17

After a trade that sacrified a bishop and a knight for two pawns and a rook, the game reached the following position: (White to move.)

[FEN "2r2qk1/1b1n1pb1/p4n1p/1p4p1/4P3/1BN2P2/PPP3PP/R2Q1RK1 w - - 0 17"]

At this point the engine was evaluating the position as -1.65.

I am confused with this evaluation. After this trade, white can go Qe2 and soon take claim of the open file with a rook and it seems that all white's pieces are with good activity and the pawn structure seems better than black's, so at first glance it seems to me that not only there is material advantage (white's up a pawn) but also a positional one.

I tried looking at the engine's suggested moves to see what I was missing and the moves are obscure to me. For instance, after 1.Qe2, the engine suggests 1...Bc6 and after 2.Rfd1 it suggests 2...Nh5. Both these moves don't seem to accomplish much to me. I can see a threat of jumping the knight to f4 eventually but it does not seem so dangerous. What makes some sense to me is that the bishop in g7 will probably be able to disrupt the pawns on the queenside either by doubling the pawns on the c-file by taking the knight or by taking on b2 if white has to move the knight due to b4 or some other threat. In any case, this should take some commitment and due to white's pieces' activity it seems that this might not be easy to go unpunished or simply stopped. Furthermore, this would justify to me perhaps an evaluation closer to 0, not -1.65.

Maybe worth noticing that the engine suggests 1.Qd2 instead of 1.Qe2, but analyzing this branch wasn't very fruitful either. For instance, one variation is as follows:

[FEN "2r2qk1/1b1n1pb1/p4n1p/1p4p1/4P3/1BN2P2/PPP3PP/R2Q1RK1 w - - 0 17"]

1.Qd2 a5 2.a3 a4 3.Ba2 Ne5 4.Rad1

At this point, the evaluation is at -0.83 and I don't see a clear problem either.

This is why I'd appreciate if someone can explain why the engine likes black in the first highlighted position so much.

6
  • 1
    Note that White is not up a pawn. Even if one were to take the values of pieces for fixed, the 1, 3, 3, 5, 9, upon which I guess you based that pawn up statement, are only some easy to work with numbers. GM Larry Kaufman for instance has suggested Pawn = 1, Knight = 3.25, Bishop = 3.35, Rook = 5, Queen = 9.75. Basing on these values, White is only up by 0.4 pawn units. Add in that Black has the bishop pair it's not clear at all that White is up material. And then of course piece values aren't static, depending on how active they can become, their value may rise and fall further.
    – koedem
    Feb 11 at 13:48
  • @koedem: Doesn't LK say Bishop = 3.25 but add .5 for Bishop pair? Using that you get exactly even material. Feb 11 at 17:38
  • @NoahSnyder yes, that's basically what I'm saying.
    – koedem
    Feb 11 at 20:35
  • @koedem: There’s a typo in your Bishop value? Feb 11 at 20:54
  • @NoahSnyder well, this is a slightly different system I guess. Larry Kaufman has posted a number of evaluation systems over time, this is the one I use in my chess engine.
    – koedem
    Feb 12 at 10:40
9

Interestingly, even without looking ahead and just evaluating the immediate position, Stockfish thinks Black is ahead by nearly half a pawn despite agreeing that White is ahead in raw material. Looking at the sub-scores, the "Imbalance" category is where most of Black's score is coming from.

"Imbalance" looks at which pieces are on the board and makes assumptions about how well they coordinate together. For example, each rook gets a bonus for each friendly knight and bishop but a penalty for other friendly pieces, and a penalty for enemy bishops and queens but a bonus for enemy knights and pawns.

In this position, Black is getting a bonus for having the bishop pair. White seems to be getting a penalty for having two rooks and a queen (the algorithm apparently thinks rooks don't work well with friendly rooks or queens) while Black's queen gets a bonus because the extra enemy rook and pawns are considered as potential targets just for existing. Overall the synergies between the various pieces actually give Black more of a score boost than the bishop pair itself does, and in total it's enough for Stockfish to think it outweighs the material deficit.

1
  • 3
    The penalty for two rooks and a queen is what Larry Kauffman calls "the principle of the redundancy of major pieces." He argues that the principle of redundancy also explains the value of the bishop pair. Feb 11 at 17:42
6

2 pieces for the rook is often better. In this position, black has a better position with more good squares for his pieces, and a clearer long term plan (attack kingside with good knights). The e5 square is a great square for the knight. The other knight can go Nh5-f4. The b2 pawn is weak and in the eye of the black bishop, which makes it harder for white to put his knight on d5. White's bishop is hitting a wall and no clear plan on how to improve it - black will just move his king and push pawns up. Black plans to break the kingside later on with a g4 push. The only plan for white is to double on d file and eventually put a knight on d5, which black is likely to be fine exchanging his bishop for.

In simple terms, a knight on e5 and another on f4 with a bishop x-raying b2 (limiting the white knight) and the rook x-raying c2 (limiting the white bishop) is a pretty clear advantage. Which side would you rather be?

Also, a knight on f4 is extremely dangerous for white - it would instantly become black's best piece. It's setting up for sacrifices, pawn pushes, and the queen swinging over to the kingside.

3

This might not answer your question as it already has some good ones. But just out of curiosity, I ran your game with two separate Stockfish engines, running at their full capacity. I wanted to see how the game would continue, assuming the players give their optimal moves. (The word optimal may not be the best choice in this context. Please take it with a grain of salt).

Here is the result:

[Title "StockFish vs. StockFish"]
[FEN "2r2qk1/1b1n1pb1/p4n1p/1p4p1/4P3/1BN2P2/PPP3PP/R2Q1RK1 w - - 0 17"]

1. Qd2 Nc5 2. Ne2 Nh5 3. g4 Rd8 4. Qa5 Nf6
5. Ng3 Rc8 6. Rad1 Ne8 7. c3 Nd6 8. Rfe1 Be5
9. Qb6 Kh7 10. Re2 Rc6 11. Qa5 Qe7 12. Bd5 Rc8
13. Kg2 Bxd5 14. Rxd5 Nc4 15. Qb4 Qa7 16. Nh5 Qb6
17. a4 Nxa4 18. Qe7 Qe6 19. Rd7 Kg6 20. Qxe6+ fxe6
21. Re7 Rc6 22. b3 Nxc3 23. Rf2 Ne3+ 24. Kh3 a5
25. Rd2 Ncd1 26. Rb7 Rd6 27. Rg7+ Bxg7 28. Rxd6 Nf2+
29. Kg3 Be5+ 30. Kxf2 Nxg4+ 31. fxg4 Bxd6 32. Ke3 Bxh2
33. Kd4 Bg1+ 34. Kd3 Kf7 35. e5 Ke7 36. Ke4 Kd7
37. Ng3 a4 38. Ne2 axb3 39. Kd3 Bh2 40. Kc3 Bxe5+
41. Kxb3 Kc6 42. Kb4 Bd6+ 43. Kb3 Kc5 44. Nc3 Be5
45. Ne4+ Kd5 46. Nd2 Bf4 47. Nf3 Ke4 48. Ne1 e5
49. Kc2 Bg3 50. Nd3 Kf3 51. Nc5 e4 52. Nb3 Kxg4
53. Nc5 e3 54. Kd3 Kf3 55. Nb3 Be5 56. Kc2 e2
57. Nd2+ Ke3 58. Kb3 Kxd2 59. Kb4 e1=Q 60. Kxb5 Qc1
61. Kb4 Qc6 62. Kb3 Qb5+ 63. Ka3 g4 64. Ka2 Qb2#

As you see, white is basically crippled after losing their bishop. Under different circumstances the result might have been different. But I think your engine's evaluation seem to be spot on.

1

To add to the other answers, Black's pieces are more active on the board than White's. They control more squares and offer more options. White's rooks are a bit cramped. Although I'm not certain about it, I believe Stockfish takes this into account in its evaluation.

0

I question whether white is actually ahead on material. The numbers representing the "values" of the pieces are just approximations based on the relative strengths of the pieces, and cannot possibly apply in all cases. For example, winning two minor pieces for a rook and a pawn is usually considered a material advantage, despite both trades ostensibly having a value of 6. The "penalty for having two rooks and a queen" mentioned in @DM's answer is almost certainly how Stockfish accounts for this.

In this case, white traded away two minor pieces for a rook and two pawns. Despite this being 6 vs. 7, I would consider this trade to be approximately equal - maybe favoring white a little, but definitely not enough to claim a definitive material advantage.

In that case, it just comes down to which side has the better position. I think black's more active pieces give him the initiative, which is enough to explain the slightly better engine-score.

0

Well Stockfish 10+ at depth 20 evaluates it to −0.7. I do not understand the other answer that says "White seems to be getting a penalty for having two rooks and a queen" because the static analysis of that position seems to have largest contribution from king-danger and bishop-pair. Also note that the static evaluation is slightly misleading because the "king-danger" is not quite real but one needs to look a few moves ahead to check that there is really not much danger.

In fact, following what appears to be one of the best lines leads to:

[FEN "2r2qk1/1b1n1pb1/p4n1p/1p4p1/4P3/1BN2P2/PPP3PP/R2Q1RK1 w - - 0 0"]

1. Qd2 Nc5 2. Rad1 Nh5 3. g3 Nf6

And at this point there is no more king-danger. There is still the bishop-pair, but the overall static analysis of this position is actually close to zero, due to other factors. So why does SF think it is still ≈−1? Because that is what happens down the road:

[FEN "2r2qk1/1b1n1pb1/p4n1p/1p4p1/4P3/1BN2P2/PPP3PP/R2Q1RK1 w - - 0 0"]

1. Qd2 Nc5 2. Rad1 Nh5 3. g3 Nf6 4. Qd6 Qe8 5. Rfe1 Nfd7 6. Nd5 Kh7 7. c3 (7. Qe7 Bxb2 8. Qxe8 Rxe8 9. Rf1 Rc8 10. Ne3) 7... Ne5 8. Kg2 Nxb3 9. axb3 Rc6 10. Qb4

The static analysis of the final position in each of the two lines shown above are −0.88 and −1.03 respectively. But humanly you can see that White has either lost a pawn or been forced into a cramped position (and is likely to lose a pawn later anyway).

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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