Game #70: the King's Gambit Declined Classical Variation, and one unguarded pawn

On January 22, 2024 I beat eejack, rated 1097, with Black in a King's Gambit Declined Classical Variation. I went into the game at 886 and came out at 898. The engine is clear about how it happened: I made two mistakes in the opening and was slightly worse, then my opponent made one blunder on move 13, my queen took a pawn with check and won a knight, and twenty moves in, White resigned.

Measured from my rating before the game, 211 points was the biggest gap I had beaten up to that point in the series (game #46 against a 1055 had been 184). It was not the best game I played that week. It is, though, a very clear example of something I needed to learn: before you move a piece, check what it was guarding.

Where I was that week

This was the week covered in the week 3 diary, a rough one: 43 games, 15 wins, 27 losses and 1 draw, and my rating went from 899 to 872. Most of that diary is about the London System with White. Game #70 is one of the better moments with Black.

The morning before it had not gone well. Game #67 was a loss, game #68 a draw against kerflup (890) and game #69 a loss against shivram007 (950) that left me at 886. Game #70 was played in the afternoon, at 10|5 rapid, and lasted about ten minutes.

How the opening got its name

The game started 1.e4 e5 2.Nc3 Bc5 3.f4 d6. White pushed the f-pawn two squares on move 3, offering it to my e5 pawn. That is the idea of the King's Gambit: a gambit is an opening where you offer material, usually a pawn, for quicker development or an attack. I didn't take it. I put a pawn on d6 to support e5, with my bishop already on c5, and that combination is why the database calls it the King's Gambit Declined Classical Variation, even though White reached it through 2.Nc3 rather than the usual 2.f4.

I didn't know any of that during the game. I just didn't want to open lines toward my own king.

Mistake 1: a pawn on f6 instead of a knight

After 4.Nf3 the position was about equal (−0.18, very slightly better for me; negative numbers mean Black is better). I played 4...f6, pushing a pawn to protect e5 a second time. The engine marked it as a mistake and its turning point of the game: the evaluation went to +1.24, slightly better for White.

r n b q k . n r
p p p . . p p p
. . . p . . . .
. . b . p . . .
. . . . P P . .
. . N . . N . .
P P P P . . P P
R . B Q K B . R

The engine's choice was 4...Nf6, developing the knight to its natural square, with a line like 5.Bc4 O-O 6.d3, about equal. My f6 did two things I didn't think about: it took the f6 square away from my own knight, and it moved a pawn that stood next to my king. White answered 5.f5, gaining space, and my knight had to go the long way round, to e7.

Mistake 2: chasing a pawn instead of castling

Over the next moves, pieces were traded on c6 and White played a mistake of their own, 10.Bxc6+, which brought the game back to about equal. After 12.Qd2 the position was −0.17, and both kings were still in the centre. Then I jumped my knight to d4.

r . . . k . . r
p p p . q . p p
. . n p . p . .
. . b . . P . .
. . . . P B . .
. . N . . . . .
P P P Q . . P P
R . . . K . . R

My knight attacked the pawn on f5, which had no defender. It looked active. The engine called it a mistake (+1.14 for White) and wanted me to castle queenside instead, with a line like 12...O-O-O 13.O-O-O Bb4 14.Rhe1, about equal. My king never did castle in this game: two moves later it walked to d7. I had 8 minutes left when I played Nd4, so this was not a clock problem. It was me looking at what I could take before looking after my king.

The blunder: the knight that was guarding e4

Here White had a good position. The engine's suggestion was simply to castle queenside, 13.O-O-O, slightly better for White. Instead my opponent played 13.Nd5, moving the knight from c3 to d5, where it attacked my queen on e7.

r . . . k . . r
p p p . q . p p
. . . p . p . .
. . b . . P . .
. . . n P B . .
. . N . . . . .
P P P Q . . P P
R . . . K . . R

The knight on c3 was the only piece defending the pawn on e4. Once it left, that pawn was loose, and so was the knight itself on d5. White's winning chance went from 60.3% to 8.6% with that one move.

The move I got right: queen takes with check

This is the one moment the review marks as good for me. I took on e4 with my queen, and it was also the engine's first choice.

r . . . k . . r
p p p . q . p p
. . . p . p . .
. . b N . P . .
. . . n P B . .
. . . . . . . .
P P P Q . . P P
R . . . K . . R

The capture was a check, so White had to deal with the king first, and from e4 my queen also attacked the knight on d5, which had no defender, the bishop on f4 and the pawn on f5. That's a double attack: one move that threatens two or more things at once, so the opponent can't answer all of them. White played 14.Kd1 and I took the knight, 14...Qxd5. The engine had me completely winning (about −6.4).

The rest was about not giving it back. After 15.Re1+ Kd7 16.c3 my knight on d4 was attacked. I took another pawn, 16...Qxf5, and let the knight go: 17.cxd4 Bb6. The engine preferred bringing a rook to e8 first (16...Rae8), but both moves kept Black completely winning. After 18.g3 Rhe8 19.Rxe8 Rxe8 20.Rc1, White resigned. In the final position I was two pawns up, each side had a queen, a rook and a bishop, and White's king was still on d1.

What the engine says about the whole game

Split by phase, my 12 opening moves scored 86.6% accuracy with two mistakes and one inaccuracy (9...Bxc6, taking with the bishop where the engine wanted the knight). My 7 middlegame moves scored 97.3%, with no errors. So the shape of the game is: I played the opening worse than my opponent, and the middlegame better, and the middlegame started with a gift.

I didn't outplay a player rated 200 points higher. I stayed close enough, and when they left something unguarded, I saw it.

The lesson: what was that piece guarding?

The blunder in this game is a pattern worth recognising from both sides of the board. A piece attacks something, so it feels active, but it was also doing a quiet job where it stood, and that job is gone the moment it moves. White's knight on c3 was guarding e4. My own 12...Nd4 a move earlier was the same kind of thinking in a milder form: I went after a pawn and forgot about castling.

The check I take from this game, before any move that attacks something, has two questions:

1. What is this piece defending right now?
2. If it leaves, what can my opponent take, and does it come with check?

If eejack had asked the first question, the knight stays on c3. If I'd asked it about my king a move earlier, I castle.

What changed afterwards

I won the next two games as well, #71 against ConflictOfTriggers (1006) and #72 against coryjaydavis (897), which put me at 917. The rest of the week went badly, and I finished at 872, as the week 3 diary explains with its London System games. An upset like this one doesn't mean much on its own; the habit behind the one good move, checking what each piece is guarding, is the part worth keeping.

For a game where the win came even more cheaply, read game #46, a Pirc Defense as White that my 1055-rated opponent abandoned. The previous week is in the week 2 diary, and everything else is on the journey page. If you're wondering whether an assistant can do this kind of review for you, here's what Claude can and can't do with chess games.