1.g4 opening is losing?

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

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
jjoshua2
Posts: 78
Joined: Sat Mar 10, 2018 5:16 am

Re: 1.g4 opening is losing?

Post by jjoshua2 » Fri Mar 13, 2020 1:31 pm

I debated over using q output that goes from -100 to 100 but decided to use winrate percentage that goes from 0 to 100 instead, since I think it's just a little bit more clear - and the CP output doesn't really mean anything with every engine using a different formula and it being so far removed from actual pawns nodays.

jp
Posts: 1345
Joined: Mon Apr 23, 2018 5:54 am

Re: 1.g4 opening is losing?

Post by jp » Fri Mar 13, 2020 2:09 pm

Ovyron wrote:
Fri Mar 13, 2020 12:20 pm
(Leela remains over-optimistic for 1.g4.)
Well, the pure Leela results reported in this thread were:

384x30-t40-1573, 186M nodes: 28.6%

Sergio 384x30-t40-1808, over 800 million nodes: +1.18.

User avatar
MikeB
Posts: 4043
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: 1.g4 opening is losing?

Post by MikeB » Fri Mar 13, 2020 3:44 pm

Jouni wrote:
Mon Oct 14, 2019 12:20 pm


When I played 30 games between SF10 and SFdev 16 games were 0-1 (here 10 first). Other games were draws. Does this indicate, that 1.g4 is losing move? BTW in TCEC superfinal 12 SF and Komodo each won with black after 1.g4. In second game white played 2.h3 which didn't help either.
It's like playing with a handicap. The Chessbase live opening book, scores it at 33% , so does Fat Fritz on my machine. In my ongoing testing between identical engines, white scores less than 33% , similar to your experience. I have white scoring about 28% ( with a few wins)- it's somewhere between 120 to 160 Elo handicap I believe. So far 7 wins 77 draws 78 losses ...still ongoing.
Image

User avatar
MikeB
Posts: 4043
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: 1.g4 opening is losing?

Post by MikeB » Fri Mar 13, 2020 3:57 pm

A rare win by white with identical engines.



Same engines were used for 32 thread vs 64 thread testing on the 32 real core Threadripper. Of course here, they are both set to 1 thread.
Image

User avatar
Ovyron
Posts: 4219
Joined: Tue Jul 03, 2007 2:30 am

Re: 1.g4 opening is losing?

Post by Ovyron » Fri Mar 13, 2020 5:31 pm

What we mean is that all those white entities would lose against black's best attack, so once you know it it makes no sense to have black playing something else.

jp
Posts: 1345
Joined: Mon Apr 23, 2018 5:54 am

Re: 1.g4 opening is losing?

Post by jp » Fri Mar 13, 2020 6:35 pm

Ovyron wrote:
Fri Mar 13, 2020 5:31 pm
... against black's best attack, so once you know it ...
We don't know Black's best attack against 2. e3, though we might guess it starts 2 ... e5.

User avatar
Ovyron
Posts: 4219
Joined: Tue Jul 03, 2007 2:30 am

Re: 1.g4 opening is losing?

Post by Ovyron » Fri Mar 13, 2020 7:18 pm

fastgm's Depth 60 attack should suffice. YOU DON'T NEED ANYTHING BETTER, CHECKING OTHER BLACK MOVES IS A WASTE OF TIME.

Exa65536
Posts: 7
Joined: Sun Nov 25, 2018 4:28 am
Full name: Alexa Stevens

Re: 1.g4 opening is losing?

Post by Exa65536 » Sat Apr 04, 2020 6:12 am

Ovyron wrote:
Fri Mar 13, 2020 7:18 pm
fastgm's Depth 60 attack should suffice. YOU DON'T NEED ANYTHING BETTER, CHECKING OTHER BLACK MOVES IS A WASTE OF TIME.
Unfortunately, I don't have the hardware or the patience to find a drawing line against depth60 SF. Failing that, here's one with one billion nodes/move for black. I could perhaps push that to ten billion given a few weeks, but no further and I doubt the effort would grant any additional insight.



(Of course, I have no proof that the final position here is genuinely drawn - but it is evaluated at 0.00 at depth 40, so clearly black has already blundered by this point)

(Note: While I lack for both hardware and patience, this search is fully automated, so if someone better-supplied with those resources wants to throw more electricity at the problem I could clean up the code and post it somewhere)

User avatar
Ovyron
Posts: 4219
Joined: Tue Jul 03, 2007 2:30 am

Re: 1.g4 opening is losing?

Post by Ovyron » Sat Apr 04, 2020 3:05 pm

Exa65536 wrote:
Sat Apr 04, 2020 6:12 am
(Of course, I have no proof that the final position here is genuinely drawn - but it is evaluated at 0.00 at depth 40, so clearly black has already blundered by this point)
You don't need to wait for 0.00, if black's score goes below -1.30 it has already blundered. The trick is that you don't continue the game, but go back to a position where black played a stronger move. If you can't find any move that keeps the score above -1.30, you have found your mainline.

I have 3...c6 as third best behind 3...Nc6 and 3...h5.
Exa65536 wrote:
Sat Apr 04, 2020 6:12 am
While I lack for both hardware and patience, this search is fully automated, so if someone better-supplied with those resources wants to throw more electricity at the problem I could clean up the code and post it somewhere)
Yes, I think it'd be really interesting.

Exa65536
Posts: 7
Joined: Sun Nov 25, 2018 4:28 am
Full name: Alexa Stevens

Re: 1.g4 opening is losing?

Post by Exa65536 » Sat Apr 04, 2020 3:25 pm

Ovyron wrote:
Sat Apr 04, 2020 3:05 pm
You don't need to wait for 0.00, if black's score goes below -1.30 it has already blundered. The trick is that you don't continue the game, but go back to a position where black played a stronger move. If you can't find any move that keeps the score above -1.30, you have found your mainline.
There is another version of the search which allows both sides to backtrack arbitrarily, but for obvious reasons has branching-factor issues that make progress very slow. I previously had it solved up to -1.71, but then realized I'd forgotten to turn off contempt and had to restart (and increased leaf node evals from 10 Mn to 100 Mn for increased stability). That search has only been running since this morning, currently has 1. g4 d5 2. e3 Nc6 3. d4 e5 4. Nc3 Be6 5. h3 h5 6. gxh5 Nf6 as a line with an eval of -1.32

Post Reply