Author: Stuart Cracraft
Date: 21:37:38 09/06/04
Tord,
Where do you save the threatmove[ply]? What
is your margin?
The only saves I have of a move in search()
is when ply == 0 and I have a new best alpha,
etc. and only when time flag hasn't fallen.
But I don't get any great result when I save
threatmove[ply] there and think it must be
some place else.
My extension code comes right after a nullmove
that returns an exact score, exact mate if found
(modified by depth) and the rest of the search
returns real values, not bounds...
#ifdef BM
if (!extend && value <= -2000 && ply >= 2 &&
((threat[ply].from==threat[ply-2].from&&
threat[ply].to==threat[ply-2].to&&
threat[ply].pro==threat[ply-2].pro) ||
(threat[ply].to==threat[ply-2].to&&threat[ply].cap==threat[ply-2].cap))) {
extend=1;
depth++;
}
#endif
I think my value <= -2000 is wrong and I think my saving
of threat[ply] is wrong.
Is there a good position to test the BM extension against?
I solve WAC 141 in about a minute so if Bm sped that up
guaranteeded, that would be a good test. Currently the
above extension has no real effect on WAC 141 for me.
Obviously it is wrong.
wac 141 without "false bm"
Alpha=-1182 Beta=-682 Maxdepth=9999999 MaxTime=999999999
1/13 g2f1 0.01 -953 945 g2f1 f4d5
2/14 g2f1 0.01 -953 1644 g2f1 f4d5 c1g5
3/14 g2f1 0.02 -953 5098 g2f1 f4d5 c1g5 d5f6
4/23 g2f1 0.09 -953 21456 g2f1 f4d5 b3d5 c6d5 c1c7 d6c7 f1g1
5/25 g2f1 0.67 -953 178724 g2f1 b5b4 b3a4 f4d5 f6g5 d5e7
6/44 g2f1 2.82 -953 694967 g2f1 b5b4 b3a4 f4d5 f6g5 d5e7 g5e7
7/44> g2f1 62.56 -703 14609806 g2f1 e8c8 h4h5 g6h5 c1f4 d6f4 h1h5 f4h6 h5h6
7/47 c1f4 98.21 5113 23871198 c1f4 e8e6 f4g5 d7e7 b3e6 e7e6 h1d1 d6e7
8/47 c1f4 108.15 5113 25960069 c1f4 e8e6 f4g5 d7e7 b3e6 e7e6 h1d1 d6e7
wac 141 with "false bm"
2/14 g2f1 0.01 -953 1644 g2f1 f4d5 c1g5
3/14 g2f1 0.02 -953 5098 g2f1 f4d5 c1g5 d5f6
4/23 g2f1 0.09 -953 21456 g2f1 f4d5 b3d5 c6d5 c1c7 d6c7 f1g1
5/25 g2f1 0.85 -953 178724 g2f1 b5b4 b3a4 f4d5 f6g5 d5e7
6/44 g2f1 3.00 -953 694967 g2f1 b5b4 b3a4 f4d5 f6g5 d5e7 g5e7
7/44> g2f1 60.07 -703 14609806 g2f1 e8c8 h4h5 g6h5 c1f4 d6f4 h1h5 f4h6 h5h6
7/47 c1f4 95.92 5113 23871198 c1f4 e8e6 f4g5 d7e7 b3e6 e7e6 h1d1 d6e7
8/47 c1f4 106.32 5113 25960069 c1f4 e8e6 f4g5 d7e7 b3e6 e7e6 h1d1 d6e7
A 3% improvement, so something is happening, but obviously not the real "it".
Stuart
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.