Don’t believe your eyes…believe your engine!

October 30, 2021 Matthew Sadler No comments exist

One of the things I like when watching engine games are evaluations that surprise me. In particular I look out for positions that seem reasonable to human eyes but which produce a high engine evaluation. It’s struck me how often I am misled by plausible-looking development schemes.

Position after 8…Bd6 (game)

In the current position, Black has placed its light-squared bishop outside the pawn chain and White’s light-squared bishop has been developed to the modest e2 square. These signals of good development practice by Black and unambitious development from White convince me intuitively that Black must have a reasonable position.

What you don’t notice immediately is Black’s lag in development as a result of playing …Nf6xd5-f6. Indeed, Black has not yet castled, its queenside is not yet developed and it’s White’s move. This is of course taken into account by the engines and leads to a +0.49 evaluation from Leela (by Leela’s standards, a clear White advantage) and an equally bullish +1.12 (the equivalent of a pawn advantage after just 8 moves!) from Stockfish.
The dynamic initiative that the engines build up in the next few moves is based on the following factors:

  1. The undefended b7-pawn that cannot be protected by a rook while Black’s queen’s knight is undeveloped
  2. The exposed bishop on f5. We notice that the White bishop’s modest placement on e2 prevents …Bg4 in reply to Nh4. In other words, White is picking up the bishop pair by force.
  3. c5 gaining a tempo on the dark-squared bishop
  4. White’s general space advantage and the lack of a challenge in the centre from Black in the next few moves.

The power of White’s initiative took me by surprise!

9.Nh4 Bg6 10.Qb3

Position after 10.Qb3 (game)

In consecutive moves, White has hit both exposed points in Black’s position: the bishop on f5 and the b7–pawn. Black has a serious choice to make: offer the exchange of queens with …Qb6 or play as conservatively as possible with …Qc7.


Defending b7 and gaining a tempo by hitting the h2–pawn.

10…Qb6 11.Be3 Qxb3 12.axb3

Position after 12.axb3 (analysis)

is again deceptively difficult for Black. Black is not developed enough to prevent White from achieving c5 followed by b4–b5 with strong queenside pressure (Black would have needed to be castled already to prevent this). It’s another example of dynamic factors in the position that don’t jump out at you at a casual glance. 12…Be4 13.Nf3 a6 14.Nd2 Bg6 15.c5 Bc7 16.b4 0–0 If Black had one more development move (…Nbd7) then its position would start to approach respectability. Without this tempo however, White is able to liquidate the doubled b-pawns and expose the b7 pawn along the h1–a8 diagonal. 17.b5 cxb5 18.Nxb5 Nd5 19.Nxc7 Nxc7 20.Nc4 Nc6 21.Nd6 Rab8 22.Rfd1 Bc2 23.Rd2 Bb3 24.Ra3 Bd5 25.Bf4 Ra8 26.Rg3 g6 27.h4 with an initiative on both sides of the board. Even Stockfish couldn’t hold this 1–0 (109) Komodo Dragon – Stockfish Matthew Engine Matches 2021

White replied to 10…Qc7 with 11.f4

Position after 11.f4 (game)

A really ambitious move from White, dealing actively with Black’s threat against h2 by clamping down on the …e6–e5 break. Less obviously, White is also preparing for a huge expansion on the kingside!

11.g3 0–0 12.c5 Be7 13.Nxg6 hxg6 14.Bf4 Qc8 15.h4 was an alternative line from Komodo Dragon which is also very strong. White grabs space all over the board with gain of tempo, netting the 2 bishops along the way. The negative aspect to this plan is that Black gains access to the d5–square. However White’s plan is to exchange off a black knight as soon as it occupies d5, destroying the outpost by forcing a black pawn to occupy the d5–square. After that White still has a space advantage on the kingside and queenside to play with.


Ethereal lets Igel carry out its plan…which turns out to be fearsome! Let’s look at some alternatives

i) 11…Nbd7 delays kingside castling to dissuade White from pushing pawns there but White has easy and powerful play: 12.c5 Be7 13.f5 Bxf5 14.Nxf5 exf5 15.Bf4 Qc8 16.Bc4 0–0 17.Rae1 forced Stockfish to give up a piece with 17…Nxc5 18.dxc5 Bxc5+ 19.Kh1 1–0 (45) Komodo Dragon – Stockfish Matthew Engine Games 2021


Position after 11…Bf5 (analysis)

was the favorite choice of my engines, and is a known device in the Caro-Kann: Black offers an exchange on f5 in order to block the advance of the f-pawn and hold back the White g-pawn. However, this costs Black another development tempo without solving the problem of Black’s king. 12.h3 Threatening g4, driving away the bishop from f5 12…h5 Holding back g4, but creating a weakness on the kingside. 13.Be3 0–0 14.Qd1 A great move, hitting the h5–pawn and also supporting g4. All my engines opted to jettison the h5 pawn to win time to complete Black’s development. 14…Qe7 15.Bxh5 Bh7 16.Bf3 Nbd7 17.Qd2 Rad8 Black has reached a fairly normal state of development… just a pawn down! However, it’s a rook’s pawn and our experiences of AlphaZero’s games have taught us that the rook’s pawn is the least damaging of all the pawns to lose! Black still has some fighting chances

Back to the game, Ethereal has just played 11…0-0


Position after 12.g4 (game)

Powerful play from Igel expanding its existing space advantage, exploiting the fact that Black is unable to challenge the white centre effectively. The modest bishop on e2 is again proving its worth by supporting the g4–pawn. Black must be very careful as its light-squared bishop is now in danger of being trapped.]


12…Na6 13.c5 Be7 14.g5 Ne8 15.Be3 Qc8 16.a3 Nac7 17.Nxg6 hxg6 18.h4 b6 19.Kg2 Rb8 20.Qc2 a5 21.b4 axb4 22.axb4 bxc5 23.bxc5 Rb4 24.Rh1 was an easy win in Stockfish – Komodo Dragon Matthew Engine Games 2021, 1–0 (44). h5 will open the h-file with mate to follow.

13.Kh1 Na6 14.f5 exf5 15.gxf5

Opening the g-file against the Black king. Notice how little pressure Black is able to exert on the white position.

15…Bh5 16.Bxh5 Nxh5 17.Qd1

Position after 17.Qd1 (game)


17…Bxh4 18.Qxh5 Bf6 19.Bf4 Qd7 20.Be5 is also very powerful

18.Rg1 Kh8 19.Qf1 Qd7 20.Qh3 Nc7 21.Nf3 We see both Igel and Stockfish redeploying the knight from h4 to f3, aiming for e5 or g5.

21…Nce8 22.Be3 Rd8 23.Rg5 Qc7 24.Rag1 Bd6 25.Qh4 Qd7 26.d5

Position after 26.d5 (game)

Freeing d4 for the dark-squared to combine in the attack against f6 and g7.

26…Rg8 27.c5 Qe7 28.Qf2 Bc7 29.d6

Carnage! Ethereal has to give up a piece and the result is never in doubt!

29…Nxd6 30.cxd6 Rxd6 31.Re1 Rd3 32.Bd4 Qd8 33.Bxf6 gxf6 34.Rxg8+ Qxg8 35.Qh4 Qd8 36.Ne4 Rd1 37.Nxf6 Rxe1+ 38.Nxe1 h5 39.Qxh5+ Kg7 40.Qg5+ Kf8 41.Nh7+ Ke8 42.Qg8+ Ke7 43.f6+ Ke6 44.Qg4+ Kd5 45.Qh5+ Ke6 46.Qh3+ Ke5 47.Nf3+ Kf4 48.Qg3+ Ke3 49.Ne5+ Kd2 50.Nxf7 Qd5+ 51.Qg2+ Kc1 52.Qxd5 cxd5 53.Nfg5 b5 54.f7 Bd6 55.f8Q Bxf8 56.Nxf8 Kc2 57.h4 d4 58.h5 d3 59.Ne4 d2 60.Nxd2 Kxd2 61.h6 Kd3 62.h7 Ke4 63.h8Q Kd5 64.Qf6 b4 65.Ng6 b3 66.Nf4+ Ke4 67.Kg2 bxa2 68.Kg3 Ke3 69.Qd6 Ke4 70.Qd5+ Ke3 71.Qd3# 1–0

Leave a Reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.