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.
zullil
Posts: 6170
Joined: Mon Jan 08, 2007 11:31 pm
Location: PA USA
Full name: Louis Zulli

Re: 1.g4 opening is losing?

Post by zullil » Mon Mar 09, 2020 9:14 am

Ovyron wrote:
Mon Mar 09, 2020 7:55 am
That's a reason I hate such things, there's just no way to know when you're done...
Just perform a full (no reductions, no extensions) alpha/beta search to depth 52. :wink:

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

Re: 1.g4 opening is losing?

Post by Ovyron » Mon Mar 09, 2020 9:34 am

Is that some survival game where the challenge is staying alive until depth 52 is over?

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

Re: 1.g4 opening is losing?

Post by jp » Mon Mar 09, 2020 1:46 pm

Uri Blass wrote:
Sat Mar 07, 2020 6:19 am
Here is another example
Stockfish needed a lot of time to find e4 and not e3 and could not see it at depth<95 (I used 1024 mbytes hash and 3 cores)
Note that chest with a single core was faster in seeing e4 is the only move that force mate in 8.



Stockfish_20030109_x64_modern:
<snip>
94/18 08:44 3,206,095k 6,111k +M9 1.e3 Ke7 2.Bb5 Nd7 3.Qg4 Nf6 4.Qg7+ Ke6 5.Nc3 Kd6 6.Qxf6+ Kc7 7.Qc6+ Kb8 8.Ba6 Ka7 9.Qb7+
95/18 10:54 4,041,755k 6,171k +M8 1.e4 Nc6 2.Qh5+ Ke7 3.Qc5+ Kd7 4.d4 Nb4 5.Bb5+ Ke6 6.Qf5+ Ke7 7.Bg5+ Kd6 8.Qe5+
Komodo 13.2 shows mate in 8 from depth 21:

+M8 1.e4 Nd7 2.d4 Nf6 3.Qf3 Nd7 4.Bg5 Ne5 5.Bb5+ Nd7 6.Qh5+ Kf8 7.Bc4 Nf6 8.Qf7#

Surely this is largely luck, though, i.e. whether a given engine happened to like 1.e4 early, before the depths when it started announcing mates?

For the N on the other side,



Komodo 13.2 does like 1.e3 to begin with, and switches to

+M8 1.e4 Nf6 2.Qf3 Nd7 3.d4 Kd8 4.Qf7 Kc7 5.Bf4+ Kc8 6.Ba6+ Kd8 7.Bb7 Ne5 8.Bg5#

around depth 25.

SF10 takes till depth 47 to find mate in 9 (with 1.e3) and did not find the mate in 8 at depths <=67.

zullil
Posts: 6170
Joined: Mon Jan 08, 2007 11:31 pm
Location: PA USA
Full name: Louis Zulli

Re: 1.g4 opening is losing?

Post by zullil » Mon Mar 09, 2020 4:32 pm

jp wrote:
Mon Mar 09, 2020 1:46 pm

For the N on the other side,



Komodo 13.2 does like 1.e3 to begin with, and switches to

+M8 1.e4 Nf6 2.Qf3 Nd7 3.d4 Kd8 4.Qf7 Kc7 5.Bf4+ Kc8 6.Ba6+ Kd8 7.Bb7 Ne5 8.Bg5#

around depth 25.

SF10 takes till depth 47 to find mate in 9 (with 1.e3) and did not find the mate in 8 at depths <=67.
Sf-dev finds mate-in-8 at depth 52 (one thread, 16 GB hash)

info depth 52 seldepth 18 multipv 1 score mate 8 nodes 34076245 nps 2512071 hashfull 12 tbhits 0 time 13565 pv d2d4 g8f6 e2e4 e8d8 d1f3 f6e8 f3f8 d8d7 c1f4 d7e6 f1b5 e8f6 c2c4 f6e4 d4d5

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

Re: 1.g4 opening is losing?

Post by jp » Mon Mar 09, 2020 6:25 pm

With both Ns, Black lasts a little longer.



Komodo 13.2 finds mate in 11:

depth=22 +M12 1.e4 Nf6 2.e5 Nd5 3.Qg4 Ne7 4.Bb5+ Kd8 5.Qh4 Nc6 6.Bxc6 Kc7 7.Qxe7+ Kxc6 8.Qd6+ Kb7 9.e6 Ka7 10.e7 Kb7 11.e8=Q Ka7 12.Qeb8#

depth=23 +M11 1.e4 Nf6 2.Qf3 Nfd7 3.d4 Ke7 4.Bg5+ Kd6 5.Bd8 Nc6 6.Qf4+ Nde5 7.dxe5+ Kd7 8.e6+ Kxd8 9.Qd6+ Ke8 10.Qd7+ Kf8 11.Qf7#


SF10 find it around depth 32:

depth=33 +M11 1.e4 Nf6 2.d4 Nc6 3.Qf3 Ng8 4.Qh5+ Ke7 5.Bc4 Nf6 6.Qf7+ Kd8 7.Bg5 Ne7 8.Qxf6 Kc7 9.Qxe7+ Kb6 10.Qd7 Ka5 11.Qb5#

fastgm
Posts: 552
Joined: Mon Aug 19, 2013 4:57 pm
Contact:

Re: 1.g4 opening is losing?

Post by fastgm » Mon Mar 09, 2020 7:07 pm

Stockfish 11 C0 vs Stockfish 11 C0, 10 cores, 32 GB Hash
1.g4 d5, every move depth 60
Duration: 33 days


zullil
Posts: 6170
Joined: Mon Jan 08, 2007 11:31 pm
Location: PA USA
Full name: Louis Zulli

Re: 1.g4 opening is losing?

Post by zullil » Mon Mar 09, 2020 8:20 pm

fastgm wrote:
Mon Mar 09, 2020 7:07 pm
Stockfish 11 C0 vs Stockfish 11 C0, 10 cores, 32 GB Hash
1.g4 d5, every move depth 60
Duration: 33 days

Thanks for the data point! I assume "C0" means "Contempt = 0". What endgame tables were the engines using?

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

Re: 1.g4 opening is losing?

Post by Ovyron » Mon Mar 09, 2020 8:32 pm

You still think it's drawn zullil? What do you think someone could do to find a single drawing line for white? (to at least put the ball on the other camp, people having to find a black win against that line)

fastgm
Posts: 552
Joined: Mon Aug 19, 2013 4:57 pm
Contact:

Re: 1.g4 opening is losing?

Post by fastgm » Mon Mar 09, 2020 8:40 pm

zullil wrote:
Mon Mar 09, 2020 8:20 pm
Thanks for the data point! I assume "C0" means "Contempt = 0". What endgame tables were the engines using?
Yes, "C0" means Contempt = 0
All 6-men syzygy tablebases were used.

Complete logfiles: http://www.fastgm.de/schach/1.g4-d5_log.7z

Andreas
Last edited by fastgm on Mon Mar 09, 2020 8:51 pm, edited 1 time in total.

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

Re: 1.g4 opening is losing?

Post by jp » Mon Mar 09, 2020 8:50 pm

Dann Corbit wrote:
Fri Mar 06, 2020 10:56 am
For absolute proofs, 13 is pretty well the limit for chest319

But ChestUCI has special parameters for much deeper mates. I have found mates of more than 60 plies with ChestUCI. However, you lose the "proof of minimal mate" in the special ChestUCI mates. So it finds a sure mate, but there could be a shorter one.
I see. That could be useful to know.
fastgm wrote:
Mon Mar 09, 2020 7:07 pm
Stockfish 11 C0 vs Stockfish 11 C0, 10 cores, 32 GB Hash
1.g4 d5, every move depth 60
Duration: 33 days
Great!
Ovyron wrote:
Mon Mar 09, 2020 8:32 pm
You still think it's drawn zullil? What do you think someone could do to find a single drawing line for white?
If it requires a "tablebase-like" draw, it's going to be very hard to find.

Post Reply