Everyone who meets the chess Italian Game as Black hits the same fork at move three. Ask the chess.com forum whether to answer 3.Bc4 with 3...Bc5 or 3...Nf6 and you get a shrug in two flavors. DrSpudnik calls it "really a matter of taste." IM pfren writes that the two are "equally good, and the latter is simpler." Stockfish 18 agrees to a degree that is almost funny: at depth 20 it scores 3...Nf6 at -0.21 and 3...Bc5 at -0.22, one centipawn apart. With the engine silent, the useful question is not which move is best but which one people actually play. We ran the maia3-5m checkpoint, see how Maia chess actually works, on that position at seven rating bands from 1000 to 2200, the same method behind our chess openings by rating dataset, and graded every legal reply with Stockfish 18 at depth 20, see what the Stockfish chess engine is. Two positions, one move order. Maia3 predicts what a player at a given rating is likely to choose. It is not a count of real games.
One opening, three names
The Italian Game is every line that begins 1.e4 e5 2.Nf3 Nc6 3.Bc4, which Wikipedia files under ECO C50 to C59. Inside it, the Giuoco Piano, Italian for "quiet game," is the 3...Bc5 branch, C50 to C54. The Two Knights Defense is 3...Nf6, C55 to C59. So the Italian is the family, the Giuoco Piano is one branch, and "giuoco piano opening" is not a synonym for "chess italian opening."
Duolingo's primer says the Italian Game "is also known as the Giuoco Piano," which squeezes the family into one branch and never mentions the Two Knights at all. Wikipedia's definition runs the other way: the Giuoco Piano page itself says the Italian Game includes the Two Knights, and the Italian Game page adds that the Two Knights was historically "treated as a separate opening." That history is why some pages still split them. Chess.com's Italian Game page, as fetched in October 2026, has no Two Knights section and no ECO codes, though the Two Knights has its own page there, so that may be layout rather than error. If a source uses the two names as one, it is using the older, looser habit, not a different opening.
The family is old. Wikipedia traces it to the 16th and early 17th century masters Pedro Damiano, Giulio Polerio and Gioachino Greco, with Greco meeting 3...Nf6 by 4.Ng5 and 3...Bc5 by 4.c3. Its two pages are not fully consistent on who owns 4.c3: the Italian Game page says Damiano played it as early as 1512, while the Giuoco Piano page says only that Damiano recorded games early in the century and Greco developed the main line in the early 17th. Today the quiet version is back in fashion. Wikipedia credits a 21st century revival to White players looking for ideas that avoid the Berlin Defense, and calls 4.c3 followed by 5.d3 the modern main line in many top level games. Chessable's blog, read in October 2026, adds that Giri, Short and Hou Yifan have used the Italian regularly against other top players.
Position A: after 3.Bc4, what Black actually reaches for
After 1.e4 e5 2.Nf3 Nc6 3.Bc4 Black has 31 legal moves. Stockfish puts 3...Nf6 first at -0.21 and 3...Bc5 second at -0.22. Then the list drops away: 3...d6 is third at -0.48, 3...Be7 fourth at -0.50, 3...h6 sixth at -0.69, and 3...Nd4 sits 17th at -1.20. In the table, "out" means the move was outside Maia's top five at that band. It does not mean zero probability.
| Rating | 1000 | 1200 | 1400 | 1600 | 1800 | 2000 | 2200 |
|---|---|---|---|---|---|---|---|
| Probability of 3...Nf6, the Two Knights (best, -0.21) | 37.3% | 33.5% | 30.3% | 28.9% | 29.4% | 33.7% | 43.1% |
| Probability of 3...Bc5, the Giuoco Piano (-0.22) | 21.5% | 25.1% | 30.6% | 36.4% | 41.9% | 45.3% | 43.8% |
| Probability of 3...h6 (-0.69) | 12.3% | 16.8% | 16.2% | 12.3% | 7.0% | out | out |
| Probability of 3...d6 (-0.48) | 7.8% | 7.2% | 7.2% | 7.4% | 7.0% | 5.0% | 2.7% |
| Probability of 3...Nd4 (-1.20) | 6.2% | 7.1% | 7.3% | 6.5% | 5.2% | 3.4% | out |
| Probability of 3...Be7 (-0.50) | out | out | out | out | out | 6.4% | 5.8% |
At 1000 the Two Knights leads, 37.3 percent against 21.5 for Bc5, and it still leads at 1200, 33.5 to 25.1. At 1400 the order flips by a hair: Bc5 at 30.6, Nf6 at 30.3. From there Bc5 is the likelier move at every band through 2200, by 36.4 to 28.9 at 1600, 41.9 to 29.4 at 1800 and 45.3 to 33.7 at 2000. At 2200 the gap shuts to 43.8 against 43.1, close to a coin flip again. Neither line is monotonic. Bc5 climbs at every band up to 2000 and then slips at 2200. Nf6 falls to its low of 28.9 at 1600 and recovers to 43.1, a U shape that this dataset describes but does not explain.
Because Stockfish's top move here is Nf6, its findability is simply Nf6's probability, which makes 28.9 percent at 1600 the weakest band for finding the engine's choice, and 43.1 at 2200 the strongest. That sounds like a problem and is not one. The likeliest human move is never more than one centipawn from the engine's best: the gap is 0 at 1000 and 1200, where Nf6 leads, and 1 at every band from 1400, where Bc5 does. The two moves together take 58.8 percent of Maia's probability at 1000 and 86.9 at 2200. From 1200 upward the choice narrows steadily toward exactly these two, and it barely matters which of them is picked.
What lower rated players reach for instead
At 1000 the model hands about 41 percent of its probability to moves other than the two main ones, and two of them stand out. 3...h6 is the third likeliest move at 1000, 1200, 1400 and 1600, at 12.3, 16.8, 16.2 and 12.3 percent. It peaks at 1200, shrinks to 7.0 at 1800 and drops out of the top five at 2000 and 2200. It is a waiting move, and at -0.69 it is the engine's sixth choice and about half a pawn worse than Nf6. 3...Nd4 is the other: 6.2 percent at 1000 and 7.3 at 1400, its high, before it fades to 3.4 at 2000 and leaves the top five at 2200. It is the Blackburne Shilling trap move, and at -1.20 it costs about a full pawn.
3...Be7 gets the opposite treatment. It does not appear in Maia's top five at any band until 2000, then takes 6.4 percent at 2000 and 5.8 at 2200, at -0.50, about three tenths of a pawn behind Nf6. The forum is split on it. Ze_Shoopuf suggests "3...Bc5 or 3...Be7" for non experts, while ChessDoofus, an NM, says "Be7 is like an acceptance that you're worse after 4. d4 d6." The engine's 29 centipawn gap sits between those opinions. It is a small concession, not a refutation.
The Two Knights deserves one honest paragraph about its sharp side. After 3...Nf6, White's sharpest try is 4.Ng5 d5 5.exd5, and Black's standard answer is 5...Na5, the Chigorin idea, followed by 6.Bb5+ c6 7.dxc6 bxc6. Wikipedia lists 8.Be2, 8.Bd3, 8.Qf3 and 8.Nh3 for White's eighth move and calls 8.Bd3 the modern main line. Black's other route is 4...Bc5, the Traxler Counterattack, which leads to "wild complications." Wikipedia also says 4.d3 is the more positional choice and is more common than 4.Ng5, often transposing to the Giuoco Pianissimo. Our data measures what Black plays at move three, and at move four in the quiet line, not how often White picks the sharp one. That question stays open.
Position B: after 3...Bc5 4.d3
The late position follows the single likeliest human move at the reference rating of 1400, two plies deeper. That is 3...Bc5, the 30.6 percent choice from the table above, and then 4.d3, which Maia gives 25.8 percent for White at 1400. The result is the quiet system that Wikipedia calls the Giuoco Pianissimo, where White delays or forgoes d4. Wikipedia notes the same quiet system can also arise from 4.c3 Nf6 5.d3. Black has 36 legal moves. Stockfish puts 4...d6 first at -0.26, 4...a6 second at -0.27 and 4...Nf6 third at -0.29, so the three leaders sit within three centipawns. 4...h6 is fifth at -0.39 and 4...Nd4 is 28th at -2.12.
| Rating | 1000 | 1200 | 1400 | 1600 | 1800 | 2000 | 2200 |
|---|---|---|---|---|---|---|---|
| Probability of 4...Nf6 (-0.29) | 46.9% | 45.4% | 43.8% | 43.2% | 44.3% | 49.3% | 61.1% |
| Probability of 4...d6, Stockfish's best (-0.26) | 26.0% | 24.2% | 24.9% | 28.3% | 33.0% | 35.5% | 30.7% |
| Probability of 4...h6 (-0.39) | 12.4% | 20.0% | 23.9% | 22.6% | 17.3% | 10.3% | 4.6% |
| Probability of 4...Nd4 (-2.12) | 4.2% | 2.6% | 1.4% | out | out | out | out |
Here the crowd is much less divided. 4...Nf6 is the likeliest move at every band, never below 43.2 percent and reaching 61.1 at 2200. 4...d6, the engine's pick, is second at every band, and 4...h6 is third at every band. The engine's pick is not a quiet favorite: d6 sits at 24.2 to 28.3 percent through 1600, climbs to 35.5 at 2000 and then falls back to 30.7 at 2200, so its findability is not monotonic either. And the eval gap between the crowd move and the engine's best is 3 centipawns at every one of the seven bands. That is the whole cost of the most popular move at every rating we measured, about three hundredths of a pawn, far too small to decide a club game. The engine's runner up, 4...a6, is a move players barely consider: it enters Maia's top five only at 2200, with 0.8 percent.
The wandering moves behave as they did one move earlier. 4...h6 peaks at 23.9 percent at 1400, the same band where its share is closest to d6's 24.9, then shrinks to 4.6 by 2200. 4...Nd4, already a pawn behind at move three, is nearly two pawns behind here, and Maia's share for it shrinks from 4.2 percent at 1000 to 2.6 at 1200 and 1.4 at 1400 before it leaves the top five at 1600. The cost of a casual knight sortie grows as the position fills out.
Plans, results and what the engine cannot say
Our tables do not say what to play after 4.d3, only what players choose. The plans themselves come from other sources. Chessable describes two. The c3 and d4 plan is the more aggressive one, with an e pawn sacrifice after 7.Nbd2. The d3 plan puts pawns on d3, c3 and h3 and delays a kingside attack. Wikipedia lists the classical Giuoco Piano main line as 4.c3 Nf6 5.d4 exd4 6.cxd4 Bb4+, and the Moller Attack as the pawn sacrifice 9.d5 after 7.Nc3 Nxe4 8.0-0 Bxc3. The Evans Gambit, 4.b4, sacrifices a pawn so White can play 5.c3 with tempo, and Wikipedia says it was "very popular until the end of the era of Romantic chess."
On results, the numbers available are not rating banded, so treat them as a rough mood only. As of October 2026, chess.com reports that in Giuoco Piano main lines Black wins 31 percent, draws 33 and loses 36, and that after the Evans Gambit Declined, 4.b4 Bb6, Black wins 41, draws 23 and loses 36. At the same check, 365chess listed 86,031 games for the whole Italian, with 3...Bc5 in 41,984 of them, 48.8 percent of its database. Its win and draw split is 41.2, 26.8 and 32 percent, though the fetch did not make clear whose perspective that is. The two sets measure different things, main Giuoco Piano lines versus the entire Italian in different databases, so they should not be set side by side. On how drawish the opening is at your own level, forum posts call it boring and no source splits the figures by rating. We would not claim an answer.
Where other pages are better
This page is narrow on purpose, and other sources are better at what it leaves out. 365chess is the stronger reference if you want move frequencies from a large database and a catalogue of more than fifteen named sub variations, from the Rousseau Gambit to the Jerome Gambit. Chess.com's page is better for quick results by variation. Wikipedia is the better history. Chessable is better at explaining the c3 and d4 plan against the d3 plan in words. Simplify Chess describes the Two Knights as "more of a counterattack than a defense," a framing our numbers cannot test. None of them splits the replies by rating, which is the gap this page fills, and none of them has an engine score for each move. What we do not measure is also worth stating: White's fourth move after 3...Nf6, the Evans Gambit, the Fried Liver and any result data. Two positions are the entire scope of the measurement.
What to take from it
Choosing between 3...Bc5 and 3...Nf6 is a taste decision, and the numbers confirm it in a specific way. At 1000 and 1200 the Two Knights is the likeliest answer, at 1400 the crowd drifts to the Giuoco Piano, and by 2200 the two are almost level. The engine was never the reason. The small number of tempting second tier moves, 3...h6 and 3...Nd4, are where the real cost lives. A pawn is a lot and half a pawn is something, but neither is the kind of mistake that decides a game on its own.
Once you are in the Giuoco Pianissimo, a fashionable and quiet position for both colors, the move most players reach for is good. The difference between 4...Nf6 and the engine's 4...d6 is three centipawns. That is the opposite of the French Defense, where our French Defense findability data found a crowd favorite costing about a pawn, and a close cousin of the Sicilian Defense lines we measured, where the likeliest human move was the engine's own pick at every band. Plans matter more than the exact move, which is the argument of why learning plans beats memorizing openings. Pick the branch you like, learn its pawn structure and the two or three moves where lower rated players drift, and spend the rest of your study time on the middlegame.
Maia3 is a model of human choice built from the maia3-5m checkpoint, not a count of real games, and every percentage above should be read that way.
Frequently asked
- Is the Italian Game the same as the Giuoco Piano?
- No. The Italian Game is the whole family of lines starting 1.e4 e5 2.Nf3 Nc6 3.Bc4, filed under ECO C50 to C59 on Wikipedia. The Giuoco Piano, Italian for "quiet game," is only the 3...Bc5 branch, C50 to C54. The Two Knights Defense, 3...Nf6, is C55 to C59. Duolingo's primer says the Italian Game "is also known as the Giuoco Piano," which merges the family into one branch. Wikipedia adds that the Two Knights was historically treated as a separate opening, which is part of why the names get tangled.
- Should Black play 3...Bc5 or 3...Nf6?
- The engine will not choose for you. Stockfish 18 at depth 20 scores 3...Nf6 at -0.21 and 3...Bc5 at -0.22. Human opinion is just as split: DrSpudnik on the chess.com forum calls it "a matter of taste," IM pfren writes that the two are "equally good, and the latter is simpler," and opticRED says only that the Two Knights feels comfortable. In our Maia3 data Nf6 is the likelier move at 1000 and 1200, Bc5 leads from 1400 through 2200, and at 2200 they are nearly tied at 43.8 and 43.1 percent. Pick the one whose follow-up positions you will enjoy playing.
- Is the Italian Game good for beginners?
- Duolingo recommends it mainly for beginners and Chessable says it suits "beginner through to grandmaster," so neither source limits it to one level. Neither page tells a beginner which small mistakes to avoid. Our data supplies one: at 1000, Maia3 gives 3...h6 12.3 percent of its predicted replies, a move about half a pawn behind Stockfish's best, and gives 3...Nd4 6.2 percent, a move about a full pawn behind. Both are easy to avoid.
- Does Black need to fear the Fried Liver Attack?
- It only arises after 3...Nf6 4.Ng5 d5 5.exd5 Nxd5 6.Nxf7. Wikipedia says 5...Na5 is generally considered Black's best and that White's attack is strong "but it has not yet been proven to be decisive." It also reports that Fischer thought 6.d4, the Lolli Attack, was "incredibly strong," which is why 5...Nxd5 is rarely played. Tarrasch called 4.Ng5 "a duffer's move" because it moves a piece twice before development is finished, and Wikipedia says 4.d3 is the more common fourth move in the Two Knights. Our data does not measure White's fourth move, so it cannot say how often a club player meets 4.Ng5.
- Is 3...Nd4 a good move against the Italian Game?
- Not by the engine. It is Stockfish's 17th ranked reply of 31, at -1.20, about a pawn worse than 3...Nf6. It is the Blackburne Shilling trap move, and Maia3 gives it 6.2 percent at 1000, 7.1 at 1200, 7.3 at 1400, 6.5 at 1600, 5.2 at 1800 and 3.4 at 2000. By 2200 it falls out of Maia's top five. It is a minority habit rather than a mainstream choice, and the numbers do not show it rising or falling steadily with rating.
Want to see what your own rating finds after 3.Bc4?
Chessdrive runs the real maia3-5m checkpoint on-device as your opponent at any rating you set from 600 to 2600, so you can find out whether Nf6, Bc5 or something stranger is what you would actually play under a clock, not just in a table. Finished games are reviewed separately with Stockfish so you can see exactly where the position turned. Available now on iOS.
Keep reading
Updated