Bishop and knight endgame collapse: game 955
My bishop and knight endgame collapse in game 955 turned a large late advantage into a loss against Prowar3 (1067) on October 1, 2024. At move sixty, Stockfish preferred a bishop check. I moved my king instead, and the saved evaluation fell from +10.14 to -0.10.
That was not my last chance to save the game. Several later decisions returned the position close to equality before I lost again. The two boards here show the missed winning check and a later defensive knight check. They make this more useful than a simple story about one decisive blunder.
The game before the ending
I played White in an Englund Gambit, with ten minutes and no increment. My recorded rating before this result was 1060 and afterwards 1052. The PGN records a loss by game abandonment after Black promoted a pawn on move sixty-six. It does not record a checkmate or explain why play stopped.
| Detail | Game 955 |
|---|---|
| Date | October 1, 2024 |
| Opponent | Prowar3, 1067 |
| My side | White |
| Opening | Englund Gambit, A40 |
| Recorded rating | 1060 → 1052 |
| Engine accuracy | White 65.8%, Black 64.3% |
| My marked errors | 17 blunders, 3 mistakes, 4 inaccuracies |
| Final recorded move | Black promoted the c-pawn to a queen |
The seventeen blunders are spread through a long game. They do not mean that every later position was lost. In fact, the score before my sixtieth move gave me a large advantage. I need to examine the opportunity that still existed, rather than treat the earlier count as a verdict that nothing could be done.
The week 40 diary gives the neighbouring results. Immediately before this game, I had lost game 952 at 1060, won game 953 at 1068 and lost game 954 at 1060. This result made two consecutive losses in that immediate sequence.
Bishop and knight endgame collapse: a check first
At move sixty, my pieces were a king on d5, a knight on b7 and a bishop on f4, with pawns on a3 and e6. Black had a king on e7, a bishop on b3 and several pawns. I moved my king to c5, where it attacked the pawn on b5.
The preferred move was bishop to g5 with check. The position checker confirms the check and the large advantage in that alternative. A check attacks the king and requires an answer. Here that forcing first move was much more valuable than beginning a king journey towards the queenside pawn.
The saved engine line sends Black's king to f8, advances my e-pawn to e7 with check and then brings my knight to d6 with another check. A fresh engine run chooses a different follow-up after the same bishop check. I do not merge those branches into a single supposedly forced sequence. Both analyses identify the bishop check as the move to examine; the numerical swing quoted above belongs to the stored game analysis.
That distinction is useful in my own review. I do not have to memorise a single long line to understand the missed candidate. First I need to see the bishop check. Then I can compare Black's legal replies and study how my king, pawn and knight cooperate in the engine continuation I am actually using.
What my king move allowed
After my king moved to c5, Black pushed the c-pawn from c4 to c3. The saved evaluation was then 0.00. On the next move I captured the b5 pawn with my king. That move changed the score to -2.42, with the engine preferring the knight to d6.
Black then took my e6 pawn with the king. That capture was also marked as a blunder in the stored analysis and returned the score to -0.01. In other words, I had lost a winning advantage, made another mistake and then been given a position close to equality again.
This sequence is why I do not describe the game as an automatic loss after move sixty. The actual moves still contained choices. The opponent did not consistently find the engine continuation either. My review needs to recognise the renewed defensive chance rather than remain fixed on the earlier missed win.
For the evaluation numbers, I use the Stockfish reading guide. A small value near zero is different from the large positive score before the bishop check, even though both positions might look complicated to me without analysis.
Position two: a knight check could preserve the fight
At move sixty-two, I moved my king from b5 to b4. The move attacked Black's bishop on b3 and the pawn on c3. Those targets are real, but the engine preferred moving my knight from b7 to c5 with check, also attacking the bishop.
My king move changed the stored score from -0.01 to -5.95. The knight alternative is a check and an attack on another piece at the same time. That makes it a useful comparison with the first board: in both cases I chose a king move while a forcing move by a minor piece was available.
A minor piece is a bishop or knight. I do not use that label to imply that the two pieces are interchangeable. The first board needs a bishop check; this board needs a knight check. The exact squares and legal replies decide the move. What connects the positions is the need to inspect forcing candidates before moving the king towards material.
The saved continuation after the knight check gives Black's king move to f5. The fresh check also shows a bishop move and the c-pawn advancing later. I can practise that branch as analysis while remembering that the real game followed my king move instead.
Even the second error was not the end of the chances
Black pushed the c-pawn to c2 after my king move. The saved analysis marks that push as another blunder and returns the score to -0.03. I again had a knight check on c5 available. I moved my king to c3 instead, and the evaluation fell to -5.86.
The next moves were different in character. Black's king reached f5, I moved the knight to c5, and Black captured my bishop on f4. My knight then captured the bishop on b3, Black's a-pawn recaptured it, my king took that pawn and Black promoted the c-pawn. Those were actual moves, ending with a new black queen on c1.
I do not need to write an imagined mating finish after the promotion. The recorded game stops there, with an engine advantage for Black and an abandonment result. The repeated returns to equality earlier make the finish more instructive: there was more than one chance to change the course of the ending.
What I can practise from the loss
My recorded clock readings after moves sixty and sixty-two were about ninety and eighty seconds. The ending had a real clock constraint, but those readings do not reveal my thought process. The review task I can justify from the moves is to look for checks by the bishop and knight before committing the king to another pawn target.
The next two games were losses at 1044 and 1036. Games 958 and 959 were wins at 1044 and 1051. That is the immediate rating aftermath; it is not proof that this ending had already changed my play. I keep the training lesson as a task, not an invented historical claim.
The beginner game review guide helps me turn these boards into a repeatable exercise. The journey index supplies the wider history, and the original Chess.com game preserves the many chances on both sides of this ending.
Frequently asked questions
Which move began the late collapse in game 955?
On move sixty I played Kc5 instead of the bishop check Bg5+. The stored score fell from +10.14 to -0.10.
Could I still save the game afterwards?
The saved analysis returned close to equality more than once. Knight checks on c5 were available on moves sixty-two and sixty-three.
How did the game finish?
Black promoted the c-pawn to a queen on move sixty-six. The PGN records a loss by game abandonment.