Author: Tim Foden
Date: 10:07:55 02/04/04
Go up one level in this thread
On February 03, 2004 at 04:06:22, Stefan Knappe wrote:
>Hi guys,
>
>at first I'd like to thank Volker and Schroer for the organization of the
>tournament and the comments on the games!
>
>I am suprised about the fact, that yace and crafty don't like Matadors sacrifice
>against Alarm.
>
>What do you and your engines think about Rxe6? Matador prefers it after 2
>seconds with a score of +1. However it was interesting and successful :-)
>
>[D] 4k3/1pq2p1r/rnpbp3/p4pB1/P2P3P/1QN2PPK/1P2R3/7R w - - 0 26
>
>Have a nice day!
>
>Regards,
>Stefan
Green Light 3.00 doesn't choose Rxe6. It prefers Rc1. It still thinks that
white is winning, although the score is dropping all the time. Here is the
analysis output...
GLC 3.00, AXP 2.1GHz, 96MB Hash:
Game stage: Middle game
Current eval: 1.422
Ply Time Score Nodes Principal variation
4 0.010 +0.986 5188 Rg1 Rh5 2. Qc2 Kd7
4 0.010 +0.986 12823 Rg1 Rh5 2. Qc2 Kd7
5 0.020 +0.946 22681 Rg1 Kd7 2. Bf6 Kc8 3. Rc2
5 0.050 +0.946 55016 Rg1 Kd7 2. Bf6 Kc8 3. Rc2
6 0.070 +1.027 76390 Rg1 Kd7 2. Rc2 Kc8 3. Nb5 Qd7
6 0.160 +1.027 170535 Rg1 Kd7 2. Rc2 Kc8 3. Nb5 Qd7
7 0.220 +0.861 231263 Rg1 Kd7 2. Rc2 Rh5 3. Bf6 Nd5 4. Nb5
7 0.520 +1.032 529057 Rhe1 Kf8 2. Rg2 Bb4 3. Re3 Nd5 4. Nxd5 cxd5
7 0.691 +1.032 726114 Rhe1 Kf8 2. Rg2 Bb4 3. Re3 Nd5 4. Nxd5 cxd5
8 1.382 +0.864 1422620 Rhe1 Kf8 2. Rg2 Be7 3. Bxe7+ Qxe7 4. Qd1 Rh5 5. f4
{ht}
8 2.513 +0.864 2604142 Rhe1 Kf8 2. Rg2 Be7 3. Bxe7+ Qxe7 4. Qd1 Rh5 5. f4
{ht}
9 3.835 +0.851 3928340 Rhe1 Kf8 2. Rg2 Be7 3. Bxe7+ Qxe7 4. Qd1 Kg7 5.
Qd3 Nd5 6. Nxd5 cxd5
9 6.169 +1.085 6374323 Rc1 Bxg3 2. Rxe6+ fxe6 3. Qxe6+ Kf8 4. Bh6+ Rxh6
5. Qxh6+ Ke8 6. Qg6+ Kd8 7. Qxg3 Qf7
9 6.589 +1.085 6882781 Rc1 Bxg3 2. Rxe6+ fxe6 3. Qxe6+ Kf8 4. Bh6+ Rxh6
5. Qxh6+ Ke8 6. Qg6+ Kd8 7. Qxg3 Qf7
10 9.513 +1.170 9988940 Rc1 Bxg3 2. Rxe6+ fxe6 3. Qxe6+ Kf8 4. Bh6+ Rxh6
5. Qxh6+ Qg7 6. Qxg7+ Kxg7 7. Kxg3 Nc4 8. Kf4
Nd2
10 10.945 +1.170 11569k Rc1 Bxg3 2. Rxe6+ fxe6 3. Qxe6+ Kf8 4. Bh6+ Rxh6
5. Qxh6+ Qg7 6. Qxg7+ Kxg7 7. Kxg3 Nc4 8. Kf4
Nd2
11 20.459 +0.974 21987k Rc1 Kf8 2. Nb5 Qd7 3. Nxd6 Qxd6 4. Qe3 Kg8 5. Bf4
Qb4 6. b3 Rh5
11 30.023 +0.974 32857k Rc1 Kf8 2. Nb5 Qd7 3. Nxd6 Qxd6 4. Qe3 Kg8 5. Bf4
Qb4 6. b3 Rh5
12 45.365 +1.139 49072k Rc1 Kf8 2. Nb5 Qd7 3. Nxd6 Qxd6 4. Bf6 Rh5 5. Rg1
Kg8 6. g4 fxg4+ 7. Rxg4+ Kf8 8. Qc2
12 1:12.01 +1.139 79172k Rc1 Kf8 2. Nb5 Qd7 3. Nxd6 Qxd6 4. Bf6 Rh5 5. Rg1
Kg8 6. g4 fxg4+ 7. Rxg4+ Kf8 8. Qc2
13 2:02.87 +0.882 133231k Rc1 Be7 2. Bxe7 Qxe7 3. d5 f4 4. dxe6 f5 5. Qd1
fxg3 6. Qd4 Qb4 7. Qf6 f4
13 5:50.45 +0.882 374352k Rc1 Be7 2. Bxe7 Qxe7 3. d5 f4 4. dxe6 f5 5. Qd1
fxg3 6. Qd4 Qb4 7. Qf6 f4
14 10:40.4 +0.563 647030k Rc1 Be7 2. f4 Kf8 3. Kg2 Qd7 4. Bxe7+ Qxe7 5. d5
Qb4 6. Qxb4+ axb4 7. dxc6 bxc6 8. Na2 Nd5 9. b3
Rh5 10. Rce1
Analyse>exit
local: t=16:51.6 nps=986763.1 n=998193772 (19.6% / 80.4%)
total: t=16:51.6 nps=986763.1 n=998193772
stats: fh=92.2%/3.14%/0.961% draws=128004
trans: probes=195499004 hits=18991007 (9.71%) draft=14448445 (7.39%)
tcuts: exact=1098 (0.00%) upper=7672801 (3.92%) lower=4146994 (2.12%)
tstor: exact=1572 (0.00%) upper=64483931 (69.71%) lower=28011945 (30.28%)
ext: check=43292720 recap=3405771 ppush=363755 1rep=2574233 thrt=0
q-moves: gen=704535753 tested=325668294 made/un=92333099 max-dep=14
max eval diff: part-1=3.661 part-2=3.486
Maybe the reason it doesn't choose Rxe6 is that it likes the root position so
much already? You can the the static evaluation value of the root position at
the top... +1.422 (for white)
Some other engines I have here that will give a score for the current position
give:
Crafty 17.11: 0.03
Yace 0.99.50: 0.09
Gothmog 0.4.5: -0.75 (well it actually says -75, but I guess centipawns?)
So maybe I have to look a bit more into how GL gets its huge 1.422 for the
current position. Maybe others could have a look at this and see if they think
it's complete garbage? :)
You can get a complete dump of the evaluation score for a position in GL by
setting a position (with the setboard or fen commands) and typing "sc". Here is
GL's take on the root position in question:
>sc
Game stage: Middle game
w pawn (f3) singly connected 20
w pawn (g3) doubly connected 35
w pawn (a4) singly connected 20
w pawn (a4) weak -97
w pawn (d4) isolated -104
w pawn (h4) passed 202
w pawn (h4) connected passed 51
w pawn (h4) singly connected 20
b pawn (a5) singly connected 20
b pawn (a5) weak -97
b pawn (f5) doubled -30
b pawn (f5) singly connected 20
b pawn (c6) singly connected 20
b pawn (e6) singly connected 20
b pawn (b7) singly connected 20
b pawn (f7) doubled -30
b pawn (f7) singly connected 20
w pawn (h4) passed outside right 100
w pawn (h4) bishop promote sq 59
eval: white safety -0.328 (wdef 12, bmat 28013)
eval: black safety -0.711 (bdef 26, wmat 28013)
*** king safety: white: -0.328 black: -0.711
development -80 (20 0)
w pawn (b2) static eval 5
w pawn (b2) blocked -80
w pawn (f3) static eval 20
w pawn (g3) static eval 15
w pawn (a4) static eval 25
w pawn (a4) blocked -60
w pawn (d4) static eval 50
w pawn (h4) static eval 25
b pawn (f7) static eval 10
b pawn (b7) static eval 5
b pawn (b7) blocked -80
b pawn (e6) static eval 25
b pawn (c6) static eval 20
b pawn (f5) static eval 35
b pawn (a5) static eval 25
b pawn (a5) blocked -60
w knight (c3) static eval 20
b knight (b6) static eval 19
b knight (b6) mobility bonus 40
w bishop (g5) static eval 21
w bishop (g5) mobility bonus 60
w bishop (g5) attack bonus 3
b bishop (d6) static eval 20
b bishop (d6) mobility bonus 60
b bishop (d6) attack bonus 3
w rook (h1) static eval -1
w rook (h1) mobility bonus 60
w rook (h1) rank mobility 39
w rook (e2) static eval 40
w rook (e2) mobility bonus 65
w rook (e2) half open file 50
w rook (e2) rank mobility 37
w rook (e2) attack bonus 3
b rook (h7) static eval -100
b rook (h7) half open file 50
b rook (h7) rank mobility -3
b rook (h7) attack bonus 3
b rook (a6) static eval -100
b rook (a6) mobility bonus -100
b rook (a6) rank mobility -30
w queen (b3) static eval -100
w queen (b3) attack bonus 1
b queen (c7) static eval 82
w king (h3) static eval -50
w king (h3) with pawns bonus 46
b king (e8) static eval -50
b king (e8) with pawns bonus 23
*** positional (white): P= 0.000 N= 0.020 B= 0.081 R= 0.290 Q=-0.100 K=-0.004
*** positional (black): P=-0.020 N= 0.059 B= 0.080 R=-0.283 Q= 0.082 K=-0.027
*** positional: white: 0.287 black: -0.109
*** attack bonus: white: 0.100 black: 0.080 (7 6)
*** attack defects bonus: white: 0.182 black: 0.072 (7*26 6*12)
*** pressure combo bonus: white: 0.400 black: 0.150 (3 7, 2 3)
*** final: white: 0.969 black: 0.193
============================
material (w): 0.000
pawn struct (w): 0.343
king safety (w): 0.383
============================
Score (white) 1.422
>
The scores are like this -50 are in millipawns (not the more common centipawns).
GL seems to like the white Rs a lot over the black ones. This gives about +0.57
to white on it's own.
It also has the three attack related terms ("attack bonus", "attack defects
bonus" and "pressure combo bonus") which total about +0.68 for white.
And then there's the pawn structure bonus of +0.343, and the king safety bonus
of +0.383.
The only major negative seems to be the Qs... about -0.19.
So, lot's of plusses and few minusses for white here! I feel this may be a good
position for helping me to debug GL's evaluation. :)
Cheers, Tim.
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.