Game #364: the win I threw away, and resigning too early in chess

On April 11, 2024, in game #364 of my rapid journey, I had siegb (1020) beaten. The engine had me at +4.65 after move 19, and my best winning chance in the game was 84.1%. I lost it twice over: first by blundering the advantage away, and then, at the very end, by resigning too early in chess terms, in a position Stockfish rated only -0.74 against me, with about six minutes left on my clock. My rating went from 989 to 982.

This is the game I go back to when I want to remember that a lost position and a position that feels lost are not the same thing.

Where I was that week

Game #364 sits in week 14 of the journey, the worst week I ever had. It started at 1051, my best rating so far, and it ended at 993. Game #364 was the tenth loss of an eleven-game losing streak, games #355 to #365, the longest of my whole journey. The three games right before it were losses too, the last one to machoem (1108). So I sat down for this game on a long run of defeats, and it shows in how it ended.

A clean opening, then chaos

The first twelve moves were fine. I played my usual setup with the bishop on f4 (1.d4 d5 2.Bf4), the bishops got traded on d6, both sides castled, and my opening accuracy was 94.4. Then the middlegame started, and over the next 33 moves my accuracy was 54.1, with ten blunders or mistakes. Black's numbers were not much better: Stockfish counted seven blunders for siegb. This was a game where the evaluation swung back and forth almost every few moves.

The first swing was mine, on move 14.

r . q . r . k .
. p . b . p p p
p . n . . n . .
. . . p p . . .
. . . . . . . .
. . P . P N . .
P P Q N B P P P
R . . . . R K .

The engine wanted a quiet move here: a rook to c1, or the pawn to e4, both "about equal". I jumped forward with the knight, Black pushed the pawn to e4, and the evaluation went from -0.30 to -2.83. I then made it worse with a pawn push to c4 (-4.91). But siegb gave it straight back, taking that pawn instead of the engine's move (+0.46). I made another mistake by taking back on e4 with the knight, and then Black put a bishop on f5, which the engine called another blunder. Suddenly I was +2.54.

Moment 1: the knight check I didn't see

r . q . r . k .
. p . . . p p p
p . n . . n . .
. . . . . b . .
. . p . N . . .
. . . . P . . .
P P Q N B P P P
R . . . . R K .

This was the turning point of the analysis. Knight takes on f6 was a check and a double attack (one move that threatens two things at once): it hit Black's king and the rook on e8, and moving the knight also let my queen on c2 attack the bishop on f5. The engine rated it +2.38. I took a pawn on c4 instead. My queen had been one of the defenders of the knight on e4, and when it left, Black could win that knight. The evaluation went from +2.54 to -5.78.

And again Black didn't take it. siegb traded the bishop on e4 and then pushed a pawn to b5 to chase my queen instead of winning my knight. I took on f6 with check, Black took back with the g-pawn, and the evaluation was +4.65 for me: a pawn up, with the pawns in front of Black's king broken. That was my best moment of the game.

Moment 2: the queen trade I refused

For the next ten moves I kept a clear advantage, sometimes +2, sometimes +3. On move 27 I let most of it slip with a rook move to c6 (+2.32 to +0.19), and Black let it come back on move 28. On move 30, Black's queen came to d6, right in front of mine.

. . . r . k . .
. . . . . p . p
p . . q . p n .
. p . . . . . .
. . . . . . B .
. P . . P . Q P
P . . . . P P .
. . R . . . K .

Taking the queen was check, and after the rook took back, the engine rated the position +3.75 for me, "white is winning". I was a pawn up with a bishop against a knight, and without queens Black would have had almost nothing to play for. I pushed my f-pawn instead, and the evaluation dropped to +1.07. I can't say from the record what I was thinking. What the record does say is that I had more than ten minutes on my clock, so this wasn't a time scramble. I simply didn't trade when I was ahead.

Moment 3: the pawn push that left my rook hanging

. . . r . k . .
. . . . . p . p
p . . . . p n .
. p . . . . . .
. . . . . P B .
q P . . P . Q P
P . . . . . P .
. . R . . . K .

Black's queen on a3 was looking at my rook on c1 along the diagonal through b2, and nothing defended the rook. The engine's move was simple: rook to c2. I pushed the pawn again, to attack the knight on g6, and the evaluation went from +1.72 to -7.95. Black could take my rook with check.

siegb didn't. Black moved the knight to e5 instead, and the position was level again (+0.10). Then I blundered on move 32 (-1.94) and again on move 34, when I moved my rook back to c1 and let Black's rook take on g2 with check (-10.16). Black gave that back too, with a rook move to h2 on move 36 that left the evaluation at 0.00. For the next several moves the game stayed close.

Moment 4: the pawn grab, and the resignation

Q . . . . n . .
. . . . . p k p
p . . . . p . .
. p . . . . q .
. . . . . . B .
. P . . P . . P
P . . . . . K .
. . . . . . . .

My king on g2 and my bishop on g4 stood on the same file as Black's queen on g5. The engine wanted the king out of that line first. I took a pawn on a6 instead, and the evaluation went from -1.78 to -6.98. Black answered with a pawn to f5, which attacked my bishop. And then I resigned.

. . . . . n . .
. . . . . p k p
Q . . . . . . .
. p . . . p q .
. . . . . . B .
. P . . P . . P
P . . . . . K .
. . . . . . . .

Black's f5 was a blunder too. In the position I gave up, the engine's best move, queen takes the pawn on b5, leaves Black only slightly better, at -0.74. My bishop was in trouble, pinned in front of my king and attacked by a pawn, and that was clearly all I could see. With about six minutes left in a 20-minute game, I didn't stop to check whether the position was really lost. It wasn't.

The lesson I took from game #364

There are two lessons here, and the second one stings more.

The first is about converting. Again and again, the engine asked for the simple move: trade queens, move the attacked rook, take the king off the dangerous file. I kept choosing the active pawn push or the extra pawn. Being ahead is when I should play the boring move, especially a queen trade that kills my opponent's counterplay.

The second is about resigning. At my level, the game isn't over until the opponent proves it. siegb blundered seven times in this game, and I resigned in a position where Black had just made another mistake. Before I resign now, I make myself look for one more move, check whether I still have time, and ask whether my opponent has really shown they can finish. In game #364 the honest answer to all three was no.

What came next

The next game, #365 against percusantos (984), was the eleventh loss in a row. Then the streak ended: I beat Olga3m777 (972) with my best accuracy of the week, 93.5, and won the game after that too. I climbed back over 1000 on April 14, finished the week at 993, and went into week 15. The whole series is on the journey page.

If you want your own collapses read like this, with the exact move where the win slipped and the move you could have played, that's what I built Chess Coach for. Here's how the analysis works, and why Chess.com's review doesn't always explain the moves. You can replay game #364 on Chess.com.