Computer Chess Club Archives


Search

Terms

Messages

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

Author: John Lowe

Date: 15:51:04 12/26/02

Go up one level in this thread


On December 26, 2002 at 11:48:33, Martin Giepmans wrote:

>On December 25, 2002 at 12:31:17, John Lowe wrote:
>
>>On December 25, 2002 at 11:55:24, Martin Giepmans wrote:
>>
>>>On December 25, 2002 at 11:27:52, Martin Giepmans wrote:
>>>
>>>>On December 25, 2002 at 08:33:00, John Lowe wrote:
>>>>
>>>>>On December 25, 2002 at 08:11:55, Martin Giepmans wrote:
>>>>>
>>>>>>On December 25, 2002 at 00:51:57, John Lowe wrote:
>>>>>>
>>>>>>>On December 24, 2002 at 18:11:51, Martin Giepmans wrote:
>>>>>>>
>>>>>>>>On December 24, 2002 at 12:32:55, John Lowe wrote:
>>>>>>>>
>>>>>>>>>On December 23, 2002 at 15:37:07, Martin Giepmans wrote:
>>>>>>>>>
>>>>>>>>>>On December 23, 2002 at 15:16:44, Mike Byrne wrote:
>>>>>>>>>>
>>>>>>>>>>>[d]R6R/3Q4/1Q4Q1/4Q3/2Q4Q/Q4Q2/pp1Q4/kBNN1KB1 w - - 0 1
>>>>>>>>>>>
>>>>>>>>>>>So far , every Palm program and Chess Tiger have fatal errors with this
>>>>>>>>>>>position.  Supossedly, this is the largest number of  possible legal number of
>>>>>>>>>>>moves, 218, available from one position in chess.  If you can prove this wrong,
>>>>>>>>>>>you'll go down in History.
>>>>>>>>>>>
>>>>>>>>>>>http://www.rescon.de/Compu/schachzahl2_e.html
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>"Qd2xb2 mate!" says my program. 29 nodes calculated to find this brilliancy.
>>>>>>>>>>
>>>>>>>>>>It didn't blow up, the monitor didn't explode in my face, the AMD-processor
>>>>>>>>>>didn't implode, even my cigarette didn't catch fire ...
>>>>>>>>>>
>>>>>>>>>>I must say that I find this at least a little bit disappointing ;)
>>>>>>>>>>
>>>>>>>>>>Martin
>>>>>>>>>
>>>>>>>>>Hi Martin,
>>>>>>>>>
>>>>>>>>>You can borrow my failure if you like.
>>>>>>>>
>>>>>>>>Thanks!  Do you really want it back?
>>>>>>>>>
>>>>>>>>>How did you manage to calculate 29 nodes before you found a mate in one?
>>>>>>>>
>>>>>>>>I don't know if you grinned when you wrote that.
>>>>>>>>Do you mean that 29 is way too much for a mate in (only) one <grin>
>>>>>>>>or do you mean that 29 is not enough?
>>>>>>>>
>>>>>>>>Martin
>>>>>>>
>>>>>>>Of course I grinned. The whole position is a party game.
>>>>>>>
>>>>>>>I'm trying to imagine which order you would have to evaluate moves in to have 28
>>>>>>>misses before finding one of the mates.
>>>>>>>
>>>>>>29 nodes is the total number of nodes visited in the searchtree; even in
>>>>>>the first iteration this is usually (much) more than the number of moves
>>>>>>tried at the root. It probably did only a few moves at the root, found a
>>>>>>mate and stopped.
>>>>>>
>>>>>>>My program generates little piece moves first and would have stumbled over the
>>>>>>>knight mate - then it would(irrationally) have finished all the moves and
>>>>>>>selected its favourite mate - which is why it crashed.....
>>>>>>>
>>>>>>>
>>>>>>>Happy Christmas
>>>>>>
>>>>>>Happy Christmas too!
>>>>>
>>>>>DIY wants to know if it can be a Spider when it grows up.
>>>>
>>>>I don't think so. A real spider is much smarter.
>>>>I always have a plastic spider with me when I play in tournaments.
>>>>Till now it scared one operator (who suffered from arachnaphobia)
>>>>but it didn't impress the engines at all.
>>>>I'll have to find a better strategy ;)
>>>>
>>>>Martin
>>>
>>>Ah, I see that I misunderstood your question.
>>>Your program DIY wanted to know if can be a Spider ...
>>>Sure DIY! If you can ask such questions you must be quite smart already.
>>>That promises a bright future!
>>>
>>>Advice: count your legs every day. As long as you have 8 legs
>>>everything is probably OK.
>>>
>>>Martin
>>
>>At least Spider wasn't spooked by nine white queens and two connected enemy
>>passed pawns.
>
>It uses a movelist at the root (room for 400 moves, just to be safe)
>but for ply 2, 3, etc there are only lists for captures.
>Other moves are generated one by one.
>That could help in a case like this.
>
>>
>>I wonder if it's possible to generate that position within the rules?
>
>What was black's last move? b2, bxa2, axb2, ..?
>Too complicated. I give up ....
>Maybe there is a program that could help here?
>In any case, if this position arose in a legal game, it must have been
>a very weird game. Monkey1 - Monkey2?
>
>Martin

I'm not very good at hex to decimal. A move takes 4 bytes and I allow 3000H per
movelist. I remember putting three white queens on the board, seeing how big the
movelist was and adding 50%

400 moves is generous and would have saved DIY from trying to update the
piece-position array for white from a black move.

Perhaps if you were to give black positive (for white) scores in eval you could
get a computer to play the game - provided checkmate is ruled out of course.



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.