Queen endgame time scramble: the win I threw away in game 1220
Game 1220 is the cleanest example in my record of a queen endgame time scramble going wrong. I was a knight up against Calibogue (1159) for about twenty moves, with Stockfish showing −6 or so for me as Black. Then, with less than a minute on my clock, I put the knight on a pinned square, lost it, missed a queen trade that would have won, and finally moved my king so that my queen hung. I lost on time a few moves after that.
The result cost me eight points, from 1157 to 1149. What makes it worth a whole post is that the engine had me winning almost all the way, and each of the decisive mistakes is a different kind of mistake.
The recorded game
I played it on December 22, 2024, in ten-minute rapid without increment, as Black. It was my fifth game of the day: I had won games 1216 and 1217, lost 1218 and won 1219, which had taken me to 1157.
| Detail | Saved record |
|---|---|
| Game | 1220 |
| Opponent | Calibogue (1159) |
| Opening | King's Indian Defense, Normal Variation (E90) |
| Result | Loss on time (1-0) |
| My Stockfish accuracy | 71.7% (White 74.6%) |
| My blunders / mistakes / inaccuracies | 11 / 3 / 5 |
| My best winning chance | 95.8% |
| Original game | Chess.com |
The phase split explains the shape of the game. My opening moves scored 89.9%, the middlegame 76.6%, and the fifteen endgame moves only 54.2%. Seven of my mistakes and blunders came in the endgame, when I had the least time.
The first chance I let go
The King's Indian is a setup where Black fianchettoes the bishop on g7, castles, and then fights for the centre with ...e5 or ...c5. In this game I swapped my light-squared bishop for White's knight on move 7 and played ...e5 on move 9. After a messy opening White blundered on move 15, retreating the knight to c3, and the evaluation jumped to −8.47.
I attacked the knight with my queen. Stockfish wanted me to take it with check: after the queen recaptures, my rook takes the pawn on e4, also with check, along the open e-file towards White's king. My move was not a disaster, but the evaluation dropped from −8.47 to −2.99. The difference is forcing moves: a capture with check gives the opponent fewer choices than a quiet attack.
Several swings later, on move 29, I was clearly winning again, and from then until move 48 every evaluation was −4.5 or better for me. By move 48 the board had only queens, kings, my knight and two pawns each. I was a knight up.
Move 48: the knight on the pinned square
White's queen had just given a check from b2, on the long diagonal that runs b2–c3–d4–e5–f6–g7. I blocked the check with my knight on e5. That is a legal block, but a knight standing between a queen and a king is pinned: it can't move without exposing the king. White answered with the obvious 49.f4, a pawn attacking the pinned knight. The evaluation went from −6.17 to −2.63.
The engine's choice was simply to step the king out of the line with Kf7. Other quiet moves, such as blocking with the queen on e5 or moving the king to h7, were also judged winning when I checked the position again. I had 53 seconds left after my move; White had 127.
The knight was gone by move 52. After 50...Qe4+, another blunder by the engine's count, White took it with 52.fxe5+ and I recaptured with the queen. I had queen and two pawns against queen and one pawn: still slightly better for me by the engine's numbers, but no longer a piece up.
Move 58: the queen trade I didn't see
White had just blundered with 58.Qh4+, and the evaluation was −6.29 for me. The stored engine line after Kf3 shows why: if the queens come off on f4, my king recaptures the last White pawn and my b-pawn is far away from White's king. A pawn endgame like that is exactly what I wanted. I moved the king the other way, to f5, and the evaluation became 0.00. I had 12 seconds after the move.
This one hurts more than move 48. Trading queens when ahead is one of the simplest endgame ideas. Here it meant walking my king towards White's queen and king instead of away from them, and I went the other way.
Move 64: the king move that dropped the queen
White's queen checked me from g3. Both king moves, Ke3 and Ke4, open the diagonal between the two queens. The difference is that from e4 my king still protects my queen on e5, and from e3 it doesn't. After Ke3, White took my queen with check, 65.Qxe5+, and the evaluation jumped from −0.05 to +13.87. I had 1.5 seconds left after the move. White kept checking, and my clock ran out after White's 67th move.
What the engine says about my queen endgame time scramble
The tempting summary is "I lost on time", and that is what the result says. But the clock was only the background. Each of the three endgame mistakes has a specific reason I can practise: a knight blocking a check on a line with my own king behind it is pinned; when I'm ahead, a queen trade is the first thing to look for; and when a king guards a piece, every king move needs a quick look at what it stops guarding.
All three are checks against my king, or my replies to them. That is the nature of a queen endgame: the side with less to lose keeps checking, and the side that is ahead has to find safe king squares quickly. With seconds left, I need the habits to be automatic. The phase numbers say I wasn't there yet: 89.9% in the opening and 54.2% in the endgame.
What came after
I lost the next two games as well, games 1221 and 1222, and fell to 1133. Then, on December 23, I played the King's Indian Normal Variation again in game 1223 against jones_961 (1247) and won. That started a run of six straight wins up to 1191, and the next day I crossed 1200 for the first time, in game 1237. The records show the sequence; they don't prove that this loss changed anything.
The surrounding games are in my weekly diary for that week, and the whole series is on the journey hub. If this pattern looks familiar, I've written more generally about why I keep losing on time and why I keep losing winning positions. You can also review a finished game like this one, move by move, with Chess Coach in Claude or ChatGPT.
Frequently asked questions
How did I lose game 1220?
On time. I played Black against Calibogue (1159) in a King's Indian Defense, was a knight up with Stockfish showing about −6 for most of the middlegame, then lost the knight, missed a winning queen trade and hung my queen with 1.5 seconds left. My flag fell a few moves later.
What is a queen endgame time scramble?
It's what my game turned into: only queens, kings and a few pawns on the board, both players checking or dodging checks, and very little time. Every move is a check or an answer to one, so one slip with the king can lose the queen.
Was I really winning?
Yes, by the engine's numbers. Every evaluation from move 29 to move 48 was −4.5 or better for me as Black, and my best winning chance in the game was 95.8%.