SFNNUE Fails!! In this Positions!!

Discussion of anything and everything relating to chess playing software and machines.

Moderators: Harvey Williamson, Dann Corbit, hgm

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
hgm
Posts: 25380
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: SFNNUE Fails!! In this Positions!!

Post by hgm » Fri Nov 20, 2020 4:01 pm

There is a difference between delaying by offering tough resistance, and delaying by silly sacrifices. I agree that in human blitz/bullet chess things are a bit different, because "the clock is a piece" there, and board positions that are game-theoretical losses can be certain wins there because of time.

I guess the thing that really counts is to maximize the distance to a trivial loss. Sacrificing a Queen or Rook in an equal position to delay a deep mate by 1 move only speeds up the loss, because the advantage of a Queen or Rook will be a trivial win for the opponent even at bullet speed. So you basically take a certain-loss-in-1 instead of the deep mate.
Get rid of the shit: vote for SHID!

User avatar
Master Om
Posts: 412
Joined: Wed Nov 24, 2010 9:57 am
Location: INDIA

Re: SFNNUE Fails!! In this Positions!!

Post by Master Om » Fri Nov 20, 2020 4:03 pm

hgm wrote:
Fri Nov 20, 2020 4:01 pm
There is a difference between delaying by offering tough resistance, and delaying by silly sacrifices. I agree that in human blitz/bullet chess things are a bit different, because "the clock is a piece" there, and board positions that are game-theoretical losses can be certain wins there because of time.

I guess the thing that really counts is to maximize the distance to a trivial loss. Sacrificing a Queen or Rook in an equal position to delay a deep mate by 1 move only speeds up the loss, because the advantage of a Queen or Rook will be a trivial win for the opponent even at bullet speed. So you basically take a certain-loss-in-1 instead of the deep mate.
Completely agree.
Always Expect the Unexpected

User avatar
MikeB
Posts: 4487
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: SFNNUE Fails!! In this Positions!!

Post by MikeB » Sun Nov 22, 2020 5:09 am

hgm wrote:
Fri Nov 20, 2020 2:09 pm

<snip>
When you are losing, going for the longest game is often a fatal mistake.
I love it when I see the computer played a mate delaying Qx check?. Now it's "I got this game won and don't blow it time" , especially by not seeing a stalemate trap
Image

User avatar
hgm
Posts: 25380
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: SFNNUE Fails!! In this Positions!!

Post by hgm » Sun Nov 22, 2020 10:48 am

In human-engine games this happens a lot. I played my Chu Shogi engine some time on the 81dojo server, and in one game it was ahead by the equivalent of a Queen and two Rooks, when it saw the opponent had a deep mate. So it sacrificed three Queen-class pieces and three Rook-class pieces in a row to delay the mate. But of course the human opponent did not see the mate at all. Being two Queens and a Rook ahead now, he had little trouble winning the end-game, though.
Get rid of the shit: vote for SHID!

Post Reply