12

In one of my recent games, I got to the following position, Black to play:

[FEN "2rr2k1/1q2bppp/p3pn2/np1p4/4PQ2/P1N2P1P/BP1R2P1/2R1B1K1 b - - 1 21"]

While playing the game, I analyzed d4 and it looked good even after Ne2 saving the rook but I considered it to be too risky to play and decided to play a safer move, dxe4. Stockfish 14+ NNUE gives -6.2 with depth 30. I further analyzed the position with the engine like about 10 moves ahead but couldn't see a material difference other than the existent +1 pawn. I don't understand how is this a positional -6.2

Can someone please explain why this is such a winning position for black? Because I couldn't understand the engine moves.

3
  • just as a comment: the "centi-pawn" evaluation has nothing to do with material anymore, but is just a way to portray the expected score. In a winning endgame with "only" one extra pawn, my engine often roars up to +70 before finding the forced mate in 30.. Stockfish finally decided to change the evaluation function in the newest 15.1 release to something sensible!
    – Hauptideal
    Dec 22, 2022 at 17:50
  • What has changed? Wasn't it already like this for a long time?
    – Minot
    Dec 22, 2022 at 18:58
  • 1
    Now there is an explicitly mentioned new convention, see here stockfishchess.org/blog/2022/stockfish-15-1 AFAIK, Stockfish was able to produce alternative evaluations before, but now it's kind of a paradigm shift that "allows for consistent evaluations in the future". And imagine this: Lc0, despite internally using a modern evaluation "format", converted it back to centipawn-values for the GUIs, because people are/were so used to it!
    – Hauptideal
    Dec 22, 2022 at 19:26

5 Answers 5

12

It's "only" -3 with standard LiChess analysis settings, but your point remains. Thing is your "10 moves ahead" (plies or moves?) are nothing when it comes to calculation - White's officers are a chicken den after the fox came in, as Germans say. Even the Q is in danger after Nh5, and White either must cede another pawn, or play Bxc4 (giving two dangerous pawns instead of one, and a dead meat Pb2), or let the Pd3 run, or immediately give the exchange after Nc4 (best move, says the engine). I'm quite a strong defender, but after some few more moves I tried it becomes clear what a desaster the White position is - and that even though Black displaces the knight - Nh5 or the seemingly tame Nd7 - in some variants. To see until the inevitable crash is way beyond my horizon even if my instincts already say White is in dire straits.

Fun experiment: Kh1 Qg1 instead of Kg1 Qf4. Black's advantage reduces to -0.5 after d4! So all these random tactics are the "real" reason of White's downfall.

4

Black is not winning immediately but the -5 to -6 evaluation looks correct to me. Here's why:

After black gets d4-d3 and Nc4, it becomes pretty tricky for white to bring their bishop into the game; hence, they have to exchange it for the knight on c4. Now, you have two connected pawns in white's territory. The plan for black is quite simple now- just bring back the knight to an active square(eg. b3), exchange the bishops, build a queen rook battery(to put pressure on weak pawns), and promote the pawns.

Of course, it's not straightforward and black has to put in some effort to win this.

2

Short answer : passed pawn. (If white dares then his pawn might fall or black might get protected passed pawn)

Black has active position as much as white. But 1. d4 makes white passive also black has a weird knight on a5. It has no job for now, but it can be so active just after a move.


[FEN "2rr2k1/1q2bppp/p3pn2/np1p4/4PQ2/P1N2P1P/BP1R2P1/2R1B1K1 b - - 1 21"]

1... d4 2. Ne2 Rxc1 3. Nxc1 {The knight on c1 is passive now. He has to struggle to become active.} Nc4 {The "weird" knight on h5 is so so good now. White can't remove the knight, just by pushing pawn. Cause the rook is under attack. So it's good to trade the bishop for the knight.} 4. Bxc4 bxc4 {White has protected passed pawn now. But b pawn is disturbing a bit. But there's no way it can help to break black's structure apart immediately.}

Another short answer as summary : The advantage is for passive pieces vs active pieces. The advantage is for a pawn advantage. The advantage is for advanced pawn. The advantage is for protected passed pawn.

After Nc4 White can just place the rook to another square, but that will make the rook passive or white can win another pawn on b2. The engine saw a way to make white passive that's why the advantage is so high. Though, according to Stockfish at 39 depth the advantage is 5.78 (something around there), it doesn't matter.

2

First black brings his pawn to d3.

Then he plays Nc4, attacking Rd2 and b2, so Bxc4 bxc4 is forced.

[Title "Analysis"]
[StartFlipped "1"]
[FEN "2rr2k1/1q2bppp/p3pn2/8/2p1PQ2/P1Np1P1P/1P1R2P1/2R1B1K1 w - - 0 25"]

Black then is up a pawn, and b2 is backward on a half open file. It will sooner or later fall. For example, after Nd7-c5-b3-d4-c2, (Q somewhere), Rb8-b7, Rdb8.

Stockfish's modern evaluation function (NNUE) is based on AI knowledge. The AI evaluates based on remembered probabilities. This type of position happens in a lot of openings, so the AI had comparable positions on board in a lot of training games and learned over time that the probability that black wins is very high, which results in the high evaluation. BTW, Leela sometimes gives an even bigger advantage for black than Stockfish.

2

Without doing actual tactical reading, I see:

  • White's light-square bishop is hemmed in and inactive; of 13 pawns on the board, 10 are on the same colour as that bishop, while Black's remaining bishop is on the opposite colour.
  • There is an open diagonal to threaten White's king, which Black's bishop and queen can move to directly. Of course this does not accomplish anything right now, but it points to White's king being generally less safe than Black's.
  • While being free to move, White's dark-square bishop can't do very much. It can form a battery with the queen, but that isn't targeting anything; it can defend the open diagonal, but being proactive is better than being reactive.
  • White's knight is corralled (it can only move backwards without being taken) and also pinned. (This is where the tactical analysis with d4 starts, since the natural way to exploit a minor piece pin is by threatening with a pawn.)
  • Black's knights don't look great either; but one of them can attack the queen directly (which can also be harassed with Bd6), while the other can easily maneuver back to the center (Na5->c6->e5; it's not a big issue to break the pin momentarily since the knight is still corralled and Black's rooks are connected).
  • Black probably has an easier time setting up on the c file, because of the queen positioning.
  • While Black's rooks are connected, White's are lined up on a diagonal (with the queen as well). This offers Black the plan of h6 - Nh7 - Bg5.

It's not hard to imagine multiple advantages like this adding up to 6 points.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.