Stockfish isn't the strongest engine in the middlegame, however, and there are bound to be forced biasses due to the way it's programmed. Both of those would show in its evaluation of 1. e4 as slightly better than 1. d4, since 1. e4 tends to move towards an ending, where Stockfish is strongest, slightly the faster than 1. d4, which has hitherto tended to be regarded as slightly the stronger opening move.
Re 1. Nf3, human bias shows in the choice of 1. .... d5 s preferential.
Stockfish evaluation at depth 50 on all possible initial opening moves
Sort:
Jan 11, 2021
0
#43
drmrboss wrote:
Chess is almost solved , It is extremely drawish game. To win a game , a player must do a series of big mistakes that worth +5 material advantage in one side or value equivalent to one Rook and above. Sometimes a player has a margin to throw away even a queen +9.5 material and still a draw( compare first two diagrams) and sometimes a player has margin to throw away two potential queens (+ 19 potential materials) and still a draw. Current SF vs SF in 2 hour standard game in common 4 cores desktop lead to > 95 percent draw. ( Statistical p value >0.05 ) and we can conclude this is a draw. In next 10-20 years, we can see the draw rate will be increasing to 97-99%.
In short, u need to do extremely stupid moves to lose a game where the game is virtually extremely drawish. We still dont have 32 men tbs, but if you look at 7 men tbs you can see that you can do series of extremely bad moves and still draw.
Chess is not even close to being solved, wdym. Chess will never be solved probably and if it is it won't be soon.
Optimissed wrote:
Both of those would show in its evaluation of 1. e4 as slightly better than 1. d4
Stockfish 15 NNUE depth 46 shows 1.d4 as the best move
No, 1. g4 can't be a forced loss. In any case, Stockfish always over-evaluates. I always thought 1. c4 and 1. Nf3 were the strongest opening choices for white.
msarabi wrote:
Optimissed wrote:
Both of those would show in its evaluation of 1. e4 as slightly better than 1. d4
Stockfish 15 NNUE depth 46 shows 1.d4 as the best move
Since I think that 1. d4 is stronger than 1. e4, was that out of context??
0
#42Stockfish isn't the strongest engine in the middlegame, however, and there are bound to be forced biasses due to the way it's programmed. Both of those would show in its evaluation of 1. e4 as slightly better than 1. d4, since 1. e4 tends to move towards an ending, where Stockfish is strongest, slightly the faster than 1. d4, which has hitherto tended to be regarded as slightly the stronger opening move.
Re 1. Nf3, human bias shows in the choice of 1. .... d5 s preferential.>>>
Yep.
Better to quote the entire post in future, rather than giving part of a sentence which gives entirely the wrong impression.
Optimissed wrote:
No, 1. g4 can't be a forced loss. In any case, Stockfish always over-evaluates. I always thought 1. c4 and 1. Nf3 were the strongest opening choices for white.
1.g4 is a forced loss in about 10/10 games or 20/20 games in TCEC games with approx 1 billions nodes per seconds search. As they are high quality games, no more need to test 1000/1000 loss to assure reliability.
1 g4? loses by force.
The other 19 first moves draw with best play by both sides and thus are equivalent.
drmrboss wrote:
Optimissed wrote:
No, 1. g4 can't be a forced loss. In any case, Stockfish always over-evaluates. I always thought 1. c4 and 1. Nf3 were the strongest opening choices for white.
1.g4 is a forced loss in about 10/10 games or 20/20 games in TCEC games with approx 1 billions nodes per seconds search.
Wait, when did this happen?
I would think that if that was even remotely true, it would have been "Breaking News" in all chess websites.
@50
See Figures 5 and 31
https://arxiv.org/abs/2111.09259
Also proof from 1 g4 to a 7-men endgame table base win:
Yeah, just what I thought.
By the way, the game that you posted is a bad joke. It is like posting a single Paul Morphy game to prove that e4 wins by force.
@52
Please try and come up with an improvement for white then. You cannot.
tygxc wrote:
@52
Please try and come up with an improvement for white then. You cannot.
There may or may not be an improvement at every single move. The whole game is totally artificial, with both sides making highly suspicious moves.
Today's engines are very very very far away from correctly evaluating openings. Long variations like this are just ridiculous.
@54
"There may or may not be an improvement at every single move." ++ Point to one. You cannot.
"both sides making highly suspicious moves" ++ Point to one for white. You cannot.
"Today's engines are very very very far away from correctly evaluating openings." ++ True
"Long variations like this are just ridiculous." ++ It goes from 1 g4? to a 7-men endgame table base loss. Try to find an improvement for white. You cannot.
As I said, there are tons of places where you can improve. For the engine to correctly evaluate the entire line, it would need to spend days (weeks?) in the key position (which might be as late as move 40). Then if you find no improvement, go back one move and repeat. The entire process to disprove the Grob would take at least billions of years, and probably much much much more. This analysis that you posted is just nothing. It is so extremely shallow that literally every point might be wrong.
@56
"there are tons of places where you can improve"
++ There are only 54 white moves. Pick any of the 54 and propose an improvement.
"The entire process to disprove the Grob would take at least billions of years" ++ No, not at all.
"every point might be wrong."
++ There are 54 points. Propose an improvement over one of the 54 white moves.
Okay, for one thing: The whole Qb3-Qxb7 thing that white played goes against the spirit of the opening. White sacrifices 5 tempi (!!!) to win back the pawn: Qb3-b7-b4-a4-d1. This cannot be the right approach. You can spend those 5 moves for development and attack, just as Henri Grob intended.
On the other hand, I don't have a supercomputer in my room, and I also don't have 10 billion years to spare. Although, to be fair, that is not needed. We only have to wait a couple of decades, and we will have some computers that are quite a bit stronger than those of today. They will easily make the whole AlphaZero analysis look ridiculous.
@58
"The whole Qb3-Qxb7 thing that white played goes against the spirit of the opening"
++ Okay, so what white move do you suggest instead of 4 Qb3?
So I got this position as the final FEN of SF evaluation of initial position of 1. Nf3 (at depth 52)
And the pgn is 50/80 200:44:26 9.890.822.131k 18.586k +0,17 1.Nf3 d5 2.g3 c6 3.c4 g6 4.cxd5 cxd5 5.d4 Bg7 6.Bg2 e6 7.0-0 Ne7 8.b3 0-0 9.Qd2 Nbc6 10.Ba3 Qd7 11.Rc1 b6 12.e3 Bb7 13.Nc3 Rfc8 14.Rc2 Nf5 15.Rac1 h5 16.Bh3 a5 17.Na4 Qd8 18.Bxf5 exf5 19.h3 Bf6 20.Kg2 Bg7 21.h4 Bf6 22.Rc3 Ne7 23.Rxc8 Nxc8 24.Nc3 Nd6 25.Bxd6 Qxd6 26.Nb5 Qd7