Game #588: how to checkmate with king and rook, and how I didn't
On June 19, 2024, in game #588 of my rapid journey, I had White against Starfiszh (993) and reached the easiest winning ending there is: king and rook against a bare king. It took no skill to know I was winning. Knowing how to checkmate with king and rook, fast, was another matter. I made 21 moves with that rook, missed two mates in two, and the game was drawn by repetition with 0.7 seconds left on my clock.
The series files this game as an opening lesson, because it started as a London System against a King's Indian setup. The opening matters, and I'll get to it. But the moment I still think about is the end.
Where I was
Game #588 sits in week 24 of the journey. That Wednesday was the worst day of the week: before this game I had lost five in a row, games #583 to #587, and my rating had gone from 1027 after game #582 down to 988. Starfiszh was rated 993, almost exactly my level.
The time control was 10 minutes with no increment. That detail decides the story. With no extra seconds per move, every move in a long ending costs time you never get back.
| Game | #588, June 19, 2024 |
| Opponent | Starfiszh (993) |
| My colour, rating after | White, 988 (unchanged) |
| Time control | 10 minutes, no increment |
| Opening | Indian Game, East Indian London System (A48) |
| Result | Draw by repetition after White's 83rd move |
| Accuracy (Stockfish) | Me 78.7, Black 79.5 |
| My blunders / mistakes / inaccuracies | 8 / 1 / 3 |
The opening: the London bishop I gave away
The moves were 1.d4 g6 2.Bf4 Bg7 3.Nf3 Nf6 4.e3 O-O 5.c3 d6 6.Nbd2: my usual London setup, with the dark-squared bishop out on f4 before e3. Black set up a King's Indian: fianchettoed bishop on g7, castled, pawn on d6. Then 6...Nd5, and the knight was attacking my bishop.
I answered with 7.Qb3, attacking the knight, which had no defender. The problem was that it was Black's move, and the knight simply took my bishop first: 7...Nxf4 8.exf4. The engine logged 7.Qb3 as an inaccuracy (+0.28 to -0.70; evaluations are from White's side, so minus means better for Black). Its move was Bg3, keeping the bishop, with the position "about equal".
The engine's own suggestion for Black on move 6 was a related idea: the knight to h5, also attacking the bishop on f4. In this kind of position the London bishop is a target, and the knight jumps to d5 or h5 are how Black hunts it. The lesson for me was simple: when a knight attacks that bishop, move the bishop first and think about counterattacks second. Apart from that, my opening was clean: 92.1 accuracy over the first twelve moves.
The middle: three blunders in a row
The middlegame was level for a long time, and then both of us lost the thread. On move 25 I pushed b2 to b3 (-0.06 to -3.73, blunder; the engine wanted the knight on b5). Black replied 25...Na6, a blunder back (+1.97), and I answered with 26.Nb5, which the engine called a blunder too, because pawn takes c4 kept the advantage (+1.97 to +0.02). Three blunders in three moves, and the evaluation ended where it started.
The ending: my pawn move that let the c-pawn run
After the queens and one pair of rooks came off, it was one rook each and a level position. Then I played 34.h4.
The engine's move was rook to c3, standing in front of Black's c-pawn. My pawn move cost about two and a half points (+0.06 to -2.56). The c-pawn went to c3 and then c2, one square from promoting, and I spent the next dozen moves defending with my rook glued to c1.
Black returned the favour on move 43 by not taking my f3 pawn, I gave it back on move 44, and Black gave it back again on move 45 with ...Rf4, letting me take the c-pawn with my rook. Then 49...Rb4 was Black's last big blunder (-0.27 to +4.17). My 50.g6 was the engine's move, and the analysis marks it as a good moment: my pawn on g6 was suddenly two steps from becoming a queen.
Moment 1: mate in two, and I pushed the pawn
By move 56 I had king, rook and a pawn on g6 against king, rook and a pawn on a3. Black's rook had just gone to e3.
Stockfish saw mate in two: 56.Ra8+ Re8 57.Rxe8#. The rook check forces Black's rook to block on e8, and then my rook takes it with mate, because my pawn on g6 and my king on h6 cover every square the black king could run to. I pushed the pawn instead, and the evaluation dropped from a forced mate to 0.00. I had a minute on my clock.
Black blundered straight back with 56...Rg3, and a few moves later we had traded everything: 59.g8=Q Rxg8 60.Rxg8, then Black's a-pawn promoted and my rook took the new queen on a1. After 62...Ke6 it was my king and rook against a lone king. My clock showed 27.6 seconds after move 63. Black had over a minute.
Moment 2: how to checkmate with king and rook, at 14 seconds
The mate with king and rook only works on the edge of the board. The rook cuts the enemy king off from the rest of the board, rank by rank or file by file, and your own king walks up to take away the squares next to it. When the enemy king has nowhere left to step, a rook check on the edge is mate.
I got Black's king to the a-file by move 73, and then I had a two-move mate twice in a row.
The engine's line was 73.Kc7 Ka7 74.Ra5#. I played 73.Kc6, which was still a forced mate in the engine's eyes but not the fastest. After 73...Ka7 it was mate in two again: 74.Rf8 Ka6 75.Ra8#, the rook going to the back rank first so the mate on a8 is ready. I played 74.Rb5 instead. Both moves kept the win in theory. Neither finished it. I had about 14 seconds.
Moment 3: the repetition
From there my rook and king circled while the black king stepped between b7 and c7. On move 79 I put the rook on d6. On moves 80 to 83 my king went d5, c5, d5, c5, and Black's king answered b7, c7, b7.
On its own, 83.Kc5 isn't a bad move. When I put the position into the engine again for this post, it still rated it a forced mate for White, and its first choice was the rook to e6. The trouble is that the position after 83.Kc5, with Black to move, had already appeared after my 79th and 81st moves. A third time is a draw by repetition, and Chess.com ended the game there. The analysis marks it as a blunder: from a forced mate to 0.00. Any rook move, the analysis's choice was Rc6, would have kept the game going. With 0.7 seconds left, though, I can't say a mate was coming.
What the engine says about game #588
My accuracy was 92.1 in the opening, 80.4 in the middlegame and 76.0 in the ending. Eight blunders, one mistake, three inaccuracies; Black made seven blunders and seven inaccuracies. My best winning chance in the game, by the engine's estimate, was 97.5%. The game lasted about twenty minutes from start to finish, and I used almost every second of my ten.
The lesson I took from game #588
The first lesson is about the opening, and it's short: in the London System, when a knight jumps to d5 or h5 to attack the bishop on f4, the bishop moves. A queen move that attacks the knight doesn't help if the knight can take first.
The second lesson is the real one. King and rook against king is a forced win, and in this game I couldn't finish it in 21 moves. Knowing that the mate exists is different from knowing the moves well enough to play them without thinking. In a 10-minute game with no increment, the second kind of knowledge is the only kind that counts. Both of my missed mates in two with the lone rook came with the black king already on the edge: the position was right there, and I didn't have the pattern ready.
The third lesson is about repetition. When I'm winning, I now check whether I'm repeating king moves. In this game a single rook move on move 83 would have kept the game going.
What came next
My next game that evening was a win against ayoop7 (881), and I finished week 24 at 1020, up from 1009. The following week, week 25, ended with another draw in an ending: a stalemate in game #627. Another ending lesson in the series is the rook against a pawn in game #284. The whole series is on the journey page, and you can replay game #588 on Chess.com.
If the words "blunder", "mistake" and "inaccuracy" in this post blur together, here's the difference between an inaccuracy, a mistake and a blunder. Chess Coach finds moments like 56.g7 in your own games, shows the mate you missed on a board, and explains it in plain words inside Claude or ChatGPT: here's how the analysis works.
Frequently asked questions
Can you checkmate with only a king and rook?
Yes. In game #588 Stockfish rated every position of my king and rook against a lone king, from move 63 to move 83, as winning for White, most of them as a forced mate. The mates it found all happen with the enemy king on the edge of the board and my king close by to cover its escape squares.
Why did game #588 end in a draw?
The same position appeared for the third time after my 83rd move, so the game was drawn by repetition. I had 0.7 seconds left on a 10-minute clock with no increment. The engine still rated the position a forced mate for White.
What was the opening lesson in game #588?
It was a London System against a King's Indian setup. Black's knight jumped to d5 to attack my bishop on f4, and instead of retreating it to g3 (the engine's move) I played the queen to b3. Black took the bishop and I had to recapture with a pawn.