 13,000, is this way to many? or about right.
It's reasonable, although it depends a lot on what position you are analyzing,
the extensions you use, and a whole host of other things.
>Also, about how much extra performance would one expect from killers? I seem to
>get a really marginal increase, what replacement scheme is good for killers?
>thanks.
I seem to remember Zappa gets 30% or so from killers. I use two killers;
this is the update code. Zappa also uses Ed's mate killers, which is the
k->mt_1 stuff. I have no idea if this is optimal; YMMV.
//update killers
if(k->m1.i == best.i) {
k->m1_count += (k->m1_count < 10); //don't let the max killer value get
too big
k->mt_1 |= (score > 9000);
}
else if(k->m2.i == best.i) {
k->m2_count++;
k->mt_2 |= (score > 9000);
if(k->m2_count > k->m1_count) { //k2 is new dominant killer, swap!
int_swap(&k->m1_count, &k->m2_count);
int_swap(&k->m1.i, &k->m2.i);
int_swap(&k->mt_1, &k->mt_2);
}
}
else {
k->m2.i = best.i;
k->m2_count = 0;
k->mt_2 = (score > 9000);
}
anthony
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.