Game #538: the king and pawn endgame mistakes that turned a win into checkmate
On June 12, 2024, in game #538 of my rapid journey, I had White in a London System against NgNguytt (957) and got to a king and pawn endgame (only kings and pawns left on the board) that Stockfish scored +7.83: three pawns against two, and close to a certain win. Then I made two king and pawn endgame mistakes in a row. The first handed Black the win, the second threw away the draw Black handed back, and the game ended with a new black queen checkmating my lone king on move 66.
My rating went from 949 to 941. The record for the game says my best winning chance was 94.7%.
Where I was that day
Game #538 sits in week 23 of the journey, the week I got back over 1000. That Wednesday had started well: four wins in a row, games #533 to #536, took me from 925 to 957. Then I lost game #537 to x-7078475191 (948) and dropped to 949, and twenty minutes later I sat down for this one.
The London System was my main weapon with White, and not a very successful one. Before this game, Chess.com had labelled 32 of my White games plain "London System": 12 wins, 18 losses and 2 draws.
| Game | #538, June 12, 2024 |
| Opponent | NgNguytt (957) |
| My colour, rating after | White, 941 (from 949) |
| Time control | 10 minutes, no increment |
| Opening | London System (D02) |
| Result | Loss, checkmated on move 66 |
| Accuracy (Stockfish) | Me 60.9, Black 79.6 |
| My blunders / mistakes / inaccuracies | 8 / 2 / 3 |
The phase numbers tell the story before any board does: 85.1 accuracy in the opening, 66.1 in the middlegame and 58.7 in the endgame. I got worse the closer I got to the finish.
Moment 1: the knight jump I didn't make
The opening went my way. On move 7 my opponent castled queenside, and the engine called it a blunder: +0.74 to +3.49. I put my bishop on b5, pinning the knight on c6 to the queen on d7 (a pin is when a piece can't move without exposing a more valuable piece behind it). Black played a knight to e4, and it was my move.
I castled. It's a natural move and it cost three and a half pawns of evaluation: +4.66 to +1.23. The knight jump to e5 attacked three things at once, and the pinned knight on c6 was one of them. I had plenty of time, more than nine minutes on my clock. I simply didn't look at the forcing move before the safe one.
Moment 2: the check I skipped in the middlegame
The middlegame was a long trade of blunders. Queens came off on move 12, my knight went hunting on the queenside and picked up pawns, and the evaluation bounced between +2 and +5 for ten moves. The engine marks my opponent's moves 13, 15, 18 and 19 as mistakes or blunders, and mine on moves 15, 16, 19 and 22 as blunders. This one is the clearest.
The knight could take the c7 pawn with check and then hit the d5 pawn too. I moved my rook instead, and +4.52 became +0.24, about equal. Two moves later I missed another one: my rook to b8 would have attacked the bishop on f8 with the rook on h8 behind it. I took a pawn on e5 instead.
Somehow I stayed ahead anyway. By move 32 we were in a knight endgame, one knight each, and I was slightly better.
Moment 3: the pawn endgame at +7.83
On move 38 Black played the knight to c4 and let me trade knights. The engine called it a blunder, +1.14 to +5.13, because the pawn endgame that followed was winning for me. Four moves later my king had taken a pawn on c4, Black's king had taken my pawn on e3, and this was the board.
My pawn on f4 was attacked, and my king on c4 was too far away to defend it. The engine's answer was to push it: f5. Its line pushes the pawn to f6 and then f7 while Black's king is busy with my other pawns, too late to stop it.
I played my king to d5. From d5 the king touches e4, e5 and e6, but not f4, so Black simply took the pawn. The evaluation went from +7.83 to -5.42 in one move. With the f-pawn gone, Black's king stood right next to my g- and h-pawns, and Black's own g- and h-pawns now had a clear road.
The clock is the part that hurts. I had 2:45 left; my opponent had 6:42. And according to the clock data, I played that king move in about one second. In a position with six pawns and two kings, where every move can be calculated, I didn't calculate.
Moment 4: the draw I didn't see either
Then my opponent gave me a second chance. On move 44 Black's king went to g4, and the engine dropped the evaluation all the way to 0.00. There was a draw on the board.
The engine's line: g6, Black takes with the h-pawn, my king goes to f7, Black's king takes my h-pawn, and my king takes on g6. Everything gets traded and nobody can win. I went straight for the g7 pawn with my king instead. Black pushed g6, my king took on h7 a couple of moves later, Black's king took both of my pawns, and the g-pawn walked down the board: g4, g3, g2 and a new queen on move 54.
I kept playing with a lone king for twelve more moves until the queen and king mated me on d7. I don't know why I didn't resign. It changed nothing, but it's in the record.
King and pawn endgame mistakes: what I learned
The two endgame moves have the same root. Both times I moved my king toward where I wanted to go, not where it was needed. On move 43 the job was to save the f-pawn or run with it. On move 45 the job was to trade off Black's last useful pawn. In a pawn endgame the pawns are the whole game, and the king's job is to serve them.
Count before you move. A pawn endgame is the one part of chess where a beginner can calculate almost everything. Who gets to promote first? Can the king catch the pawn? Every one of those questions has an answer you can count with your fingers. I had nearly three minutes and used one second.
Forcing moves first, even early. Moments 1 and 2 were the same mistake in a different shape. A knight jump that attacks the queen and a knight capture with check are the first moves to look at, before castling or tidying up a rook. I covered this pattern in game #509 too, where four of my missed wins were checks.
A big lead isn't a finished game. I was ahead for most of the 66 moves and still lost by mate. The engine's final numbers for me were 8 blunders, 2 mistakes and 3 inaccuracies; Black made 9 blunders. Two careless players, and the last blunder decided it.
What came next
The next morning I won game #539 against namespacetaken (924), lost the next one, and then won four in a row, games #541 to #544. The week finished at 1009, back over 1000. For another endgame from the series, a rook against a pawn, see game #284. The whole series is on the journey page.
If the words blunder, mistake and inaccuracy are new to you, here's what each one means. And if you want your own endgames explained in plain words like this, that's what Chess Coach does inside Claude: how the analysis works. You can replay game #538 on Chess.com.
Frequently asked questions
What was the king and pawn endgame mistake in game #538?
On move 43 my king stood on c4, Black's king on e3 was attacking my pawn on f4, and I moved my king to d5. That doesn't protect f4. Stockfish wanted the pawn pushed to f5, out of reach, and rated it +7.83. After my move it was -5.42, winning for Black.
Could you still have drawn the endgame in game #538?
Yes. Black made a mistake on move 44 and the engine scored the position 0.00. Pushing my pawn to g6 to swap off Black's h-pawn kept the draw. I walked my king to f7 instead, Black's g-pawn ran through and promoted, and I was checkmated on move 66.
How much time did you have when you made the mistake?
2 minutes 45 seconds in a 10-minute game, against 6 minutes 42 for my opponent. The clock says I played the king move in about one second.