It seems that a lot of people do not trust Stockfish's opinions on openings - saying that it cannot properly evaluate opening positions. I was wondering why this is, and if anyone has an answer.
Because stockfish is a computer, and not a human & it can look through millions of variations humans can not. So I'll give you an example of what I mean:
If stockfish sees two different openings.
(When I say "lines" I'm saying after they are trimmed down by the engine. These are ones the engine will look into. Generally, an engine won't evaluate too deeply into terrible lines once it sees it's winning it moves on to next line. So if you are looking at a move tree your legal moves will be viewed as this by the engine. So these represent all legal moves assuming a strong response)
Opening 1 =
major advantage in 1% of lines
moderate advantage in 1% of lines
slightly advantageous in 2% of lines
about equal in 16% of lines
slight disadvantage 10% of lines
moderate disadvantage 25% of lines
major disadvantage 45% of lines
Opening 2 =
major advantage in 0% of lines
moderate advantage in 15% of lines
slightly advantageous in `````25% of lines
about equal in 40% of lines
slight disadvantage 5% of lines
moderate disadvantage 5% of lines
major disadvantage `10% of lines
For a human opening 1 would be VERY hard to play, but possible to gain a great advantage if you played super accurately. on the other hand opening two would be much safer and more likely to give you an advantage with reasonable play. However since the engine spotted that one super narrow path with a great advantage in opening 1 it deems opening 1 to be the better of the two openings. It doesn't care that 80% of the lines are a disadvantage and only 4% are of any advantage at all. It spots that 1% shot with a great advantage and narrows in on it like a laser and goes that way. On the other hand the 2nd much less risky line where you got a 40% shot at an advantage & only 40% chance of disadvantage it just doesn't care it sees that one narrow path on line one deeming it the better line. It doesn't care if you gotta get 8 super accurate hard moves in a row to get that advantage and sac 3 pieces at just the right moment & if you don't see all that you are way worse off. It scoffs at that other line where most the moves are safer. Almost all humans would greatly prefer opening 2 to opening one. It's just the safer way to go for a win for a human. Odds are you got 40% at an advantage in that line, but the engine don't care it sees that long shot you can't and will drag you down a much riskier path in hopes for a slightly more accurate win in spite of it being in risky woods where blunders are around ever corner.
If stockfish sees two different openings.
(When I say "lines" I'm saying after they are trimmed down by the engine. These are ones the engine will look into. Generally, an engine won't evaluate too deeply into terrible lines once it sees it's winning it moves on to next line. So if you are looking at a move tree your legal moves will be viewed as this by the engine. So these represent all legal moves assuming a strong response)
Opening 1 =
major advantage in 1% of lines
moderate advantage in 1% of lines
slightly advantageous in 2% of lines
about equal in 16% of lines
slight disadvantage 10% of lines
moderate disadvantage 25% of lines
major disadvantage 45% of lines
Opening 2 =
major advantage in 0% of lines
moderate advantage in 15% of lines
slightly advantageous in `````25% of lines
about equal in 40% of lines
slight disadvantage 5% of lines
moderate disadvantage 5% of lines
major disadvantage `10% of lines
For a human opening 1 would be VERY hard to play, but possible to gain a great advantage if you played super accurately. on the other hand opening two would be much safer and more likely to give you an advantage with reasonable play. However since the engine spotted that one super narrow path with a great advantage in opening 1 it deems opening 1 to be the better of the two openings. It doesn't care that 80% of the lines are a disadvantage and only 4% are of any advantage at all. It spots that 1% shot with a great advantage and narrows in on it like a laser and goes that way. On the other hand the 2nd much less risky line where you got a 40% shot at an advantage & only 40% chance of disadvantage it just doesn't care it sees that one narrow path on line one deeming it the better line. It doesn't care if you gotta get 8 super accurate hard moves in a row to get that advantage and sac 3 pieces at just the right moment & if you don't see all that you are way worse off. It scoffs at that other line where most the moves are safer. Almost all humans would greatly prefer opening 2 to opening one. It's just the safer way to go for a win for a human. Odds are you got 40% at an advantage in that line, but the engine don't care it sees that long shot you can't and will drag you down a much riskier path in hopes for a slightly more accurate win in spite of it being in risky woods where blunders are around ever corner.
A position can only be: won or draw
And those weird numbers - what does this mean? Win or not?
Can you prove that the basic position a piece up is a win?
And those weird numbers - what does this mean? Win or not?
Can you prove that the basic position a piece up is a win?
@Sarg0n The weird numbers are different lines. Suppose there is 2 positions with 20 legal moves. As a human, I'd much prefer an easier to play position like the second one. Engines will always take the first one.
1 move = 1.9 to +3.5 on engine analysis (Great advantage)
0 move = 1.1 to +1.8 on engine analysis (Moderate advantage)
0 move = 0.5 to +1 on engine analysis. (Slight advantage)
2 moves = -0.5 to 0.5 (roughly equal)
6 moves = -0.5 to -1 (Slight disadvantage)
5 moves = -1.1 to -1.8 (Moderate disadvantage)
7 moves = -1.9 to -3.5 (Great disadvantage)
Of the 20 moves here you are not going to be super happy with 19 of them here. 17 you will be worse off to some degree, 1 you will be in good shape, and 2 you will be about equal. An Engine would be super happy and comfortable playing a position like this. This is a position an engine would love to have. One little inaccuracy though and you may wind up losing.
Lets say there is another position though.
0 move = 1.9 to +3.5 on engine analysis (Great advantage)
5 move = 1.1 to +1.8 on engine analysis (Moderate advantage)
5 move = 0.5 to +1 on engine analysis. (Slight advantage)
5 moves = -0.5 to 0.5 (roughly equal)
3 moves = -0.5 to -1 (Slight disadvantage)
1 moves = -1.1 to -1.8 (Moderate disadvantage)
1 moves = -1.9 to -3.5 (Great disadvantage)
In a position like this second one you have 10 moves that are an advantage to you, and 5 that are about equal and only 5 that could be bad for you. The engine would scoff at a position like this compared to the other position. It doesn't care that most moves are easy to make and it's easy to maintain an advantage. It doesn't care that most moves are safe, and you can slowly grind that advantage up. It simply doesn't care. It only sees that 1 big good move in the other line and determines that the other position is a better one in spite of the winning / advantageous path being much more narrow it's slightly taller. Engines 100% prefer the first position they don't care they have to calculate and be super accurate on every move for the next 8 moves in a row. They don't care if 99.99% of moves are losing if they see the 1 winning move. That means nothing to them they already ran the numbers.
1 move = 1.9 to +3.5 on engine analysis (Great advantage)
0 move = 1.1 to +1.8 on engine analysis (Moderate advantage)
0 move = 0.5 to +1 on engine analysis. (Slight advantage)
2 moves = -0.5 to 0.5 (roughly equal)
6 moves = -0.5 to -1 (Slight disadvantage)
5 moves = -1.1 to -1.8 (Moderate disadvantage)
7 moves = -1.9 to -3.5 (Great disadvantage)
Of the 20 moves here you are not going to be super happy with 19 of them here. 17 you will be worse off to some degree, 1 you will be in good shape, and 2 you will be about equal. An Engine would be super happy and comfortable playing a position like this. This is a position an engine would love to have. One little inaccuracy though and you may wind up losing.
Lets say there is another position though.
0 move = 1.9 to +3.5 on engine analysis (Great advantage)
5 move = 1.1 to +1.8 on engine analysis (Moderate advantage)
5 move = 0.5 to +1 on engine analysis. (Slight advantage)
5 moves = -0.5 to 0.5 (roughly equal)
3 moves = -0.5 to -1 (Slight disadvantage)
1 moves = -1.1 to -1.8 (Moderate disadvantage)
1 moves = -1.9 to -3.5 (Great disadvantage)
In a position like this second one you have 10 moves that are an advantage to you, and 5 that are about equal and only 5 that could be bad for you. The engine would scoff at a position like this compared to the other position. It doesn't care that most moves are easy to make and it's easy to maintain an advantage. It doesn't care that most moves are safe, and you can slowly grind that advantage up. It simply doesn't care. It only sees that 1 big good move in the other line and determines that the other position is a better one in spite of the winning / advantageous path being much more narrow it's slightly taller. Engines 100% prefer the first position they don't care they have to calculate and be super accurate on every move for the next 8 moves in a row. They don't care if 99.99% of moves are losing if they see the 1 winning move. That means nothing to them they already ran the numbers.
@#2 like <3.
Also.. I'm pretty sure stockfish has a book fed to it when it plays because its normal opening choices are abysmal? (somebody can correct me)
Also.. I'm pretty sure stockfish has a book fed to it when it plays because its normal opening choices are abysmal? (somebody can correct me)
This topic has been archived and can no longer be replied to.