Chess Scandinavian Defense: What Every Rating Plays

Ask the chess.com forum which queen retreat to use in the Scandinavian Defense after 3.Nc3 and you get a three way argument: d8, d6 or a5. The sources agree on the menu and then split on the order. Wikipedia calls 3...Qa5 the main line and the most popular option, ChessCompass calls 3...Qd6 the most popular modern choice, and Chess.com calls 3...Qd8 passive but playable. Our own engine run adds a fourth voice that is not an opinion. We ran the maia3-5m checkpoint on the position after 1.e4 d5 2.exd5 Qxd5 3.Nc3 at seven rating bands from 1000 to 2200 and graded every legal reply with Stockfish 18 at depth 20. It is the same method behind our chess openings by rating dataset; for background, see how Maia chess actually works and what the Stockfish chess engine is. 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 Scandinavian Defense begins 1.e4 d5 and is filed under ECO B01. Wikipedia also lists Center Counter Defense and Center Counter Game as names, and Scandinavian Defence is the British spelling of the same opening. After 2.exd5 Black chooses between the Queen Variation, 2...Qxd5, and the Knight Variation, 2...Nf6, which Chess.com calls the Modern Variation. This page covers the queen line, because that is where the biggest rating differences sit.

The opening is old. Wikipedia calls it arguably the oldest recorded opening, depicted in the Catalan poem Scachs d'amor from about 1475 in a fictional game in Valencia between Francesc de Castellvi and Narcis Vinyoles. Wikipedia says Jacques Mieses played it often and developed its theory greatly in the late 19th and early 20th centuries. In modern play Anand used it against Kasparov in the 1995 World Championship match and, in Wikipedia's words, obtained an excellent position, though Kasparov won the game. Carlsen beat Caruana with it at the 2014 Tromso Olympiad and drew him with it at the 2016 Olympiad. As of October 2026, Chess.com says it is Black's seventh most popular reply to 1.e4, though rankings like that move with the database.

The standard setup

ChessCompass gives the usual Black formation: ...Nf6, ...c6, ...Bf5 or ...Bg4, ...e6, ...Nbd7, ...Be7 or ...Bd6 and then castling short. The bishop leaves the pawn chain before ...e6 closes it. The early queen sortie typically costs one or two tempi, but White's central pawn is gone, so the time is harder to use. That is the bargain: Black accepts a small lead in development for White in return for a solid structure with no weaknesses.

Position A: after 3.Nc3, which queen move?

After 1.e4 d5 2.exd5 Qxd5 3.Nc3 the queen must move, and Black has 47 legal moves. The top ten all sit between -0.63 and -1.31 for Black, then the list falls off a cliff: the eleventh move, 3...Qc6, is -4.85, and 3...Nf6, which leaves the queen en prise to the knight, is -5.51. The table shows the five queen moves Maia3 rates likeliest, and they are the same five at every band.

Stockfish's order is Qa5 at -0.63, Qd6 at -0.75, Qd8 at -0.78, Qe5+ at -0.93, Qd7 at -1.05, Qe6+ at -1.09 and Qf5 at -1.20.

Three things stand out. First, the engine's choice is not the crowd's choice for most of the range. Qa5 climbs at every step, from 17.7 percent at 1000 to 53.5 at 2200, but it only becomes the likeliest move at 1800, where it edges Qd8 by 38.8 to 38.1. Second, Qd8 is the favorite from 1200 to 1600, peaking at 38.7 percent at 1600, and then loses ground to 30.7 at 2000 and 21.7 at 2200 as stronger players move to Qa5 and Qd6. Third, Qd6 is the slow riser: 4.9 percent at 1000 and 19.1 at 2200. At 2200 it is nearly level with Qd8, 19.1 to 21.7, which fits ChessCompass calling it the most popular modern choice.

Because Qa5 is Stockfish's best move, its findability is simply its probability: 17.7 percent at 1000, the lowest of any band, and 53.5 at 2200, the highest. The cost of the crowd move is small once you pass 1000. The gap between the likeliest human move and the engine's best is 15 centipawns at 1200, 1400 and 1600, where Qd8 leads, and 0 from 1800, where Qa5 does. That is the difference between -0.63 and -0.78, and it backs the sources calling Qd8 passive but playable: Wikipedia says it "remains playable, although it is considered somewhat passive," and ChessCompass says Black loses the most time with it but has no weaknesses, a solid Caro-Kann like game. If that sounds familiar, our Caro-Kann Defense rating data covers the other opening that Qd8 resembles.

The queen checks at 1000

The odd band is 1000. There the likeliest move is not a retreat at all but 3...Qe6+, at 24.0 percent, with 3...Qd8 second at 22.1, 3...Qe5+ third at 20.2 and Qa5 only fourth at 17.7. The gap to the engine's best at that band is 46 centipawns, the largest in either position we measured, because Qe6+ scores -1.09 against -0.63. Both checks fade with rating: Qe5+ goes from 20.2 to 13.2, 10.1, 8.3, 6.5, 4.6 and 3.3 percent, and Qe6+ from 24.0 to 21.1, 15.8, 10.6, 5.9, 3.0 and 1.8.

Are the checks actually bad? Wikipedia lists both as inferior third moves. After 3...Qe5+ 4.Be2 c6 5.Nf3 Qc7 6.d4, it says White has a handy lead in development. After 3...Qe6+ 4.Be2 Qg6 5.Nf3 Qxg2 6.Rg1 Qh3 7.d4, it says White's massive lead in development far outweighs Black's extra pawn. The picture is less one sided in opinion posts. One chess.com forum poster, writing about the Letterman Variation, calls the check "not great, but not terrible either," while also noting that the queen can get trapped after 4.Be2 Qg6 5.Bf3 Qxg2. That is forum opinion, not an authority, and we would not lean on it. The firmer anchor is the engine: -1.09 against -0.63 is a real cost, about half a pawn, but not a collapse.

For a 1000 rated player the practical point is that the check is tempting because it looks forcing, even though it develops nothing. Qa5 and Qd8 do the same job of saving the queen with a better score, and you can choose between them on taste.

Position B: after 3...Qd8 4.Nf3

The late position follows the single likeliest human move at the reference rating of 1400, two plies deeper. That is 3...Qd8, the 35.4 percent choice from the table above, and then 4.Nf3, which Maia gives 44.8 percent for White at 1400. Black has 31 legal moves. Stockfish puts 4...Nf6 first at -0.72, 4...e6 second at -0.73, 4...Bf5 third at -0.76 and 4...g6 fourth at -0.83. The three leaders are separated by a handful of centipawns.

Here the crowd and the engine agree at every band. 4...Nf6 is both Stockfish's best and Maia's likeliest move from 1000 to 2200, so the gap is 0 throughout, and its share climbs without a single dip, from 25.1 percent to 58.9. The top of the list is nearly flat, as it was in our Italian Game rating data: 4...e6 is one centipawn behind Nf6, yet it is the less popular of the two at every band and falls to 8.8 percent at 2200.

The lower band habits are easy to see. 4...Nc6 is the second likeliest move at 1000, at 23.8 percent, and then shrinks to 18.9, 14.0 and 9.8 before dropping out of the top five at 1800. At -0.98 against -0.72 it is a mild inaccuracy, not a blunder. 4...c6 goes the opposite way: it is outside the top five until 1800, then takes 9.0, 13.3 and 16.2 percent at 1800, 2000 and 2200. The stronger players are reaching for the pawn structure that goes with ...Bf5, even though the engine scores it at -0.85 and slightly prefers other moves.

Which queen move should you pick?

The numbers support a practical rule. At 1000, forget the checks and choose between Qa5 and Qd8. From 1200 to 1600, Qd8 is both the crowd move and only 15 centipawns off the engine's best, and the follow up positions are calm: develop ...Nf6, ...Bf5 or ...e6 and castle. From 1800, Qa5 is the likeliest move and the engine's best, but ChessBase says the plan has cost Black time and the queen often has to retreat to c7 after ...c6. ChessCompass adds the tactical watch points: b4 tricks and Bd2 hitting the queen. Qd6 avoids the tempo gaining Bd2, according to ChessCompass, and Wikipedia says that although it was once poorly regarded because of Nb5 or Bf4, numerous grandmaster games have since shown it to be quite playable, and that it has grown in popularity.

That makes the engine and the practical verdict diverge. Stockfish ranks Qa5 first, while sources describe Qd6 as perfectly playable. In our data Qd6 at -0.75 sits between Qa5 at -0.63 and Qd8 at -0.78, so the engine does not punish the choice much, and a structure you understand is worth more than the difference. Plans matter more than the exact move, which is the argument of why learning plans beats memorizing openings.

What the Scandinavian does to your results

As of October 2026 Chess.com says White wins 44 percent of master games against the Scandinavian, against 40 for the French and 37 for the Sicilian. Of the three, the Scandinavian gives White the highest score, which fits the engine's edge of a bit more than half a pawn. For another answer to 1.e4, our French Defense findability data shows how a different early commitment plays out across ratings. For the best moves in other positions, see our guide to the best chess moves by position.

As of October 2026, on the Lichess opening explorer, for ratings 1000 to 2500, White answers 1.e4 d5 with 2.exd5 in 67 percent of games, then e5 in 11, Nf3 in 8, Nc3 in 6 and d4 in 3. That is a useful reminder that the lines on this page assume White takes the pawn, which most do. Lichess's explorer remains better than this article for raw game counts, and Chess.com's opening page is better for a quick view of master win rates. What neither does is split the queen retreat by rating, or tell you what the engine thinks of each one, which is the gap this page fills.

Where other pages are better

This page is narrow on purpose. Wikipedia has the fuller variation tree and the better history. ChessBase's learning page explains the Qa5 plan and the queen retreat to c7 more clearly than we do. ChessCompass is better at tactical traps and at side lines such as the Portuguese Gambit, a pawn offer for fast development that it calls objectively not fully sound. Lichess is the place to go for how real games turn out at your level. What we do not measure is also worth saying: 2...Nf6 and every line after it, whether players recapture on d5 at all, White's third move alternatives, and any result data. Two positions are the entire scope of the measurement. The unmeasured 2...Nf6 line can also arise from 1.e4 Nf6 2.Nc3 d5 3.exd5 Nxd5, and our Alekhine Defense page gives an engine score for that move order.

What to take from it

The Scandinavian is a one decision opening for the first few moves, and the decision is which queen move to make after 3.Nc3. The engine prefers Qa5, the crowd of 1200 to 1600 players prefers Qd8, and 1000 players are drawn to checks that cost a third to half a pawn. None of the three sensible choices costs more than 15 centipawns against the best. Choose the one whose structure you will play for a hundred games, learn the usual setup of ...Nf6, ...c6 or ...e6, ...Bf5 and castling, and spend 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.