Author: Will Singleton
Date: 08:54:15 06/08/04
Go up one level in this thread
On June 07, 2004 at 22:43:14, James Swafford wrote:
>On June 07, 2004 at 22:20:19, Will Singleton wrote:
>
>>Here's one from a game today on ICC, Alarm-Amateur. Amateur doesn't find the
>>right move, and loses. Shredder does well here, but not many (any?) others.
>>Anyone find Qd5 in less than a minute?
>>
>>[d]1rb1k2r/5p1p/p1nQpp2/1p6/4qP2/PN6/2P1B1PP/R3K2R b KQk -
>
>Mine does not. It likes Bb7. It doesn't know much though. :)
>
>What makes Qd5 'the right move'?
>
>I'm also interested in how you go about picking 'critical'
>positions from your games (since you come up with so many).
>Do you just look for where Amateur's score starts dropping,
>or do you analyze your games with a strong engine (like Shredder)
>and look for major score differences between its analysis and
>Amateur's, or ...?
>
>--
>James
Hi James,
Some games, like this one, feature score reversals, so I naturally want to look
to see why Amateur didn't find the right move. In this case, there are two main
variations: Bb7 leads to a position where white has control of the d file, plus
good attacking chances. Qd5 looks like it frees up black's pieces, giving black
a chance to attack, but leaving the d pawn isolated. Perhaps the reason it
takes so long here to cut thru the chaff is due to the open position, with
all kinds of tactics to explore, along with dynamic positional possibilities.
1... Bb7 2. Nc5 Qd4 3. Rd1 Qxd6 4. Rxd6 Ba8 5. Kf2 Rd8 {+0.85}
1... Qd5 2. Qxd5 exd5 3. O-O-O Ne7 4. Bf3 b4 5. axb4 Rxb4 {+0.15}
So, after looking at this, I conclude that my eval isn't accurate enough to help
quickly prune off bad lines, and therefore it takes way too long to find the
right move. And, perhaps recognition of positional features at interior nodes
would help to reduce the tree.
Will
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.