1.g4 opening is losing?

[Zenmastur]

I think if there is a draw, it will have an attraction (like a gift pawn) and then a cyclic draw. Or perhaps a very deep tactic. Most likely, the Grob is a loss, but I am not entirely convinced yet.

I had about a million positions, which I whittled down to 10K interesting ones by using SQL queries.
I analyzed the 10K positions to 26 plies and I am now analyzing all of the positions that were 1.5 pawns or less to 36 plies (about 7K, IIRC)

I started with a much smaller group and did a mass analysis. I took the top moves from the most important positions ( about 1,500 of them) and did analysis to 35 ply. This and old-fashion analysis to much greater depth on certain lines. With my prior and current analysis I'm even more convinced it's dead lost. But if you find a line you think has possibilities I'd like to see it.

Regards,

Zenmastur

[jjoshua2]

The threshold for a won position is probably about 70% or perhaps slightly higher, 71 or 72%, based on everything I've observed. For 1 g4?, it is about 68% for Black with 1...d5, so probably not a forced win although close enough to the line to leave that as a possibility.

Leelenstein 13 already is below 30% winrate after g4 d5.

Code: Select all

info depth 17 seldepth 45 time 430635 nodes 8239992 score cp 2917 hashfull 1000 nps 19133 tbhits 0 multipv 1 pv c2c4 e7e5 f1g2 d5c4 b1c3 g8e7 d1a4 b8c6 g1f3 e7g6 a4c4 c8e6 c4a4 g6f4 h1g1 f4g2 g1g2 d8d7 d2d3 h7h5 g4g5 e6g4 f3h4 e8c8 c1e3 a7a6 a1c1 c8b8 a2a3 f8e7 c3e4 c6d4 a4d7 d8d7 e3d4
info depth 17 seldepth 45 time 430635 nodes 8239992 score cp 2916 hashfull 1000 nps 19133 tbhits 0 multipv 2 pv f1g2 c8g4 c2c4 c7c6 d1b3 e7e6 b3b7 b8d7 b1c3 g8e7 c4d5 e6d5 d2d4 a8b8 b7a6 b8b6 a6d3 e7g6 h2h3 g4e6 g1f3 f8d6 h3h4 h7h6 g2h3 e6h3 h1h3 e8g8 h4h5 g6e7 h3h1 c6c5 b2b3 c5d4 c3a4 b6c6 f3d4 c6c8 c1b2 d6e5 a1d1
info depth 17 seldepth 45 time 430635 nodes 8239992 score cp 2844 hashfull 1000 nps 19133 tbhits 0 multipv 3 pv h2h3 h7h5 g4h5 e7e5 d2d3 b8c6 f1g2 g8f6 b1c3 c8e6 c1g5 f8e7 e2e3 d5d4 e3d4 e5d4 g5f6 g7f6 c3e2 d8d6 e2g3 e8c8 g1e2 f6f5 d1d2 e7f8 e1g1 f8h6 f2f4

EDIT: go 30 million easily did it. Probably still dropping too; thats why multiPV ordering is funny.

Code: Select all

info depth 25 seldepth 60 time 568285 nodes 24554674 score cp 2871 hashfull 1000 nps 16504 tbhits 0 multipv 1 pv c2c4 e7e5 f1g2 d5c4 b1c3 b8c6 d1a4 g8e7 g1f3 e7g6 a4c4 c8e6 c4a4 d8d7 d2d3 e6g4 c1e3 f8e7 e1c1 a8d8 d3d4 g4f3 g2f3 c6d4 a4d7 d8d7 f3b7 e8g8 c1b1 f8d8 b7a6 g6f4 e3c1 f4h3 h1f1 e7h4 f2f3 g7g6 e2e3 d4e6
info depth 25 seldepth 60 time 568285 nodes 24554674 score cp 2873 hashfull 1000 nps 16504 tbhits 0 multipv 2 pv f1g2 c8g4 c2c4 c7c6 d1b3 e7e6 b3b7 b8d7 b1c3 g8e7 c4d5 e6d5 d2d4 a8b8 b7a6 b8b6 a6d3 e7g6 h2h3 g4e6 g1f3 f8d6 h3h4 h7h6 g2h3 d8f6 h4h5 g6e7 b2b3 e6h3 h1h3 e8g8 c1b2 c6c5 d4c5 d7c5 d3d2 f6f5 h3h4 f8c8 a1d1 c5e4 c3e4 d5e4 f3d4 f5f6 d4f3 f6e6 f3d4 e6d5 e1f1 d6e5 d4f5 d5d2 f5e7 g8f8 d1d2
info depth 25 seldepth 60 time 568285 nodes 24554674 score cp 2884 hashfull 1000 nps 16504 tbhits 0 multipv 3 pv g4g5 e7e5 f1g2 b8c6 d2d4 e5d4 g1f3 h7h6 c2c3 f8e7 h2h4 h6g5 h4g5 h8h1 g2h1 d8d7 f3d4 d7g4 e1f1 e7g5 h1d5 g5c1 d1c1 c6d4 c3d4 g4d4 d5g2 g8f6 b1d2 f6g4 e2e3 d4b6 c1c3 c8e6 a1c1 e8c8 d2e4 e6d5 c3g7 f7f5 e4g3 d5g2 f1g2

LS14 that Pohl tested 46 elo stronger than LS13 after go 50M nodes.

Code: Select all

info depth 37 seldepth 90 time 1772501 nodes 27383748 score cp 2597 hashfull 1000 nps 15142 tbhits 0 multipv 1 pv f1g2 c8g4 c2c4 c7c6 d1b3 e7e6 b3b7 b8d7 b1c3 g8e7 c4d5 e6d5 d2d4 a8b8 b7a6 b8b6 a6d3 e7g6 h2h3 g4e6 g1f3 f8d6 h3h4 h7h6 g2h3 e6h3 h1h3 e8g8 h4h5 g6e7 b2b3 c6c5 h3h1 d8c8 c3a4 b6c6 d4c5 d7e5 d3c3 e5f3 c3f3 d6e5 c1b2 e5b2 a4b2 c6c5 b2d3 c5c2 e1f1 e7f5 d3b4 c2c1 a1c1 c8c1 f1g2 c1g5 g2h3 f8e8 f3g4 e8e4 g4g5 h6g5 b4d5 e4e2 h3g4 f5h6 g4f3 e2a2 h1c1 a2a3 c1c3 a7a5 f3g2 g8h7 c3c5 a3b3 c5a5 h6g4 d5e7 f7f6 e7f5 b3b2 g2g3 g4h6
info depth 37 seldepth 90 time 1772501 nodes 27383748 score cp 2507 hashfull 1000 nps 15142 tbhits 0 multipv 2 pv c2c4 e7e5 f1g2 d5c4 b1c3 b8c6 d1a4 g8e7 g1f3 e7g6 a4c4 c8e6 c4b5 e6d7 b5c4 g6f4 h1g1 f8d6 d2d3 e8g8 g2h1 c6a5 c4e4 f8e8 h2h4 d7c6 e4e3 b7b6 c3e4 a5b7 f3g5 d8d7 e4d6 c7d6 h1c6 d7c6
info depth 37 seldepth 90 time 1772501 nodes 27383748 score cp 2253 hashfull 1000 nps 15142 tbhits 0 multipv 3 pv h2h3 h7h5 g4h5 e7e5 d2d3 b8c6 f1g2 g8f6 b1c3 f8e7 c1g5 c8e6 e2e4 d5e4 d3e4 f6h5 g5e3 h5f4 e3f4 e5f4 c3d5 d8d6 d1d2 e8c8 e1c1 c6b4 a2a3 b4d5 e4d5 e6f5 g1e2

So winrate dropped 2.74% or 16.8 elo per %. So linearly, we need 436 elo stronger engine/search to solve to 0% winrate

[jp]

The threshold for a won position is probably about 70% or perhaps slightly higher, 71 or 72%, based on everything I've observed. For 1 g4?, it is about 68% for Black with 1...d5, so probably not a forced win although close enough to the line to leave that as a possibility.

So winrate dropped 2.74% or 16.8 elo per %. So linearly, we need 436 elo stronger engine/search to solve to 0% winrate

By win rate under 30%, I guess you mean draw rate about 70%, so total score under 65%? (I assume Larry's 72% mark is total score.)

I see its top choices are Bg2, c4 and h3.

[Zenmastur]

I see its top choices are Bg2, c4 and h3.

Good luck with c4 and h3 lines. They are no better than g5 or Bg2.

[jp]

Actually, I'm a bit confused by the LS output, i.e. "score cp 2597, ... 2507, ... 2253".

[MonteCarlo]

That output is expected score in hundredths of a percent, so it thinks white's best is an expected score of just under 26%

[jp]

Oh, of course. Thanks. I'm so used to "cp" meaning centipawn, not centi-percent.

[Ovyron]

What is Leela's percent translated to centi-pawns? My claim was that a -1.60 cp score was unstoppable but I have no idea what is its percentage equivalence.

[jp]

My claim was that a -1.60 cp score was unstoppable but I have no idea what is its percentage equivalence.

If I've got it right, 160 cp translates to 80.77% score for Leela.

[Ovyron]

Thanks. I guess that's not much (Leela remains over-optimistic for 1.g4.)