Is the Halloween Gambit good? A knight for a pawn on move four
Objectively, no. White gives up a knight for a pawn, and the engine rates the position as close to two pawns up for Black. But at club level White actually scores a little more after this sacrifice than in the average game from the same position.
8 
7 
2 
a 
c 
d 
e f 1 h The Four Knights Game has a reputation as one of the calmest openings there is. Everyone brings their knights out and gets ready for a long positional game. And then on move four White takes on e5 with the knight and gives away a piece.
Boo.
From there White's pawns chase the black knight: d4, then d5 or e5, then f4. Every move the knight has to find a new square.
Why is it called the Halloween Gambit?
The name comes from the German player Steffen Jakob. In his words: "Many players are shocked, the way they would be frightened by a Halloween mask, when they are mentally prepared for a boring Four Knight's, and then they are faced with Nxe5."
Trick or treat, in other words. Black gets a free knight. Then comes the trick.
The sacrifice itself is much older. In 1888 the theoretician Oskar Cordel wrote that it used to be played a lot in Leipzig, under the nickname "Gambit Müller und Schulze". There were no such players. In German, "Müller and Schulze" just means "anybody", like "Tom, Dick and Harry" in English.
What the games say
We looked at rated Lichess games (blitz, rapid and classical) since 2013, grouped by the average rating of the two players. You can check any of these numbers yourself in the Lichess opening explorer.
After 1.e4 e5 2.Nf3 Nc6 3.Nc3 Nf6, between 2 and 5 White players in 100 go for the sacrifice 4.Nxe5. Here is how they score (a win is one point, a draw is half):
| Average rating | After the sacrifice 4.Nxe5 | All games after 3...Nf6 |
|---|---|---|
| 1400–1599 | 54% | 51% |
| 1600–1799 | 55% | 51% |
| 1800–1999 | 55% | 52% |
| 2000–2199 | 54% | 52% |
| 2200+ | 52% | 53% |
So in practice the sacrifice works, at least below 2200. It's still unsound, mind you. Defending against it is simply hard. Black needs several accurate moves in a row, often against a gambit they've never seen.
Among players rated 2200+ it already scores a little below average.
How should Black answer?
Take the knight: 4...Nxe5. At every level at least 95 players in 100 do, and they're right.
After 5.d4 the knight has two squares to go back to: 5...Ng6 and 5...Nc6.
1 2 7 h f e d c 8 a The simplest is 5...Ng6. The engine likes it best and gives Black about a pawn and a half. And in the games White scores only 49–52% after it at every level. The extra knight starts to tell.
After 5...Nc6 White scores more, 55–59%, because the pawns keep chasing the knight with 6.d5, then f4 and e5.
1 2 7 h e d c 8 a Unless Black finds 6...Bb4! Grandmaster Larry Kaufman has called it the refutation of the gambit, an idea of the Polish IM Jan Piński. The bishop pins the knight on c3, and the pawn storm loses its punch.
The engine rates 6...Bb4 about the same as the natural 6...Ne5. But in the games the difference is huge: after 6...Bb4 White scores 42–49%, after 6...Ne5 about 59% (below 2200). Part of that is probably who plays it. Whoever knows 6...Bb4 usually knows what comes next, too.
Few people find it. With an average rating of 1400–1999, fewer than 3 Black players in 100 answer 6.d5 with 6...Bb4. Even among 2200+ it's only one in four.
Can you remember 6...Bb4? Then play it, it scores best. If not, 5...Ng6 does the job.
Whichever you pick, don't grab every pawn on offer. Just get your pieces out.
Should you play it with White?
For blitz and for fun, yes. Especially around 31 October. Your opponent is often seeing this sacrifice for the first time and gets confused. But against someone who knows 5...Ng6 or 6...Bb4, you're a piece down for a pawn and a big centre.
More on this opening: the Halloween Gambit traps, where the pawn storm after 5...Nc6 6.d5 Ne5 plays out move by move, our Four Knights Game guide, and the Naroditsky Variation, another surprise on move four (4.Nd5).
Halloween Gambit traps →