Author: Dann Corbit
Date: 18:24:32 08/29/02
Go up one level in this thread
On August 29, 2002 at 20:40:44, David Hanley wrote:
>Ok, i have my program working and fairly complete--it has all the rules in and
>plays decent positional chess. I'm pleased.
>
>I implemented all the mechanisms that i think i'm supposed to--nullmove, hash
>move first, then two killers, then winning static exchange moves sorted by
>expected material win, then moves sorted by their history score. I sort moves
>at the root with each iterative deepening.
>
>However, the branching factor seems 'off.' My branching factor seems to vary
>between 4 and 7. For example, here's an 8 ply search from the opening position:
>
>Score: -53
>Searched 414821 total nodes
>Got 114215 hash hits (27 percent hits)
>There were 240894 quiesce nodes
>The pawn hash contains 17855 entries
>Branching factor is 4.522705
>
>Here's a 6 ply search a bit later: (white just thought and chose e3)
>
>+---+---+---+---+---+---+---+---+
>| R | N | | Q | K | B | | R |
>+---+---+---+---+---+---+---+---+
>| P | P | P | | P | P | P | P |
>+---+---+---+---+---+---+---+---+
>| | | | | | N | | |
>+---+---+---+---+---+---+---+---+
>| | | | P | | B | | |
>+---+---+---+---+---+---+---+---+
>| | | |*p | | | | |
>+---+---+---+---+---+---+---+---+
>| | |*n | |*p |*n | | |
>+---+---+---+---+---+---+---+---+
>|*p |*p |*p | | |*p |*p |*p |
>+---+---+---+---+---+---+---+---+
>|*r | |*b |*q |*k |*b | |*r |
>+---+---+---+---+---+---+---+---+
>Score: -44
>Searched 134392 nodes
>Got 27504 hash hits (20 percent hits)
>There were 87728 quiesce nodes
>The pawn hash contains 6098 entries
>Branching factor is 5.999756
>
>
>I guess i wonder if i'm fussing over nothing. Or perhaps if my search is broken
>somehow. I tried turning off move sorting and launched a search from the
>opening position. It seems to indicate to me that the move sorting is having a
>big positive effect. Here's a six ply search from the opening position, no move
>sorting:
>
>Score: -55
>Searched 591677 nodes
>Got 128337 hash hits (21 percent hits)
>There were 395070 quiesce nodes
>The pawn hash contains 31050 entries
>Branching factor is 7.623291
>
>It took more nodes than the 8 ply search with sorting!
>
>Maybe i should be happy with my code, but i hear people talking of branching
>factors like two and three. So maybe i should look for a bug.
================================================
Yace looked at less than ten percent of that many nodes:
================================================
white ( 1): sd 6
search depth=6
white ( 1): go
usetime = 9999999999.00, mintime = 9999999999.00 maxtime = 9999999999.00 tl
9999999999.00 ml 0
1 0.025 0.27 1t 1.e4 {0}
20 0.030 0.27 1. 1.e4 {0}
42 0.037 0.00 2t 1.e4 e5 {0}
117 0.044 0.00 2. 1.e4 e5 {0}
296 0.050 0.19 3t 1.e4 d5 2.Nc3 {0}
382 0.057 0.19 3. 1.e4 d5 2.Nc3 {0}
917 0.065 0.00 4t 1.e4 d5 2.Nc3 Nf6 {0}
2708 0.075 0.00 4. 1.e4 d5 2.Nc3 Nf6 {0}
5475 0.088 0.18 5t 1.e4 d5 2.exd5 Qxd5 3.Nc3 Qd4 {0}
6798 0.098 0.19 5t+ 1.d4 d5 2.Nc3 Nf6 3.Nf3 {0}
6938 0.103 0.19 5t 1.d4 d5 2.Nc3 Nf6H 3.Nf3H {0}
9347 0.121 0.19 5. 1.d4 d5 2.Nc3 Nf6 3.Nf3 {0}
13968 0.138 0.00 6t 1.d4 d5 2.Nc3 Nf6 3.Nf3 Nc6 {0}
34240 0.210 0.00 6. 1.d4 d5 2.Nc3 Nf6 3.Nf3 Nc6 {0}
34240 0.214 0.00 6f. 1.d4 d5 2.Nc3 Nf6 3.Nf3 Nc6 {0}
34240 Nodes, 24.82% Leavenodes, 160000 Nodes/sec
27160 eval, 58.52% score, 98849 genmoves, 14.28% captures le 95/84
ext: pawn 0, rcp 26, chk 226, repchk 60, null 0, prune 599
htable: 28874 store, 0 rejected, 34372 probe, 28.1% f/p, 33.4% f/s
entries 5416665 age 1 renew 0
egtb probes 0, found 0 max_depth 14
white ( 1): d4
================================================
GLC-218 shows an even more drastic difference:
================================================
>sd 6
>go
fen rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1
Game finished test: 0
time limit 10.00 (1 moves left) max time limit 10.00
Generating internal KP-K end-game table...
Done (0.320 secs)
Game stage: Opening
Current eval: 0.000
Ply Time Score Nodes Principal Variation
6 0.02 +0.000 6174 Nf3 Nf6 2. Nc3 Nc6 3. d3 e6
6 0.04 +0.000 12970 Nf3 Nf6 2. Nc3 Nc6 3. d3 e6
local: t=0.05 nps=259404.0 n=12970 (37.2% / 62.8%) fh=91.7%
total: t=0.05 nps=259404.0 n=12970 draws=0
trans: probes=4826 hits=858 (17.78%) draft=342 (7.09%)
tcuts: exact=7 (0.15%) upper=309 (6.40%) lower=9 (0.19%)
tstor: exact=18 (1.52%) upper=920 (77.83%) lower=244 (20.64%)
ext: check=25 recap=13 ppush=0 1rep=3 thrt=0
q-moves: gen=2712 tested=2040 made/un=941 max-dep=5
max eval diff: part-1=0.721 part-2=0.809
!m Ng1-f3 0 1
move g1f3
time left 10.00
-> Score=0 PV=Nf3 Nf6 2. Nc3 Nc6 3. d3 e6
Game finished test: 0
>
================================================
Gnu Chess:
================================================
GNU Chess 5.04aTE
Transposition table: Entries=1024K Size=32768K
Pawn hash table: Entries=16K Size=1024K
White (1) : post
White (1) : book off
book now off.
White (1) : sd 6
Search to a depth of 6
White (1) : go
Thinking...
Root = -35, Phase = 1 Depth = 6
Ply Time Eval Nodes Principal-Variation
1+ 0.01 41 1 Nf3
1. 0.02 41 21 Nf3
2& 0.02 -10 61 Nf3 Nf6
2. 0.04 -10 99 Nf3 Nf6
3& 0.04 40 203 Nf3 Nf6 Nc3
3. 0.06 40 222 Nf3 Nf6 Nc3
4& 0.07 -10 376 Nf3 Nf6 Nc3 Nc6
4. 0.08 -10 869 Nf3 Nf6 Nc3 Nc6
5& 0.10 29 2001 Nf3 Nf6 Nc3 Nc6 e4
5. 0.12 29 3056 Nf3 Nf6 Nc3 Nc6 e4
6& 0.14 -10 5771 Nf3 Nf6 Nc3 Nc6 e4 e5
6. 0.21 -10 17175 Nf3 Nf6 Nc3 Nc6 e4 e5
Time = 0.2 Rate=83407 Nodes=[15329/1846/17175] GenCnt=19137
Eval=[4444/7463] RptCnt=0 NullCut=746 FutlCut=2442
Ext: Chk=203 Recap=25 Pawn=167 OneRep=19 Horz=1 Mate=0 KThrt=19
Material=[3600/3600 : 4400/4400] Lazy=[177/129] MaxPosnScore=[226/204]
Hash: Success=25% Collision=6% Pawn=73%
black KQkq
r n b q k b n r
p p p p p p p p
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . N . .
P P P P P P P P
R N B Q K B . R
My move is : Nf3
================================================
Phalanx needs less than 10K nodes:
================================================
Depth=6, Value=31, Time=0.18, Last turn=0.00, Nodes=9829, N/s=54605
PV = Pd2-d4 Pd7-d5 Ng1-f3 Nb8-c6 Pe2-e3 Ng8-f6
my move is Pd2-d4
1. ... d2d4
[ black, 1 ]
================================================
Genesis:
================================================
White(1): bookoff
opening book off.
White(1): on
depth time score variation
1-> 0.01 0.22 1. d2d4
nodes 21 nps 2100 q 0% hhits 0%
2-> 0.02 0.00 1. d2d4 d7d5
nodes 73 nps 7300 q 2% hhits 1%
3-> 0.04 0.18 1. d2d4 d7d5 2. g1f3
nodes 204 nps 10200 q 3% hhits 19%
4-> 0.08 0.03 1. e2e4 g8f6 2. e4e5 f6e4
nodes 1812 nps 45300 q 18% hhits 16%
5-> 0.15 0.17 1. e2e4 e7e5 2. d2d4 d7d5 3. g1f3 d5e4 4. f3e5
nodes 3941 nps 56300 q 24% hhits 7%
6-> 0.60 0.01 1. d2d4 d7d5 2. c1e3 g8f6 3. b1c3
nodes 34531 nps 76565 q 27% hhits 13%
----------------------------------------------------------
nodes 40582 time 0.60 nps 67524
leaf 24179 q 10898 eval 29531
null 4205 null-cut 2053
hhits 5205 hexact 2751 hleaves 2751
hput 7847 hoverwrites 0
leaves 59% quiesc 26% null 10%
null-cut 65% aspwin-suc 100%
hhits 12% hoverwrites 0% husage 0%
White(1): d2d4
This page took 0.01 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.