Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Does this position blow up your program? 2

Author: Heiner Marxen

Date: 11:50:32 12/26/02

Go up one level in this thread


On December 26, 2002 at 04:29:01, Uri Blass wrote:

>On December 25, 2002 at 23:23:48, Drexel,Michael wrote:
>
>>[D] Q2Q2Q1/1R6/3BR3/3k3N/2RB2Q1/5R2/Q2Q2Bp/3N3K w - - 0 1
>>
>>Problem:
>>White to move
>>If white gives check it has to be mate in one.
>>Find a legal position with the highest number of possible checks.
>>
>>I discovered this 5 years ago. Took me about 16 hours to find it.
>>I think it is possible to find a proof that it is not possible to find more:
>>
>>85 mates in one! :-)
>>
>>I must have been crazy.
>>
>>Fritz 7.0.0.8 engine crashed in Analysis mode if I increase the number of lines
>>to 64 (maximum). Shredder 7 not.
>>I have checked it under Fritz 7 GUI and Shredder7 (=Fritz8)GUI.
>
>perft 1=142
>perft 2=12
>perft 3=1603
>perft 4=415
>perft 5=53576
>perft 6=45655(0.12 seconds)
>perft 7=5844247(movei optimized for perft 0.431 seconds)(yace 1.054
>seconds)(movei normal version for today 0.491 seconds)
>perft 8=3651048(13.139 seconds)(yace 23.774 seconds)(movei normal version for
>today 13.5 seconds)
>perft 9=441,943,521(37.504 seconds)(yace 90.728 seconds)(movei normal version
>42.271)

Chest agrees completely:

FEN: Q2Q2Q1/1R6/3BR3/3k3N/2RB2Q1/5R2/Q2Q2Bp/3N3K w - -
Counting width of legal tree, 9 plies deep...
Time (user) = 35.41 sec
dep          #nodes     quot        sum nodes
  0               1 [  0.000]               1
  1             142 [142.000]             143
  2              12 [  0.085]             155
  3            1603 [133.583]            1758
  4             415 [  0.259]            2173
  5           53576 [129.099]           55749
  6           45655 [  0.852]          101404
  7         5844247 [128.009]         5945651
  8         3651048 [  0.625]         9596699
  9       441943521 [121.046]       451540220
(Athlon 1500+)

>perft 10=950,008,655(1031.73 seconds)
>The last number was not verified by other programs.
>
>Finally we get perft 10>perft 9
>
>The surprising fact in these results is that
>perft 8/perft 7<perft 6/perft 5 and
>perft 4/perft 3<perft 6/perft 5.
>
>I give another challange
>
>Find the biggest number n such that perft(n)=0 and perft(n-1) >0
>
>here is a simple example that proves that n>11
>
>[D]k1q5/P2QQQQQ/KP6/PP6/4q3/5q2/6q1/7q w - - 0 1
>
>perft 1=3
>perft 2=4
>perft 3=4
>perft 4=6
>perft 5=6
>perft 6=8
>perft 7=8
>perft 8=10
>perft 9=10
>perft 10=10
>perft 11=5
>perft 12=0

Again, Chest agrees.

>It is also easy to see that n>12 by the following example
>
>[D]k1R5/P2QQQQQ/KP6/PP6/4q3/5q2/6q1/2q4q b - - 0 1
>
>Uri

Just one ply deeper:
FEN: k1R5/P2QQQQQ/KP6/PP6/4q3/5q2/6q1/2q4q b - -
Counting width of legal tree, 14 plies deep...
Time (user) = 0.00 sec
dep          #nodes     quot        sum nodes
  0               1 [  0.000]               1
  1               1 [  1.000]               2
  2               3 [  3.000]               5
  3               4 [  1.333]               9
  4               4 [  1.000]              13
  5               6 [  1.500]              19
  6               6 [  1.000]              25
  7               8 [  1.333]              33
  8               8 [  1.000]              41
  9              10 [  1.250]              51
 10              10 [  1.000]              61
 11              10 [  1.000]              71
 12               5 [  0.500]              76

Cheers,
Heiner



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.