Week 111: how to spot a knight fork before it lands

Fifteen games from March 23 to 29, 2026, and the rating barely moved: 6 wins, 8 losses and a draw, 1242 → 1233 (−9). With White I scored 4–2–1; with Black only 2–6. The review had two threads. The first is an old friend from week 106: a black knight landing on f2 after I castled queenside and forking my rooks. The second is the mirror image, a fork I could have played. So this week is about how to spot a knight fork, from both sides of the board — plus one game I'd rather forget, a winning ending that I turned into a stalemate.

The week in numbers

Three of my wins and three of my losses ended in checkmate. Thirty-three of the 43 blunders were in the middlegame. The strongest opponent was Revati1987 (1438), a loss in 18 moves.

The fork on f2 again: game 2012

A knight fork is a knight move that attacks two pieces at once. Knights are the classic forking piece because their jump is hard to see and nothing can block it. Against Maxmudjon8538 (1287), a 30-minute game, I castled queenside on move 14, Black's knight came to e4, and I played 16.Bd3. The engine wanted 16.Rdf1, covering f2 (+3.04). After 16.Bd3 Nxf2 the knight attacked both my rooks.

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

At this point I'm losing a rook for the knight either way. The engine's line after 17.Qc2 is 17...Nxd1 18.Qxd1 and I'm still slightly better (+0.99). After 17.Rdf1 the knight takes the other rook, 17...Nxh1 18.Rxh1, and Black's second knight also takes my e3 pawn, 18...Nxe3 (−1.21). That's exactly what happened, and the game ended with 35...Qxa2# (game). I had 1572 seconds on my clock at move 17 — more than 26 minutes — and still didn't stop to ask where that knight could jump.

The fork I didn't play: game 2010

The same week I had the chance to be the one with the knight. As Black against MRMREazy (1324), in another 30-minute game, the position was level on move 16, and White's pawn on f4 was attacking my knight on e5.

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

16...Nf3+ is a fork: check to the king and an attack on the rook on e1. The knight can be taken, but the engine's line shows why it still works: 17.gxf3 Qg3+ 18.Kh1 Qxh3+ with my queen close to White's king (+0.19, about equal). I played 16...Be6, which attacks a bishop but leaves my knight hanging; the engine answers 17.Bxe6 or 17.fxe5 (+2.53). My opponent resigned three moves later anyway, but that was luck (game). When one of my pieces is attacked, the first question is whether it can move with a threat — and a knight with a check is the best threat there is.

The stalemate: game 2017

Against yougottaanticipate (1247) I played 62 moves and made 12 blunders, the most of any game this week. And yet I reached a king and pawn ending at +6.90, completely winning. On move 59 I pushed my pawn, 59.d6, and the evaluation dropped to +0.37 in one move; the engine wanted 59.Kb5, going for Black's a-pawn first. Three moves later, after 61.d7+ Kd8 62.Kd6, Black's king had no legal move and it was stalemate, a draw (game).

The lesson of the week

Knight forks are predictable: a knight can only jump to eight squares. When an enemy knight is in my half of the board, I look at those squares and check whether any of them attacks two of my pieces or my king plus a piece. In the London, with my king castled long, f2 is the square I keep forgetting. And when my own piece is attacked, I look for a knight check before retreating.

All games this week

Previous: week 110 · Next: week 112 · The whole journey · Review your own games like this