Game #499: how I lost a rook endgame with connected passed pawns
On June 3, 2024, in game #499 of my rapid journey, I reached the kind of endgame beginners are told is "just technique": a rook and two connected passed pawns (two pawns side by side with no enemy pawn able to stop them) against a lone rook. Stockfish had me at -7.30, completely winning. I had 16 minutes on my clock. On move 62 I put my rook on a square next to White's king, the king took it, and I left the game.
I lost to Bobthebuilder222333 (993) and dropped to 953. My best winning chance in the game, by the engine's estimate, was 93.6%.
The first fourteen moves of this game, a Giuoco Piano where I chased White's bishop and never took it, are in a separate post about the opening. This one is about everything after move 25, when the game slowly turned in my favour and I kept handing it back.
Where I was
It was a Monday in week 22 of the journey, the week of 91 blunders, and this was my fourth loss in a row. Earlier that day I'd lost to StrawHat_Mike47, marivet88 and Victorlutte. The week summary lists game #499 among nine games that week where I had at least an 80% chance of winning and didn't win.
| Game | #499, June 3, 2024 |
| Opponent | Bobthebuilder222333 (993) |
| My colour, rating after | Black, 953 (from 960) |
| Time control | 15 minutes + 10 seconds per move |
| Opening | Giuoco Piano Game (C50) |
| Result | Loss, game abandoned after White's 63rd move |
| Accuracy (Stockfish) | Me 58.8, White 77.1 |
| My blunders / mistakes / inaccuracies | 8 / 3 / 6 |
| Endgame accuracy (29 moves) | 52.1 |
How the advantage arrived
After the opening White was a bit better, and by move 23 queens were off and the position was level. Then White started drifting: 25.Bb4 and 26.Rc1 were inaccuracies, and I was -2.53. My knight went to b2 (a blunder by the engine's count; its move was 26...Rc8), and White gave it straight back with 27.Rc6. I was -4.92, and a move later my knight was on d3.
Moment 1: the knight that took one pawn too many
I took a pawn on f2 with the knight, and the evaluation went from -5.13 to -0.35. The engine's reply for White was 30.Kg3, attacking the knight, and its line had me giving the knight back for two pawns. The pawn push d5 kept my advantage. White didn't find Kg3 (30.Rc3 was a blunder), and I was -5.34 again. Two moves later I moved my rook from e8 to c8 instead of e6 and dropped from -6.10 to -2.16. On move 33 I gave the knight back for the pawn on e4.
Moment 2: the wrong capture with check
By move 39 a pair of rooks had been traded, it was my rook against White's rook, and White's king had come up to f4. I was two pawns up and -5.11.
My capture looked like a double attack (one move that threatens two things at once): check and the d4 pawn. But it opened lines for White, and the evaluation went from -5.11 to -1.91. The rook check came first in the engine's line: 39...Rh4+ 40.Ke3 dxe5 41.dxe5, and the order of the moves kept me winning. White blundered back on the very next move (40.dxe5, when the king should have taken), and I was -5.14.
Moment 3: three pawns up, and the king step
Three pawns up. My king went to attack the pawn on e6, and the evaluation went from -5.67 to -0.30. The square e7 put my king on the seventh rank, and White's rook came to d7 with check and took on f7. After 44.Rd7+ Kf8 45.Rxf7+ the engine called it exactly level.
Then it kept swinging. On move 48 White's rook went to b7 (a blunder, -3.74), I pushed my h-pawn instead of bringing the rook to g5 (0.00 again), and on move 49 White pushed the e-pawn to e7 with check (another blunder, -4.82). By move 52 I had taken that pawn: my rook and two connected passed pawns on g7 and h4 against White's lone rook. White had no pawns left.
Moment 4: ten moves of technique, one move of nothing
For the next ten moves I mostly did it right. The pawns went forward, h3 and then g5 and g4, my rook came to g2, and my king came to g6. Apart from two inaccuracies the engine had no complaints, and the evaluation went from about -4.6 to -7.30.
The rook on g3 stood right next to White's king on f4, and nothing defended it: my pawns on g4 and h3 cover f3, h3 and g2, not g3. -7.30 to +4.07. 63.Kxg3, and I had two pawns against a rook. When I checked the final position for this post, Stockfish scored it about +82, completely winning for White.
I had 16 minutes and 28 seconds on my clock. In a game with a 10-second increment, I'd played 62 moves fast enough to finish with more time than I started with. Across the whole game I never dropped below 14 minutes. Losing on time wasn't the risk here. Not using it was.
What I learned from game #499
Look at every square next to the enemy king. It's the most basic check there is: before I put a piece somewhere, is it defended, and can the king reach it? With an endgame this simple I didn't do it.
Connected passed pawns need a plan, not just pushes. The engine's top moves on move 62 were 62...Kh5 and 62...h2: the king walks with the pawns, or the pawn goes to the second rank. My rook had no job on g3. It wasn't blocking a check, protecting a pawn or cutting off the king.
Use the clock when the evaluation swings. Six times in this game I had the engine at -3.7 or better and played a move that dropped it to -2.2 or worse: moves 29, 32, 39, 43, 48 and 62. And I was playing quickly the whole time. If the rook vs pawn endgame of game #284 taught me anything, it's that endgames reward counting.
Resign, don't leave. The record says the game was abandoned. That happened again in game #506 the next day and in game #509 two days later. Resigning is one click.
What came next
That evening I won two in a row, against 0yyyzu (1015) and alonsomorris (987), and climbed back to 972. The next day began a seven-game losing streak, the one that included game #506 and ended at game #510. The week is in week 22, the whole series is on the journey page, and you can replay game #499 on Chess.com.
If you look at your own evaluation graph and see a line that swings up and down like this one, I wrote how to read Stockfish analysis. Chess Coach reads it for you inside Claude or ChatGPT and points at the move where it turned: how the analysis works.
Frequently asked questions
How do you win a rook endgame with two connected passed pawns?
In my game the engine's plan was simple: keep the rook safe and bring the king and the pawns forward together. On move 62, with my rook on g2 and pawns on g4 and h3, its choices were 62...Kh5 and 62...h2, both rated about -6.7, completely winning for Black.
What was the losing move in game #499?
62...Rg3. The rook moved next to White's king on f4 with nothing defending it, and 63.Kxg3 took it. The evaluation went from -7.30 to +4.07 in one move.
Were you short of time?
No. It was a 15-minute game with 10 seconds per move, and I had more than 14 minutes on my clock for every move of the game. I had 16:28 left when I played 62...Rg3.