Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Poor assessment by Crafty 16.14

Author: Dave Gomboc

Date: 17:05:22 07/27/99

Go up one level in this thread


On July 27, 1999 at 19:51:17, KarinsDad wrote:

>On July 27, 1999 at 19:28:17, Dave Gomboc wrote:
>
>[snip]
>>>>>>
>>>>>>White: Kg2, Qd5, pawns on b3, f2, g3, h4.
>>>>>>Black: Kf6, Qe5, pawns on b4, g7, h7.
>>>>>>White to move.
>>>>>>
>>>>>>I played 43. Qc6+, and my opponent blundered with 43...Qe6??  I went on to win
>>>>>>the pawn ending after 44. Qxe6+ Kxe6 45. Kf3.
>>>>>
>[snip]
>>>
>>>43 ... Ke7 44. Qb7+ any 45. Qxb4 winning a pawn
>>>43 ... Kf7 44. Qb7+ Qe7 (to protect the b pawn) 45. Qxe7 and black's king is
>>>even further back than with what was played
>>>43 ... Kf5 44. Qf3+ Kg6 45. Qg4+ any 46. Qxb4
>>>43 ... Kf5 44. Qf3+ Ke6 45. Qb3+ any 46. Qxb4
>>>
>>>>
>[snip]
>>>
>>>I don't see Crafty not picking a swindle move in a lost game as being inferior.
>>>That's all.
>>
>>If I back up in the game, Crafty also picks a bad move earlier, ...Qe5, based on
>>the incorrect PV that is given when analyzing the position that I posted.  Bad
>>assessments are costly, whether they affect the PV in the particular position or
>>not, because there'll be a different position where it _does_ affect the PV.
>>
>>It's not a matter of swindling at all.  It's a matter of playing moves that make
>>the opponent's task difficult.  This is _extremely_ important.  See Jose's post
>>for more on that.  If Crafty knew (as Bob says it will soon) that the majority
>>will create an outside passer, it would have preferred to drop its b-pawn and
>>make White work harder for the win.  Crafty may also not have played ...Qe5,
>>leading to the position that I posted.
>>
>
>Staying with the original position you posted (and not some earlier move), if I
>look at the 4 variations above (which are the only ones I see off the top of my
>head other than the one Crafty picked), it appears that one leads to black's
>king being backed up an additional square and the rest lead to black losing a
>pawn and no apparent way to force perpetual check or to win the pawn back. If
>gaining the pawn was difficult to defend, then I may agree with you.
>
>Granted, for a human, the ones that lead to white winning the pawn are more
>complex positions due to the checks. For a computer, this is probably not true
>(at least I would hope not when searching 12+ ply down).
>
>So, a human may allow the loss of a pawn in order to get the checks in, but a
>computer would try to hang onto the pawn if it saw no way to force something by
>losing it.
>
>Your point is well taken, but this is a lousy example (IMO) and Crafty played as
>good as the position allowed.
>
>KarinsDad :)

It's a lousy example in the sense that other alternatives lost anyway.  It's
still a good example for showing that what it chose lost easily.

But see my reply to myself :) that shows Crafty definately losing a position
that may not even be a win... or a torturously long win, if it is one.

Dave



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.