Belgrade Gambit (1.e4 e5 2.Nf3 Nc6 3.Nc3 Nf6 4.d4 exd4 5.Nd5)
The Belgrade Gambit is the Four Knights Game's hidden weapon: White leaps a knight to d5 after Black recaptures the pawn, sacrificing material for disruptive piece activity. The engine isn't fully convinced, but the positions are chaotic enough to trip up unprepared opponents. Try it against the engine below.
Play the Belgrade Gambit against the engine
Free, no signup — you play white, the engine adapts to your level.
You just played the Belgrade Gambit against the engine. Create a free Chessy account for AI coaching that explains every twist in these sharp positions.
Create a free account →The engine verdict: Black is better
After 5.Nd5, Stockfish evaluates at −0.34 at depth 16 — a third of a pawn in Black's favour. That's a meaningful edge: the gambit is genuinely dubious by engine standards, not just "near-equal with compensation." The engine's prescribed refutation is Nb4, threatening the e4-pawn and the d5-knight simultaneously (Nxf6+ Qxf6 a3). Nb4 doesn't even appear in the top-6 most-played moves — most players naturally capture on d5 or e4 instead, which explains why White scores above 50%.
Why White still wins in practice
Across 107,474 Lichess games White scores 52.5%. The engine's best move (Nb4) is rare; what players actually play is:
- Nxd5 (35,689 games) — takes the outpost knight; White scores 57.8%
- Nxe4 (30,522 games) — grabs the e-pawn instead; White scores 54.7%
- Bc5 (15,971 games) — White only 46.6%
- Bb4+ (6,225 games) — White only 46.3%
The knight captures look natural but leave White better in practice. The quieter piece moves (Bc5, Bb4+) are where Black's engine edge actually shows.
When Black goes wrong
The three catalogued mistakes all miss the engine's correct Nb4. Nxe4 loses 52 centipawns — Black grabs the pawn but allows the Nd5 to activate further. Bb4+ costs 56 cp — a checking move that doesn't threaten anything concrete. d6 costs 69 cp — passive and gives White time. All three are classified as inaccuracies: the Belgrade doesn't punish dramatically, but it steadily accumulates small advantages through practical complications.
Playing it as White
The Belgrade works because its threats look scarier than the engine admits. After 5.Nd5, your knight on d5 eyes f6 and c7 simultaneously. If Black takes on d5 with Nxd5, exd5 gives you a strong passed pawn and a dangerous position. If they take on e4 instead, your Nd5 stays active while Black has to defend. Go in knowing you're objectively slightly worse — the bet is that your opponent doesn't find Nb4 over the board.
Results across 107,474 Lichess games
| Most-played continuation | Games | White wins |
|---|---|---|
| Nxd5 | 35,689 | 57.8% |
| Nxe4 | 30,522 | 54.7% |
| Bc5 | 15,971 | 46.6% |
| d6 | 8,952 | 47.2% |
| Bb4+ | 6,225 | 46.3% |
| Be7 | 5,718 | 42.8% |
Frequently asked questions
Is the Belgrade Gambit sound?
Not objectively. The engine evaluates it at −0.34 (Black's advantage), and the refutation Nb4 is well-known. It's a dubious but playable practical weapon, especially at club level where Nb4 is rarely found.
What's the engine's best reply to the Belgrade Gambit?
Nb4, attacking both the e4-pawn and the Nd5-knight. After Nxf6+ Qxf6 a3, Black consolidates the extra pawn. It's the best-scoring line for Black but the least played of the main alternatives in practice.
Why does White score 52.5% if the engine favors Black?
Because the engine's best move, Nb4, is almost never played. The natural replies (Nxd5, Nxe4) both score above 54% for White. The gambit's dubious character only shows when Black finds the correct technical response.
Is the Belgrade Gambit related to the Halloween Gambit?
Both appear from Four Knights territory and sac material for activity, but they're distinct. The Halloween sacrifices a knight outright on move 4 for two pawns; the Belgrade delays one move and aims for a positional outpost on d5 rather than a pawn storm.
How many games feature the Belgrade Gambit?
Over 107K Lichess games have reached the Belgrade Gambit position. White wins 52.5%, Black wins 44.0%, with 3.5% draws — based on real rated games.
What is Stockfish's evaluation of the Belgrade Gambit?
At depth 16, Stockfish rates the Belgrade Gambit as a slight advantage for Black (-0.34) from White's perspective. This is the computer's assessment of the position after the main opening moves.