Game #465: London System vs Stonewall Dutch, and the capture on b5
On May 20, 2024, in game #465 of my rapid journey, I played the London System against DewaCatur_19 (1274), who set up a Stonewall Dutch: pawns on c6, d5, e6 and f5. On move 10 I took a pawn on b5 with my bishop, Black took back with another pawn, and Stockfish's evaluation fell from +0.65 to -6.08. I fought back to a winning endgame, missed the same rook capture three times in four moves, and then left the game in a lost position on move 49.
The series picked this game as a lesson for this opening, and it is one. Not because of deep theory, but because the decisive mistake came from a simple question I didn't ask: who else is guarding that pawn?
Where I was
Game #465 was the first game of week 20 of the journey, a Monday morning. I had lost my last game of the previous week to bshtylla (1254), and this was the second opponent in a row rated over 1250. I came in at 1038 and left at 1035, a small loss because of the rating gap.
| Game | #465, May 20, 2024 |
| Opponent | DewaCatur_19 (1274) |
| My colour, rating after | White, 1035 (from 1038) |
| Time control | 20 minutes, no increment |
| Opening | Queen's Pawn Opening, Horwitz Defense (A40), by Chess.com's label |
| Result | Loss, I abandoned the game after Black's 49th move |
| Accuracy (Stockfish) | Me 53.3, Black 66.2 |
| My blunders / mistakes / inaccuracies | 7 / 2 / 0 |
Black's game wasn't clean either: seven blunders, one mistake and two inaccuracies. It was that kind of game.
The opening: a queen on a4 and a pawn on b5
The moves were 1.d4 f5 2.Bf4 Nf6 3.Nf3 e6 4.e3 Be7 5.c3 d5 6.Ne5 O-O 7.Nd2 a6. Black's first move, f5, is the Dutch Defense, and with pawns on d5, e6 and f5, plus c6 a move later, Black had built the Stonewall. My setup was the London as usual: bishop on f4, pawns on d4, e3 and c3, and a knight jumping to e5.
Then I got ambitious on the queenside. 8.Qa4 wasn't the engine's choice: the game analysis wanted the bishop to e2, and a recheck for this post listed g4, bishop to d3 and knight to f3 as its top three. Black played 8...c6, and I pushed 9.b4. Again, not the engine's move: it preferred the queen back to c2 or quiet moves like bishop to e2 and h3. Black answered 9...b5, attacking my queen.
I took on b5 with the bishop. The engine called it a blunder: +0.65 to -6.08 (evaluations are from White's side, so minus means better for Black). The reason is on the board. The pawn on b5 had two defenders, the pawns on a6 and c6. Black took back with the c-pawn, and the new pawn on b5 attacked my queen again. I had given a bishop for one pawn and still had to move the queen.
The game analysis picked queen to b3 as the best move. When I checked the position again, Stockfish put three queen retreats within a tenth of a pawn of each other: c2, d1 and b3, all "about equal" or "slightly better for white". Any of them was fine. The only bad option was the one I chose.
Moment 1: the fightback
I was down a piece for a pawn, but Black gave some of it back. On move 14 Black put the bishop on d7, which the engine marked as a blunder (-4.57 to -1.74), and my knight took it. The engine agreed with my capture: its description says the knight took the bishop and attacked both knights and the rook on f8. After 15...Nfxd7 16.Bc7 Qc8, the evaluation was still around -1.5, but the game was alive.
I pushed the b-pawn. The engine wanted queen takes e6 with check, a move that also attacked the undefended bishop on e7 and both knights. My pawn push dropped the evaluation from -1.46 to -5.42.
Then Black blundered back. On move 17 the rook went to a7, and suddenly I was better: the engine had me at +2.77.
I took a pawn on a6 with my rook. Black's rook took my bishop on c7, and the evaluation went from +2.77 to -7.09. The engine's move was the same as the move before: queen takes e6, check. I passed on it twice in a row. I did play it on move 19, one move too late.
Moment 2: the knight on b6
The middlegame went on. On move 21 my queen took a rook on f8 and Black's queen took it back, and by move 33 the position was a strange endgame: my rook, knight and four pawns against Black's queen, knight and two pawns. On move 33 Black's king came to f6, and the engine called it a blunder: I was winning, +2.89.
The winning move for me was the same for the next four moves: rook takes the knight on b6. I played 34.Kg2, then 35.f5+ and 36.Kg3. Black put the queen on a7, and here's the last chance.
The knight on b6 looked protected by the queen on a7, and I think that's why I left it alone. But it wasn't really protected. I asked the engine to play rook takes, queen takes back, and the reply was knight to d7 with check: a knight fork on the king and the queen, scored at +11.10. Instead I played the king to h4, and the evaluation flipped from +5.23 to -5.25. The same capture was the engine's move on moves 34 and 36 too, and both of those king moves were also logged as blunders.
Black blundered once more with the knight to c8, I missed that chance with my rook to b7, and from there Black's d-pawn ran down the board. After 49...Kxf7 I had a rook and a pawn against a queen and a pawn on d2, and I left the game with eight and a half minutes on my clock. I'm not proud of that. Resigning takes one click; abandoning just makes my opponent wait.
What the engine says about game #465
My accuracy by phase was 71.3 in the opening, 55.4 in the middlegame and 48.5 in the endgame. Nine of my moves were mistakes or blunders: one in the opening, two in the middlegame and six in the endgame. The clock had nothing to do with it. At the move-37 blunder I still had more than ten minutes.
The lesson I took from game #465
For the London System vs Stonewall Dutch, I took two things away:
- Don't send the queen out early to the queenside. Queen to a4 and pawn to b4 weren't the engine's moves, and they gave Black a pawn push (...b5) that attacked my queen with tempo. Quiet development, like bishop to e2 or d3, was what Stockfish wanted.
- Before taking a pawn that attacks your queen, count its defenders. The pawn on b5 had two. Taking it with a bishop cost me the bishop.
And one lesson for the whole game, not just the opening: "protected" is a question, not an answer. The knight on b6 was protected by the queen, but recapturing would have walked the queen into a knight fork. I now try to play the capture out in my head one more move before deciding a piece is safe.
What came next
That same afternoon I beat bobbo6767 (1286) in twelve moves, my biggest upset. The next time I faced this opening, game #475, I lost again, to KriOnBoard2016 (1060); the time after that, game #494, I beat Mick210869 (959). The full week is week 20, and the series lives on the journey page. You can replay game #465 on Chess.com.
If the difference between an inaccuracy, a mistake and a blunder in these numbers isn't clear, here's how the three labels work. And if you want your own games read the same way, with the missed capture shown on a board, that's how Chess Coach's analysis works.
Frequently asked questions
What went wrong in the opening of this London System vs Stonewall Dutch game?
My queen went to a4 and my pawn to b4, and when Black pushed ...b5 to attack the queen I took that pawn with my bishop. Black had a second pawn guarding b5, took back, and the new pawn attacked my queen again. Stockfish scored it +0.65 before and -6.08 after.
What is the Stonewall setup?
It's a pawn shape for Black with pawns on c6, d5, e6 and f5. In game #465 Black reached it after 1.d4 f5, the Dutch Defense, though Chess.com labelled the game Horwitz Defense.
Was game #465 lost after the opening?
No. Black gave material back twice, and in the endgame the engine had me winning on moves 34, 36 and 37, each time by taking a knight on b6 with my rook. I didn't take it, and later I abandoned a lost position on move 49.