@1
Stockfish (3390) > Silman (2383)
Not that it matters much: those positions are completely equal.
By the way your move order to reach that position is questionable: 4...Nxe4!
Here is a better move order.
Note white refrained from Nc3 and preferred c3 instead.
They prepared with much more powerful cloud engines.
https://www.chessgames.com/perl/chessgame?gid=2136417
During the Candidates tournament one of the players said something that's obvious after you hear it, and answers questions like this.
If you use an engine at high depth, and actually put the moves on the board, letting it reach high depth each time, and explore many possibilities... then moves will either tend towards 0.00 or they will be a win/loss.
In this case the meaning is, there is no difference between 7...Bxe3 and 7...Bb6 since the engine will think they're both 0.00 if you do a thorough analysis.
The only difference that exists (for us) is how easy the move is for a human to play, in which case I agree with Silman that the solid center and open f file make white's game easier to play... this is also reflected in the statistics which say humans do better (at every level) with 7...Bb6.
I heard once (I think it was Dan Heisman who I paraphrase) that all of the evaluation numbers are sort of a fudge factor and with an infinitely powerful computer or perfect algorithm there would be no numbers. All positions would just be Win/Lose/Draw, like the tablebases show us.
#4
Yes, the initial position is a draw.
After one player makes some error it becomes a win for the other player.
So the point is to avoid errors and to steer towards positions where the opponent is more likely to make an error, and if he makes an error, then to prey on it.
I heard once (I think it was Dan Heisman who I paraphrase) that all of the evaluation numbers are sort of a fudge factor and with an infinitely powerful computer or perfect algorithm there would be no numbers. All positions would just be Win/Lose/Draw, like the tablebases show us.
Yeah, the only "true" (for lack of a better word) evaluations are mate in __ or draw.
Having said that, I do think that, on average, as black, I'd find it easier to play a 0.00 position than a 0.8 position even if both are a draw... of course sometimes the reverse is true, but like I said, on average, I'd tend to prefer 0.00 as black to 0.8.
But when the differences are very small, it's extremely logical to ignore the engine since even the programmers say the error is something like + or - 0.1
A fun way to test this is with a position like this
White to move must capture the bishop to restore material equality, however if you ask the engine to show the top 3 moves, it will start evaluating 3 different promotions... what you've done is trick the engine into calculating the same line 3 times
--
Here's an example of a result for that position at depth 30
-
Note this means I essentially tested the same position 3 times with identical hardware, software, and thinking time. The result was reaching 3 different evaluations spanning 0.13
#6
Yes, the engines consider completely meaningless underpromotions, kind of a bug.
Another phenomenon is oscillation of evaluation.
White typically starts +0.33, then after black's reply down to about 0.00, then after a white move +0.33 again etc. This reflects the value of the initial tempo up: a third of a pawn and it vanishes every time black evens the ply count.
#6
Yes, the engines consider completely meaningless underpromotions, kind of a bug.
Another phenomenon is oscillation of evaluation.
White typically starts +0.33, then after black's reply down to about 0.00, then after a white move +0.33 again etc. This reflects the value of the initial tempo up: a third of a pawn and it vanishes every time black evens the ply count.
Yeah, a lot of people don't seem to realize it's not enough to let it reach something like depth 30. It's useful to put a few of the engine's moves on the board. Sometimes when you do, you'll see the evaluation immediately shift a bit.
@8
Chess gets more complicated going from 32 to 26 men and then gets less complicated.
Thus engine evaluations of positions of > 26 men cannot be trusted: the positions are too complicated and the horizon effect obscures the evaluation.
32 men: 1.89 × 10^33
31 men: 1.71 × 10^34
30 men: 1.64 × 10^35
29 men: 1.53 × 10^36
28 men: 5.46 × 10^36
27 men: 1.05 × 10^37
26 men: 1.08 × 10^37
25 men: 6.14 × 10^36
24 men: 3.19 × 10^36
23 men: 5.66 × 10^35
Table 3
https://arxiv.org/pdf/2112.09386.pdf
However Stockfish computes 7...Bxe3 to be the best move for black, with 7...Bg4 second best 7...Bb6 further down the pecking order.
Is Silman's advice still valid or good here? I have looked at the opening stats and seen that at the amateur level, white wins more often after 7...Bxe3 and less often after 7...Bb6
At the master level it is usually a draw after 7...Bxe3 and black often wins after 7...Bb6