Ruy Lopez: What Every Rating Actually Plays After 3.Bb5

Open any Ruy Lopez guide and the first statistic you meet, on Wikipedia as of October 2026, is that 3...a6 turns up in about two thirds of all games. It is true, and it is also an average over every rating that ever played the opening, which is why it tells you very little about the player across from you. After 1.e4 e5 2.Nf3 Nc6 3.Bb5, a 1000 rated player and a 2200 rated player are choosing from the same list of legal moves and landing on different ones. 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 reply with Stockfish 18 at depth 20, see what the Stockfish chess engine is. The likeliest reply changes twice across those seven bands, so three different moves each lead somewhere. Maia3 predicts what a player at a given rating is likely to choose. It is not a count of real games.

One convention before the numbers. Every evaluation on this page is from White's side, so +0.29 means White is better by 0.29 of a pawn and -2.88 means Black is winning clearly. The figures in our table are Stockfish's score for the position after Black's move, restated that way.

One opening, several names

The Ruy Lopez is 1.e4 e5 2.Nf3 Nc6 3.Bb5. Wikipedia says it is named after Ruy Lopez de Segura, a 16th century Spanish priest, and it is also called the Spanish Opening or the Spanish Game. He wrote about it in his 1561 book, the Libro del Axedrez. Chess.com files the root position under the code C60, and Wikipedia gives the Berlin as C65 to C67. A search summary of Wikipedia also says the theory of the Ruy Lopez is the most extensively developed of all Open Games, with some lines analysed beyond move thirty. We could not open that page's passage ourselves, so treat the claim as reported, though nothing in our own work contradicts it.

What matters for a club player is that Black's third move names the system. Wikipedia lists 3...a6 as the Morphy Defence, 3...Nf6 as the Berlin Defence, 3...d6 as the Steinitz Defence, 3...Bc5 as the Cordel Defence, 3...f5 as the Jaenisch or Schliemann Gambit, and 3...Nd4 as Bird's Defence. Chess.com's opening page lists seven lines: Main Line, Closed, Berlin, Exchange, Open, Schliemann-Jaenisch Gambit and Marshall Attack. Some names are third moves and some are structures that arise later, after 3...a6 4.Ba4 Nf6 5.O-O.

After 3.Bb5, what Black actually reaches for

Stockfish's ranking of Black's third move is tight at the top and then falls away. Its choice is 3...Nf6 at +0.29, followed by 3...a6 at +0.35, 3...Bc5 at +0.50, 3...Nge7 at +0.52, 3...Nd4 at +0.54, 3...g6 at +0.59, 3...Be7 at +0.61, 3...f5 and 3...a5 at +0.68, and 3...d6 at +0.77. In the table, "out" means the move was outside Maia's top five at that band. It does not mean zero probability.

Read the rows as three leaders and two handovers. At 1000 and 1200 the likeliest reply is the Berlin, 28.2 and 23.6 percent, and it is also the engine's top move, so the gap between the crowd move and the best move is zero. At 1400 and 1600 the likeliest reply becomes 3...d6, 23.6 and 26.6 percent. That is the weakest of the top five by the engine, at +0.77, 48 centipawns behind the best move. Only 18.8 percent at 1400 and 15.8 at 1600 play the engine's 3...Nf6. From 1800 the Morphy takes over, 27.6 percent at 1800, 39.0 at 2000 and 56.2 at 2200, at +0.35. The likeliest move is then six centipawns from the best, and 3...Nf6 sits near 15 percent from 1800 up, at 14.8, 15.5 and 15.2.

Two details stand out. First, the Morphy is not absent at low ratings: 3...a6 is the second likeliest reply at 1000, 1200, 1400 and 1600, at 23.3, 20.4, 20.6 and 22.8 percent. It never disappears, it just stops being the likeliest reply until 1800. Second, the Old Steinitz rises and falls like a hump: 11.2 percent at 1000, a peak of 26.6 at 1600, 14.1 at 2000, and out of Maia's top five at 2200. That shape is what the two thirds average cannot show, since the strong players who dominate game databases play 3...a6 and wash the rest out. Our aggregate figures say the same thing from the other side: the top five replies cover 80.5 percent of Maia's probability at 1000 and 91.8 at 2200, so the field narrows as ratings rise.

The Cordel, 3...Bc5, is a steady minority choice, from 10.9 percent at 1000 to 15.4 at 2000 and 9.8 at 2200. 3...Nd4 appears in the top five only at 1000 and 1200, with 6.9 and 7.2 percent. 3...Nge7 enters at 1400 with 7.1 percent and fades to 4.4 at 2200. The Schliemann, 3...f5, shows up only at 2200, with 6.2 percent.

What this means if you play White

The practical reading is a preparation order, not a prediction. If you play White around 1000 to 1200, expect the Berlin first and the Morphy second, so those two lines deserve your time before anything exotic. From 1400 to 1600 the crowd answer is the Old Steinitz, and that is good news, because of the five likeliest replies it is the one that scores best for White. 4.d4 is +0.73 in a local Stockfish check at depth 22, with the engine's own continuation being 4...exd4 5.Nxd4 Bd7 6.Nc3 Nf6 7.Bxc6, and 3...d6 itself is +0.77 in the dataset. You get a small, stable edge. From 1800 up you need the Morphy, which means the theory the opening is famous for.

The 4.Bxc6 Exchange Variation is a legitimate alternative if you would rather not memorize. After 3...a6 4.Bxc6 dxc6 the position is +0.17 for White in our depth 22 check, with 5.O-O the engine's choice, so the structure is level in engine terms, not a failure. The forum reads it as the friendly entry: FrogCDE writes that "there's a lot to be said for starting with the exchange variation, as favoured by Fischer at one time."

One late position is trivial. After 3...d6 4.Bxc6+ the recapture 4...bxc6 is played at 97.1 to 99.4 percent at every band, and it is the engine's choice, at +0.32.

What this means if you play Black

If you want the engine's choice with the least reading, play the Berlin. 3...Nf6 is first by the engine at +0.29, and it keeps the position simple to find: the top of the list is a developing move, not a theory exam. The Open Berlin is 4.O-O Nxe4 and then 5.d4 or 5.Re1, and only 5.d4 leads to one of the most famous endgames in chess. After 5.d4 Nd6 6.Bxc6 dxc6 7.dxe5 Nf5 8.Qxd8+ Kxd8, our depth 22 check gives +0.22 for White, which is a draw to most eyes, even though Black's king has lost castling rights. 5.Re1, which avoids that endgame, is +0.19 with Black to move. Wikipedia says Kramnik used the Berlin as a drawing weapon against Kasparov at the 2000 Classical World Championship, drawing games 1, 3, 9 and 13, and quotes him saying "This was all new to Kasparov." The line became popular in the 21st century, and Wikipedia nicknames it the Berlin Wall.

As of October 2026, Chess.com recommends the Berlin as a solid choice for Black and reports White 33 percent, draws 44 percent and Black 22 percent for it. Hold that figure loosely, since the page states no game pool or rating band.

3...a6 is the theory heavy choice. At +0.35 it costs six centipawns against the Berlin, which is nothing, and it keeps more of the middlegame alive. The price is memory. The traditional Closed main line is 3...a6 4.Ba4 Nf6 5.O-O Be7 6.Re1 b5 7.Bb3 d6 8.c3 O-O 9.h3, from which the Chigorin, Breyer, Zaitsev and other systems branch. After 8...O-O our check gives +0.39 for White, with 9.h3 the engine's pick, so Black is well within reach of equality. Chess.com recommends the Arkhangelsk for tactical players, and Chessable names it as 6...Bc5. Chessable also warns that "at club level, the Marshall Attack can be a dangerous weapon, and White should know how to deal with it." The Marshall is 7...O-O 8.c3 d5, a pawn sacrifice for active counterplay. We did not measure it, and no page we read gives a rating split for it.

The named defences, with engine numbers

The Old Steinitz, 3...d6, is the quiet, solid reply and the weakest in the engine's top group. The cost is real but small: 0.77 against 0.29 is about half a pawn, and a half pawn is a slow squeeze, not an attack. It is the clearest case in our data of the likeliest human move sitting far from the best one.

The Cordel, 3...Bc5, scores +0.50. It is playable, a little less accurate than the Berlin or the Morphy, and never the likeliest reply.

Bird's Defence, 3...Nd4, is +0.54 in the dataset. The line is 4.Nxd4 exd4, and our depth 22 check there reads +0.62 for White with 5.O-O the engine's pick. Maia3 places it in the top five only at 1000 and 1200, at 6.9 and 7.2 percent. Falling_Fast's forum advice to learn "those 4 defenses for black," meaning the Morphy, Berlin, Steinitz and Bird, is a sensible short list for a player meeting the Bird at 1000 or 1200, and the engine does not punish the Bird heavily.

The Schliemann or Jaenisch Gambit, 3...f5, is +0.68 in the dataset and +0.61 in the depth 22 check with 4.d3 as the engine's reply. It is the sharpest option on the list, a gambit by name, and it appears in Maia's top five only at 2200, at 6.2 percent.

Two traps worth knowing

The Noah's Ark Trap is a family of ideas in which a white bishop is trapped on b3 by black pawns. In the move order Wikipedia gives, the moves run 3...a6 4.Ba4 d6 5.d4 b5 6.Bb3 Nxd4 7.Nxd4 exd4 8.Qxd4?? c5 and the queen has to move again, leaving the bishop on b3 with no retreat: 9.Qd5 Be6 10.Qc6+ Bd7 11.Qd5 c4. In our depth 22 check, the position after 8...c5 is -2.88 for White, with Black clearly winning, and it is -3.05 after 11...c4. The bishop is lost for two pawns and still behind. Wikipedia lists 5.c3, 5.Bxc6+ and 5.O-O as better fifth moves for White, and 8.Bd5 or 8.c3 as better eighth moves. A search summary credits an Endre Steiner against Capablanca game at Budapest in 1929 as a historic example, but we could not open that page, so we cannot confirm it.

The second is a trap that does not work for White. After 3...a6 4.Bxc6 dxc6, it is tempting to grab 5.Nxe5, since the e5 pawn looks undefended. After 5...Qd4 the fork on the knight and the e4 pawn leaves White at -0.62 in the depth 22 check. The engine's best is 6.Nf3 Qxe4+ 7.Qe2 Qxe2+ 8.Kxe2, so Black regains the pawn and has a slightly better ending. 4.Bxc6 does not win a pawn yet.

Should a club player learn it at all?

The forum answer is a split, and the data does not settle it. ponz111 writes that "a 1500 rated player needs to study all aspects of chess and thus should not be bogged down on one particular opening." Ethan_Brollier says "there is far, far too much to remember" and suggests the Scotch or the Italian, adding to avoid theory until about 1050. If you take that route, the Scotch Game by rating shows which reply to 4.Nxd4 your band is likeliest to meet. On the other side, pfren calls the Spanish "extemely versatile," with many setups and pawn structures, and "an ideal school on positional chess." tygxc's answer is "Just play it. Think carefully. When you lose, analyse."

Chessable, for its part, puts White's win rate at 40 percent, a rate it calls impressive, though the page gives no rating split. Our contribution is smaller and more concrete. The Ruy Lopez is not one opening at club level. It is a Berlin at 1000 to 1200, an Old Steinitz at 1400 to 1600 and a Morphy past 1800, and the right amount of study is the line you will actually meet. If you want to compare it with a quieter 1.e4 e5 choice built on the same dataset, our Italian Game page does the same exercise for 3.Bc4, where 3...Bc5 and 3...Nf6 are nearly level. For a 1.d4 comparison, the Queen's Gambit Accepted page applies the same method. The board for this opening is on the Ruy Lopez opening page, and for the general question of when a move is actually good, see how to find the best chess moves.

Where other pages are better

This page is narrow on purpose. Our data covers Black's third move after 3.Bb5, at seven ratings, and the late position after 3...d6 4.Bxc6+. Everything past those points comes from a handful of local Stockfish checks and from other writers, and it is not a database. If you want move frequencies from millions of games, a master game explorer on chess.com or lichess is the right tool, and no figure on this page replaces it. Wikipedia is the stronger reference for the history, the full list of named variations and the theory. Chessable is better for step by step courses on the Closed lines and the Marshall. None of them splits Black's reply by rating, which is the gap this page fills, and none shows an engine score for every third move.

We also do not measure White's fourth move or results by rating.

What to take from it

The reply you meet after 3.Bb5 depends on who you are playing. At 1000 and 1200 it is likeliest the Berlin, which is also the engine's top move. At 1400 and 1600 it is likeliest the Old Steinitz, the engine's weakest of the five. From 1800 it is the Morphy, and by 2200 it is more than half of Maia's probability. As White, prepare in that order for your band. As Black, the Berlin is the engine's choice and simple to find, and the Morphy is the choice with the deepest theory. Neither is wrong, and the engine will not decide it for you.

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.