Week 76: how to get better at chess endgames, one clean and one thrown away
Week 76 was short and good: 4 rapid games between July 8 and 11, 2025, 3 wins and 1 loss, rating 1248 → 1264 (+16), and an average accuracy of 91.0%, the best of any week in this part of the journey. It also gave me a neat pair of examples for how to get better at chess endgames: one endgame I played almost perfectly, and one rook endgame, two pawns up, that I let slip in two moves.
The week in numbers
| Games | 4 (3 White, 1 Black) |
| Record | 3 W · 1 L · 0 D |
| Rating | 1248 → 1264 (+16) |
| Average accuracy | 91.0% (my opponents: 86.5%) |
| Blunders | 6 (my opponents: 10) |
| Openings | London System, Nimzowitsch Williams Variation, Zukertort Variation, Old Benoni (1 each) |
Two of the four games had zero blunders: game 1609 against ahmadmaniee_40 (1228), 96.7%, and game 1610 against tbm1806 (1232), 96.8%, which I won on time. Five of my six blunders came in one game, the chaotic London against GMLashio, which I still won.
The clean one
Game 1609, Zukertort Variation, White (game). The engine's phase summary is the nicest I've had in a while: 95.9% in the opening, 94.9% in the middlegame, and 97.6% over 27 endgame moves, with no mistakes and no inaccuracies. It finished with my rook, bishop and five pawns against Black's lone king, and my opponent resigned on move 47. Nothing spectacular happened, and that's the point. An endgame without errors is just one move after another that doesn't give anything back.
The one I let go
Game 1608 against miszczers (1250), Black, my only Black game and my only loss (game). After 24...Rxa2 I had seven pawns against five in a rook endgame, and the engine had me at −5.14: winning.
I pushed my b-pawn, thinking about my extra queenside pawns. But White's king was already on f4, one step from my kingside, and after b5 it could go to g5 and start eating pawns. h6 stops that: a pawn on h6 covers g5, so the king can't use it. With h6 the engine still has me winning (−5.05). After b5 it dropped to −1.37, and after my next pawn move, 27...b4+, to +0.21. White's king walked to f6, right next to my f7 pawn, and at that point the record says I abandoned the game.
Being two pawns up wasn't the problem. Leaving the enemy king free to walk in was.
Moment 2: the rook sacrifice I didn't play
Game 1607 against GMLashio (1224), White, London System (game). A wild middlegame: Black's queen had reached a2, on my second rank, with my king still on e2, and Black's last move, 19...f6, attacked my knight on e5.
Taking the bishop was the safe-looking move, and it gave the advantage away: +3.12 to −0.15. Rxh6 goes straight at Black's king along the open h-file. The rook can be captured, but the engine's line shows the attack still works, and it rates the position as clearly better for White. I won anyway after 117 half-moves, when Black resigned facing my pawn about to promote on f8.
The lesson of the week
Endgames reward patience more than ideas. In game 1609 I didn't do anything clever, I just didn't make a mistake for 27 moves. In game 1608 I had a clear plan, push my extra pawns, and forgot what the engine keeps showing me in endgames: deal with the opponent's king before you run with your own pawns. A pawn move like h6, which takes a square away from the enemy king, often matters more than a pawn move that goes forward.
All games this week
| Game # | Date | Colour | Opponent (rating) | Result | Opening | Accuracy | Link |
|---|---|---|---|---|---|---|---|
| 1607 | Jul 8 | White | GMLashio (1224) | Win | London System | 83.8% | game |
| 1608 | Jul 8 | Black | miszczers (1250) | Loss | Nimzowitsch Defense Declined Williams Variation | 86.9% | game |
| 1609 | Jul 10 | White | ahmadmaniee_40 (1228) | Win | Queen's Pawn Opening Zukertort Variation | 96.7% | game |
| 1610 | Jul 11 | White | tbm1806 (1232) | Win | Old Benoni Defense | 96.8% | game |
Back: week 75 · Forward: week 77 · Journey hub · See your own endgames phase by phase with AI chess analysis from Chess Coach.
Frequently asked questions
How do you get better at chess endgames?
My week 76 suggests starting with the opponent's king. In my lost rook ending the mistake was a pawn push that let White's king walk in; the engine's choice, h6, simply took a square away from it.
Is being two pawns up in a rook endgame always winning?
Not automatically. In game 1608 I was two pawns up and the engine had me at −5.14. Two pawn moves later it was about equal, +0.21.
What does a clean endgame look like at 1250?
My game 1609: 27 endgame moves at 97.6% accuracy with no mistakes, until my opponent resigned against my rook, bishop and five pawns.