Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Another Toga 1.0 blunder (blitz game)

Author: Dagh Nielsen

Date: 10:51:59 10/13/05

Go up one level in this thread


On October 13, 2005 at 13:21:29, Günther Simon wrote:

>Well S9.1 makes the same 'blunder' at least in a fast game...
>What speed has your machine? I did this analysis on my slow machine P3 650,
>but as it still holds c6/depth 21 as best after 47 sec, it will hold it for at
>least 14 sec on a P4 2.67 (ratio 3.3:1 measured here, P3s are relatively
>faster...)
>
>I guess people should stop concluding too much from bullet/fast blitz sudden
>death tcs, this just leads often to wrong conclusions like here.
>(the result was just a random result due to thinking times of 0-3 sec from
>an equal game at around move 30.)

Thanks for your comment.

Sorry, I did not examine the move c7-c6 in depth, I just know that when I
watched the game being played, I instantly feared disaster when it chose to play
c7-c6. My point is, for Black to draw this type of positions, usally a blockade
is necessary (or otherwise a passer protected by the bishop).

Before posting, I quickly let Fruit look at the position before c7-c6 with
multiple candidate moves displayed. It also evaluated h6-h5 as equal, and in the
principal variation the move c7-c6 was not included, so I just assumed black had
another positionally more sound way of playing.

However, it doesn't really matter if the position is lost, either before or
after c7-c6. It's just one example where Fruit needlessly enters this kind of
endgames because it evaluates them as equal, where they often are not. You may
notice that Fritz 9 evaluated it as giving white a sligt advantage, and punished
Fruit very well playing white moves leading to a win, moves that Fruit evaluates
as inferior for white (if you look at each move with Fruit). The same is true in
other examples against Schredder, the position is equal, but Fruit enters an
exchange down endgame, evaluate them as equal, while Scredder or Fritz 9
recognizes they are inferior.

My machine is an AMD Athlon 2400.

I do not agree that we should not conclude anything from blitz games. These are
features of the evaluation functions, and Fruit would make the same type of
flawed evaluations also with longer time controls.

Frankly, I have seen it happen too many times for you to simply discard the
phenomenon because I present a blitz example.

Also, what is blitz today, was long time control a few years ago, and what is
long time control today, will be blitz in a few years, due to computers
increasing their speed. If you know what you are doing, you are quite able to
make conclusions about engine weaknesses from blitz games, I would even say that
it is more useful to play many blitz games instead of a few long control games
because you get more examples to judge from.



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.