Three Knights Opening Traps: The Nb5 Corner Raid, Move by Move

The Corner Raid

Your c3-knight has one route memorised: b5, then c7, then a8 — where his rook is sitting. A quarter of opponents open the door.

After 3.Nc3 your queen's knight is one hop from d5 and b5 — the two squares that hit c7. That matters because a knight landing on c7 checks a king on e8 and attacks a rook on a8 in the same instant. So the knight waits. You take on e5, you push d4, and you wait for his queen to step off d8 to defend something. The moment she does, the tour is on: b5, c7, a8. Which of the two squares you jump to decides the game — 33% of players pick the wrong one and lose their queen for it, and this tree shows you exactly why. Better still, the most common defence gets checkmated on move nine, because his knight never left g8 and his own rook cannot reach f8 through it. Every number below was measured — real Lichess frequencies at 1000-1600, Stockfish at depth 18.

The line, move by move

1. e4 e5 2. Nf3 Nc6 3. Nc3 Bc5 4. Nxe5 Nxe5 5. d4 Bxd4 6. Qxd4 Qf6 7. Nb5 Kd8 8. Qc5 c6 9. Qf8#

1. e4 (you) Standard first move. Nothing here hints at what your queenside knight is going to do.
1... e5 (opponent) He mirrors — the most played answer in chess.
2. Nf3 (you) Nf3, hitting e5. Still nothing unusual.
2... Nc6 (opponent) He defends the pawn with a knight. Now the move that names the opening.
3. Nc3 (you) Nc3. It develops, it guards e4, and it parks the knight one jump from both d5 and b5. Both of those squares attack c7, and c7 is the square where a knight forks a king on e8 and a rook on a8. Hold that thought for six moves.
3... Bc5 (opponent) …Bc5 — 24%, the second most popular move, and the one that leaves e5 defended by exactly one piece while his bishop stands on a square a pawn can hit. That is the whole invitation.
4. Nxe5 (you) Nxe5! Only 10% play this — 39% develop with Bc4, which is genuinely fine and a hair better on the engine (+0.6 against +0.2). Bc4 gets you a game; Nxe5 gets you everything on this page. And if he answers 4…Bxf2+, which 18% do, simply take with the king: 5.Kxf2 leaves you at +1.1 with the whole centre and his best bishop gone.
4... Nxe5 (opponent) 73% recapture, and they are right to. Now the move that turns the sacrifice into a loan.
5. d4 (you) d4! It forks his bishop on c5 and his knight on e5 in one push. He cannot hold both, so he takes.
5... Bxd4 (opponent) …Bxd4 — 54%, the main line by a distance. He grabs the pawn with the bishop and expects to stay a piece ahead for a while.
6. Qxd4 (you) Qxd4! Everything is back to even and his knight on e5 is attacked with nothing defending it. How he saves that knight is the entire game.
6... Qf6 (opponent) …Qf6 — 15%. The knight is defended, the queen is developed, and it is the most active-looking move available. It is also the move that walks his queen off the only square guarding c7.
7. Nb5 (you) Nb5!! Two jobs in one move: it attacks c7, where a capture forks his king on e8 and his rook on a8, and it covers d4 from behind your own queen. That second job is the entire point. 33% of White players jump to d5 instead — it hits his queen AND c7, so it looks like the stronger square. It loses on the spot: 7.Nd5 Nf3+! 8.gxf3 Qxd4 and nothing is guarding d4, measured at −3.5 for you. Nb5 is played by 9% and measures +3.6. Same knight, one square apart, a seven-pawn swing.
7... Kd8 (opponent) …Kd8 — 44%, and it makes total sense: the king steps across and defends c7 personally. He has given up castling forever to do it.
8. Qc5 (you) Qc5! 42% of players find it and it reads +7.9. The queen lines up on c7. His king 'defends' that square, but Kxc7 would be illegal — your knight on b5 covers it. Qxc7+ is unstoppable. He has something far worse to worry about first.
8... c6 (opponent) …c6 — 39%, the most common move here and a perfectly logical one: it attacks your knight on b5 and shuts the c-file. Before you move that knight, look at his back rank.
9. Qf8# (you) Qf8#. Walk his escape squares: c7 is covered by your knight, d7 has his own pawn on it, c8 has his own bishop, e8 sits on your queen's rank. His rook on h8 would cover f8 — except his own knight is still parked on g8 in the way, and his queen's route from f6 is blocked by his own pawn on f7. (19% play …Nc6 here instead: identical mate.)

Every branch cashes out

8 coached branches in this family. Engine-checked finishes:

Bc5 / Qf6 / Kd8 / c6Forced mate
Checkmate on move nine, delivered onto a square his rook was covering until his own undeveloped knight got in the way. Nothing he played was a howler: Bc5, Bxd4, Qf6, Kd8, c6 all look reasonable one at a time.
Bc5 / Qf6 / Kd8 / Ne7Forced mate
Two different mates from one position, two moves apart, and the move that dodged the first one set up the second. Blocking your own pieces is not a beginner's mistake — it is THE mistake, and it costs him twice in a row.
Bc5 / Qf6 / Nf3++5.9 by force
+5.9 — a whole piece, won by a knight move you made two moves earlier for a completely unrelated reason. Nb5 was aimed at c7; the fact that it also guards d4 is what turns his most forcing move into a losing one.
Bc5 / Qf6 / Nc6+6.3 by force
+6.3. The knight that started the game on c3 finished it in a corner it may never leave, having eaten a rook to get there. Trading queens first is the entire difference between this and losing your own queen to …Nf3+.
Bc5 / Nc6+17.2 by force
+17.2 — his queen for a bishop, plus the g7 pawn, with his rook still trapped on h8 and his king still on e8. The bishop on h6 did two jobs at once: it protected your queen and it baited his.
Bc5 / d6Sound line
+1.5 and a pawn against the move 45% of opponents choose — no fireworks, because the solid defence has earned a solid answer. Take the pawn, steer for the endgame, and let the extra pawn do the work.
Nf6Sound line
0.00, and that is the honest number. Half your opponents delete the entire trap with one natural developing move, and recognising the moment it happens — instead of playing the trap into a position that no longer contains it — is worth more rating than the trap ever was.
Bb4Sound line
+0.7 and a pawn in six moves, off a bishop move that pinned nothing at all. His knight on b4 will poke at c2 for a while; a2-a3 sends it home and the pawn stays yours.

Why it works

The whole opening rests on one square. A knight landing on c7 checks the king on e8 and attacks the rook on a8 in the same instant, and from c3 the knight is one jump from b5 and d5, the two squares that reach it. The catch is that c7 starts the game defended by the queen on d8, which is why the version of this trap chess folklore quotes, 3...g6 4.d4 exd4 5.Nd5 with Nxc7+ to follow, does not actually exist: the queen just takes the knight, and in the explorer 100 percent of the games that reached that position did exactly that. The real trap waits. You capture on e5, push d4, and let his queen walk off d8 to defend something. The moment she does, the tour is on.

The move that decides everything is the one that looks weaker. After ...Qf6 defends the loose knight, 33 percent of White players jump Nd5, because it attacks the queen and c7 at once. It loses on the spot: ...Nf3+ forks king and queen, and after gxf3 Qxd4 nothing guards d4, measured at -3.5. Nb5 attacks the same c7 square while also covering d4 from behind your own queen, and it measures +3.6. Only 9 percent of players find it. Same knight, one square apart, a seven pawn swing.

The finish is prettier than it has any right to be. In the main line the king guards c7 himself with ...Kd8, Qc5 lines the queen up anyway, and Qf8 is checkmate on move nine, because his rook on h8 cannot cover f8 through his own undeveloped knight on g8. Nothing Black played was a howler: Bc5, Bxd4, Qf6, Kd8 and c6 all look reasonable one at a time. The sidelines cash out too, from a +17.2 queen win when he hunts the wrong piece to a clean +6.3 rook raid when queens come off first.

How to punish the Three Knights Opening

The refutation is one developing move. 3...Nf6 turns the game into a Four Knights, and 49 percent of players find it without knowing they are defusing anything. It matters because the whole mating pattern needs the knight parked on g8, blocking the rook off f8; move it and the tree above stops existing. White's best after that is Bb5 at +0.6, a normal game with a normal edge, and the honest number at the end of that line is 0.00.

If you are already in the Bc5 line and White has taken on e5 and pushed d4, defend the attacked knight with 5...d6 rather than the active looking ...Qf6. The pawn does the job, your queen stays home guarding c7, and the worst that happens is White grinds a pawn out of the e5 arithmetic at around +1.5. Every disaster in this opening starts with the queen leaving d8.

How to avoid falling into it

The alarm is a knight sitting on c3 while pieces come off in the centre. Before your queen leaves d8 for any reason, ask what happens on c7 the move after: if a knight can reach b5, the fork is armed and the rook on a8 is the price. The same goes for your kingside: get the g8 knight out early, because it is the difference between your rook covering f8 and a queen delivering mate there.

And when the raid does land, do not trust forcing moves to save you. The ...Nf3+ fork looks like a refutation and loses a clean piece at +5.9, because the knight on b5 was quietly covering d4 the whole time. Count what defends the capture square before you check, not after.

Questions people actually ask

Is the Three Knights Opening any good?: Yes, and that is part of the appeal: 3.Nc3 develops a piece and guards e4, so nothing is risked to set the trap. Against the most common reply, 3...Nf6 at 49 percent, you get a standard Four Knights where Bb5 measures +0.6. The corner raid and the move nine mate are a bonus that a quarter of opponents walk toward with 3...Bc5.
Does the Three Knights c7 fork actually work?: The version the internet quotes does not. After 3...g6 4.d4 exd4 5.Nd5, the fork 6.Nxc7+ just loses a knight to ...Qxc7, because the queen never left d8, and 100 percent of the explorer games that got there played the capture. The fork only becomes real once the queen steps away, for example to f6, which is exactly the moment this tree teaches you to strike.
Should the knight jump to b5 or d5?: b5, and the numbers are brutal about it. 7.Nd5 attacks queen and c7 at once and is chosen by 33 percent of players; it loses to ...Nf3+ and ...Qxd4 at -3.5. 7.Nb5 hits c7 and covers d4 in the same move, is found by only 9 percent, and measures +3.6.
What if Black just defends with 5...d6?: Then there is no trap, and pretending otherwise costs you your advantage. 45 percent of opponents play it. The honest plan is Bf4, take on e5 once two attackers beat one defender, and bank about +1.5 and a pawn. Solid defence earns a solid answer, not fireworks.
At what rating does the Corner Raid work?: The frequencies behind this tree come from blitz and rapid games between 1000 and 1600, drawn from 56.4 million games after 3.Nc3. In that band 24 percent play the 3...Bc5 that opens the door, and 44 percent of those who face Nb5 answer with the losing ...Kd8. Higher up, more opponents choose 3...Nf6 and you simply get the sound Four Knights instead.

Drill the Three Knights Opening against a real opponent

The trainer above plays every branch with a coach explaining each move. When you are ready, take it into a live game, or paste your Chess.com or Lichess username and see which openings actually cost you rating.

Play it for real Get your free scout report

More opening weapons

New to openings? Learn openings without memorizing explains the approach behind every walkthrough here, and the free AI chess coach guide covers what the rest of the app does without a card.