Queen or pawn capture with check: my game 1130

Queen or pawn capture with check was the decisive comparison in my game 1130 against leon751: Stockfish preferred taking with the c-pawn on move thirty, but I used my queen. The saved evaluation swung from +6.13 to -2.78, and I eventually lost.

The recorded game

I played this game on 2024-11-18 as White, using the Chess.com account nacsol3. The record lists leon751 (1201) as my opponent and 1142 as my rating. This was rapid chess: the time control was 600 seconds, ten minutes without increment. The game belongs to the rapid sequence used for the From 631 to 1385 journey.

Queen or pawn capture with check

Both capture candidates at move thirty go to b6 and both give check. That makes this position a more specific lesson than simply remembering to look for checks. I did choose a check. The failure was in selecting which piece should make the capture and calculating what the resulting arrangement allowed.

My opponent was rated 1201 in the metadata, against my 1142. I played White in a ten-minute game, and the saved analysis gives my play 60.5 percent accuracy. Eleven blunders do not reduce the game to one move, but the move-thirty swing supplies a particularly clear exercise. I want to compare two legal candidates that share a target square and still produce very different assessments.

Position 1: the choice at move 17

My actual move was 17.Bh6. The saved Stockfish assessment changed from -0.28 before it to -5.56 afterward. These values use White's perspective. The analysis classifies this position as middlegame and records 388 seconds on my clock after the move; that reading cannot tell me what I was thinking.

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

The verified move descriptions are concrete: Bishop from e3 to h6; attacks the rook (g7), and behind it the rook (f8); double attack: it threatens two things at once; but the bishop can now be captured; the preferred alternative is Pawn from f2 to f4. The stored engine branch is 17. f4 Nxd4. I read that branch as a calculation to compare with my decision, not as a quotation of the actual continuation. The original game link above is the source for the moves that were played.

An earlier bishop move needs its own explanation

Before reaching the main comparison, I look back at move seventeen. I moved my bishop to h6 while the engine preferred pushing the f-pawn. The evaluation after my move was strongly favorable to Black. That is useful context: the later positive evaluation was a recovered opportunity, not an advantage that survived intact from the opening.

I should resist turning that history into an invented emotional story about a comeback. The record shows a sequence of moves and evaluations. It does not tell me whether I believed the position had improved. What I can learn now is that I need to reassess after each reply. A previous mistake does not determine the value of every later position.

Position 2: the choice at move 30

My actual move was 30.Qxb6+. The saved Stockfish assessment changed from +6.13 before it to -2.78 afterward. These values use White's perspective. The analysis classifies this position as middlegame and records 227 seconds on my clock after the move; that reading cannot tell me what I was thinking.

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

The verified move descriptions are concrete: Queen takes the pawn (b6), one of the pawns protecting the king; check; attacks the knight (c6); attacks the bishop (b7), and behind it the rook (b8); double attack: it threatens two things at once; the preferred alternative is Pawn takes the pawn (b6), one of the pawns protecting the king; check. The stored engine branch is 30. cxb6+ Kc8 31. Qf8+ Nd8 32. Ra7. I read that branch as a calculation to compare with my decision, not as a quotation of the actual continuation. The original game link above is the source for the moves that were played.

Calculating beyond the first check

For the preferred pawn capture, the stored best line continues with Black moving the king, my queen giving another check, Black interposing a knight and my rook moving to the seventh rank. Those moves are an engine branch. They were not the continuation of the game after my queen capture, and I will not splice them into the score.

The value of that branch is that it forces me to look beyond the shared first destination. After the pawn captures, the queen remains on its old square for a different job. After the queen captures, the board has a different arrangement. I can inspect those positions directly rather than relying on a broad rule that a cheaper piece should always capture first. Such a rule would be too crude.

The narrow lesson I want to keep

A forcing move limits the opponent's legal answers, but it still needs calculation. This game gives me a good test of that distinction because I found a checking capture and still missed the stronger one. When I replay it, I would name both candidates before asking the engine to explain the reply. I would then compare the position after each capture.

I do not claim that this review changed my rating or cured a habit. Those would need later evidence. The supported lesson is smaller: game 1130 contains two checks to the same square, one preferred pawn capture and one played queen capture, followed by a large evaluation swing. That is enough to build a useful review around without inventing a wider story of progress.

Where this sits in the journey

The adjacent rapid records put game 1129 at 1149, this game at 1142, and game 1131 at 1135. Those are recorded ratings, not a forecast or evidence that one review caused improvement. The weekly diary gives the surrounding games and results.

For a broader method, I use the beginner game-review guide and the guide to reading Stockfish analysis. They explain how to move from a numerical swing to a question I can answer at the board. I can also review a finished game through Chess Coach in Claude or ChatGPT, keeping the actual score and suggested alternatives visible together.