Author: Stuart Cracraft
Date: 23:30:37 01/30/98
Go up one level in this thread
On January 30, 1998 at 18:06:17, Bruce Moreland wrote:
>
>On January 30, 1998 at 14:50:18, Stuart Cracraft wrote:
>
>>8/7p/1p2KPb1/p5k1/P1P5/1P6/8/8 b - - 0 1
>>
>>The right move looks to this human to be h5. Only a very
>>long search saw this for my program and I am wondering what
>>heuristic I am missing:
>>
>>Phase = Endgame, 60 moves in 300 secs, target = 32767 sec, Root score =
>>828
>> 1- 0.0 8 23 Bb1
>> 1. 0.1 392 24 Bb1
>> 2+ 0.1 828 34 Bb1
>> 2. 0.1 108 96 Bb1 Ke5
>> 3+ 0.1 492 162 Bc2
>> 3. 0.1 108 596 Bc2 Ke5 Kg4
>> 4+ 0.1 592 1212 b5
>> 4. 0.2 208 2692 b5 Ke5 Kg4 cxb5
>> 5+ 0.2 428 7680 Bf7+
>> 5. 0.3 108 12337 Bf7+ Kxf7 Kg4 b4 axb4 Kf8
>> 6+ 0.4 328 17455 b5
>> 6. 0.8 308 40822 b5 f7 bxa4 f8=Q Bf7+ Kxf7 axb3
>> 7- 2.2 208 124063 b5
>> 7& 2.3 192 132253 b5 f7 Bxf7+ Kxf7 b4 Kf8 Kf4 Ke7 h5
>> 7. 2.4 192 136224 b5 f7 Bxf7+ Kxf7 b4 Kf8 Kf4 Ke7 h5
>> 8+ 3.8 528 222438 b5
>> 8. 6.2 508 381529 b5 f7 Bxf7+ Kxf7 bxa4 bxa4 Kf4 Kf8 h5
>>Kg8
>> Ke3
>> 9- 18.8 408 1163781 b5
>> 9& 24.9 28 1524091 b5 cxb5 Kh6 Ke7 Kh5 b4 axb4 f7 Bxf7
>> 9& 28.4 108 1735916 h5 c5 h4 cxb6 Bf7+ Kxf7 Kg4 b7 Kf3 b8=R
>> 9. 33.9 108 2051702 h5 c5 h4 cxb6 Bf7+ Kxf7 Kg4 b7 Kf3 b8=R
>>10+ 37.7 228 2281466 h5
>>10& 47.2 528 2831861 h5 Ke7 h4 f7 Bxf7 b4 Bxc4 bxa5 bxa5 Kd7
>>Kg4
>> Ke7
>>10. 81.9 528 5035141 h5 Ke7 h4 f7 Bxf7 b4 Bxc4 bxa5 bxa5 Kd7
>>Kg4
>> Ke7
>
>I don't think you are necessarily missing a heuristic, I think it's
>worse than this, I think you have some kind of horrendous bug.
>
>My evidence is that 25 seconds into the search, you are playing 1. ...
>b5, which manages to *lose*.
>
>bruce
Correct. It was a bug in square-of-the-pawn evaluation in my
static evaluator. Basically, I was considering every pawn a passed
pawn and applying the rule of the square-of-the-pawn. With that
fixed, behavior looks better. Solution in approx. 3.5 seconds on
a slow 486.
There seems to be some residual horizon effect in the PV, but the
program sees the advancing h-pawn.
K2 1/26/98 0.25
Type ? for help.
Transposition table: Entries=100K Size=4800K
Pawn transposition table: Entries=31K Size=2750K
? ? ? 60 seconds per problem.
weirdpos
h5
black
. . . . . . . .
. . . . . . . p
. p . . K P b .
p . . . . . k .
P . P . . . . .
. P . . . . . .
. . . . . . . .
. . . . . . . .
Phase = Endgame, 60 moves in 300 secs, target = 60 sec, Root score = 28
1& 0.3 41 1 Kf4
1. 0.3 41 18 Kf4
2& 0.3 23 43 h5 Ke5
2+ 0.4 359 76 Bf5+
2. 0.4 141 126 Bf5+ Ke5 Kg4
3+ 0.6 293 595 Bf7+
3& 0.6 293 688 Bf7+ Kd6 Kxf6 c5
3. 0.6 293 737 Bf7+ Kd6 Kxf6 c5
4- 0.9 217 2251 Bf7+
4& 1.0 59 2536 Bf7+ Kxf7 h5 Ke6 Kg4
4& 1.0 152 2635 h5 f7 Bxf7+ Ke5 Bxc4 bxc4
4. 1.0 152 2675 h5 f7 Bxf7+ Ke5 Bxc4 bxc4
5+ 1.3 268 3563 Bf7+
5& 1.6 268 5013 Bf7+ Ke5 h5 b4 axb4 Ke4
5. 2.2 268 8508 Bf7+ Ke5 h5 b4 axb4 Ke4
6+ 3.5 358 14327 h5
6& 3.7 387 15480 h5 f7 Bxf7+ Ke5 h4 Ke4 Bg6+ Ke3
6. 5.7 387 25404 h5 f7 Bxf7+ Ke5 h4 Ke4 Bg6+ Ke3
7- 10.9 309 54697 h5
7& 11.8 159 58410 h5 f7 Bxf7+ Kxf7 Kg4 b4 axb4 Kf8 h4
7. 13.6 159 67539 h5 f7 Bxf7+ Kxf7 Kg4 b4 axb4 Kf8 h4
8+ 14.8 269 72125 h5
8& 16.1 456 78201 h5 f7 Bxf7+ Ke5 h4 Ke4 Bg6+ Kf3 h3 c5
8. 29.4 456 149031 h5 f7 Bxf7+ Ke5 h4 Ke4 Bg6+ Kf3 h3 c5
9- 60.8 381 317626 h5
9& 67.7 159 361420 h5 f7 Bxf7+ Kxf7 Kg4 b4 axb4 Kf8 h4 Kg7 h3
9. 79.6 159 419150 h5 f7 Bxf7+ Kxf7 Kg4 b4 axb4 Kf8 h4 Kg7 h3
10. 79.7 159 419167 h5 f7 Bxf7+ Kxf7 Kg4 b4 axb4 Kf8 h4 Kg7 h3
c5
My move is : h5
Time = 79.7 Rate=5259 Nodes=[382001/37166/419167]
Eval=33428 RptCnt=1029 NullCut=11914
Extensions: Check=8717 Recapture=2126 Pawn=0 PassedPawn=0
OneMove=0 Threat=0 DeepSearch=0 Mate=0
Material=[400/600] Worst=[159/-632]
Hash: Success=57% Collision=40%
Hash Counts: ReadHits=83896, ReadProbes=145133, WriteColl=34388,
WriteTries=85548
Pawn Hash: Success=96% Collision=2%
Pawn Counts: ReadHits=64413, ReadProbes=66736, WriteColl=61,
WriteTries=2323
white h6
. . . . . . . .
. . . . . . . .
. p . . K P b .
p . . . . . k p
P . P . . . . .
. P . . . . . .
. . . . . . . .
. . . . . . . .
+++ h5 h5
Correct=1 Total=1 Percent=100%
********* Missed *********
********* Grand totals **********
Correct=1 Total=1 Percent=100%
Depth= 10.0/16.0 Time = 79.7 Rate=5260 Nodes=[382001/37166/419167]
EvalCnt=33428 RptCnt=1029 NullCut=11914
Extensions: Check=8717 Recapture=2126 Pawn=0 PassedPawn=0
OneMove=0 Threat=0 DeepSearch=0 Mate=0
Hash: Success=57% Collision=40%
Pawn Hash: Success=96% Collision=2%
?
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.