Week 121: how to get better at chess endgames, starting with the one I threw away
Two games in the week of June 3 to 7, 2026, one loss and one win, and the rating didn't move: 1231 → 1231. The average accuracy was a good 87.2%, but that number hides the most painful game of the month. In game 2103 I was winning an endgame by more than five pawns' worth, and I lost it. If you're looking for how to get better at chess endgames, my honest starting point is this game: the opening and middlegame were fine, and almost everything went wrong after the queens came off.
The week in numbers
| Games | 2 |
| Wins / losses / draws | 1 / 1 / 0 |
| Rating | 1231 → 1231 (0) |
| Average accuracy | 87.2% |
| My blunders | 3, all in one endgame |
| Colour | 1 game with White (loss), 1 with Black (win) |
| Openings | Accelerated London System, Van't Kruijs Opening |
Game 2103: the passed pawn I let go
White against brmvkrishna (1237), an Accelerated London. Things went well early. On move 10 Black's rook took my bishop on h5 and my queen took the rook back with check, so I came out with a rook and a pawn for my bishop, and the queens were traded two moves later. By move 27 the engine had me at +5.33, winning.
The engine split my accuracy by phase, and it tells the story on its own: 95.7% in the opening, 93.3% in the middlegame, and 66.2% in the endgame. All three of my blunders came in the endgame. So did the turning point.
Black had just pushed the a-pawn from a7 to a5. That pawn is now a passed pawn: no white pawn stands in front of it or on the next file to stop it, so only my pieces can. My knight on c6 could take it right away, and the engine kept me winning after that (+5.03 in its line). I played king to f1. The pawn advanced to a4, and the engine's evaluation after my king move was 0.00: its line is just a series of knight checks.
What makes this hurt is the clock: I had 4 minutes and 47 seconds left when I played king to f1. It wasn't time trouble: the pawn had moved on the previous turn, and I played around it. Two moves later Black gave me another chance, and I took a bishop on a6 with my pawn when a knight check was better. On move 35 a pawn move on the king's side turned +1.77 into −4.91, and I lost on move 45 (game 2103).
Game 2104: a clean win, and the engine's trade
Black against YN589 (1235), who opened with 1.e3, the Van't Kruijs Opening. This was the opposite kind of game: 96.3% accuracy, no blunders, and a win on move 32. On move 23 I was already winning (−4.64), but White's rook had reached my second rank on c7, where it was pinning my bishop on d7 to my king on h7.
Both moves keep Black winning. I attacked the rook with bishop to d8. The engine slightly preferred bishop to e7, which attacks White's bishop on a3, which has no defender, and offers a trade: in its line the bishops come off and Black stays winning at −4.76. Trading pieces when ahead takes away the other side's chances. In the game it didn't matter: White answered with a knight move, left the rook on c7 for my bishop, and the game ended on move 32 (game 2104).
The lesson of the week
In an endgame, a passed pawn comes first. Before any king move or quiet improvement, I look at every passed pawn on the board, mine and theirs, and ask whether I can take it, block it or push it. In game 2103 Black's pawn had moved on the previous turn, right in front of my knight, and I still played a king move. My endgame accuracy that day, 66.2%, against 95.7% in the opening, is a clear sign of where my training time should go.
All games this week
| Game # | Date | Colour | Opponent (rating) | Result | Opening | Accuracy | Link |
|---|---|---|---|---|---|---|---|
| 2103 | Jun 3 | White | brmvkrishna (1237) | Loss | Queen's Pawn Opening, Accelerated London System | 78.0% | game |
| 2104 | Jun 7 | Black | YN589 (1235) | Win | Van't Kruijs Opening | 96.3% | game |
Go back to week 120, forward to week 122, or see every week on the journey hub. If you have an endgame like this in your own history, Chess Coach can review it with Stockfish.
Frequently asked questions
How do I get better at chess endgames?
Start with your own lost endgames. In my game 2103, my accuracy was 95.7% in the opening and 93.3% in the middlegame but 66.2% in the endgame, and all three of my blunders came there. The first one was ignoring a pawn that had just become passed.
What is a passed pawn?
A pawn with no enemy pawns in front of it on its own file or on the files next to it, so only pieces can stop it. In game 2103 Black's a-pawn became one on move 27, and the engine's best move was to take it at once with my knight.
Should you trade pieces when you're ahead?
Usually that's the engine's advice. In game 2104, where I was winning, Stockfish preferred a bishop move that offered a trade over my bishop move that attacked a rook.