Playing11 min read

Reti Opening: Players Most Often Take on c4, Then Slip With ...b5

The pages we opened give the Reti a verdict in one line: 2...dxc4 is discouraged, 3...b5 is dubious. None shows how big the edge is or what players actually play. Chessdrive's Maia3 and Stockfish data covers two positions at seven rating bands. After 1.Nf3 d5 2.c4, the engine's top reply 2...d4 (+0.09) is predicted at 9.1 percent at 1000, while 2...dxc4 leads up to 1800. After 3.e3, below 1800 the likeliest move is 3...b5, which costs Black about half a pawn. We check the refutation, the forum claim and the 2...d4 line.

The starting position

The Reti Opening is a hypermodern way to start the game as White: develop the knight, play c4, and fianchetto a bishop instead of building a pawn centre at once. The name has two meanings, and the pages you will find disagree about which one is standard. Wikipedia says that in modern usage it means 1.Nf3 d5 2.c4, ECO A09, and that 1.Nf3 alone is the Zukertort Opening. As of October 2026, chess.com and Duolingo call plain 1.Nf3 the Reti, and chess.com calls 1.Nf3 d5 2.c4 the Reti Gambit. This page covers 1.Nf3 d5 2.c4, the narrower meaning, because that is where the opening makes real demands on Black.

Every page we read gives a verdict in a line: 2...dxc4 is discouraged, 3...b5 is dubious. None shows how big the edge is or which moves players at 1000 or 2000 actually choose. We ran the maia3-5m checkpoint (see how Maia chess actually works) on two Reti positions at seven rating bands from 1000 to 2200, and graded every legal move with Stockfish 18 at depth 20. Percentages in the tables are the model's predicted move probabilities for a player at that rating. They are not game database frequencies and not win rates.

Every evaluation is from White's side, so +0.29 means White is better by 0.29 pawns. The two dataset positions are the one after 1.Nf3 d5 2.c4 and the one after 1.Nf3 d5 2.c4 dxc4 3.e3, both with Black to move. Our line checks are separate local Stockfish runs at about depth 20 to 22, called local checks below.

Where the Reti comes from

Richard Reti was born on 28 May 1889 in Bazin, now Pezinok, and died of scarlet fever in Prague on 6 June 1929. With Aron Nimzowitsch he was a leading voice of hypermodernism, and his book Modern Ideas in Chess appeared in 1923. Wikipedia says Reti dated the opening to early 1923 and gives his game against Rubinstein at Karlovy Vary that year as its origin. At the time almost any game that began with Nf3 and c4 was called a Reti. Hans Kmoch named the attacking system the Reti Opening or Reti System, and Tartakower called it the Reti and Zukertort Opening.

The fame came in 1924. At New York, Reti beat the world champion Capablanca with 1.Nf3 in round 5, in 31 moves, ending an unbeaten run that went back to January 1916. ChessBase says Capablanca equalised without much trouble in the opening and lost the thread in the middlegame. One caution: chessgames.com files that game under ECO A15, an English with a King's Indian formation, so it was not a 1.Nf3 d5 2.c4 game. Chess.com says Reti won a famous game against Bogoljubov in the same tournament with 1.Nf3 d5 2.c4, but we could not obtain the move list.

Move 2: the engine wants 2...d4, humans take on c4

After 1.Nf3 d5 2.c4, Stockfish ranks 2...d4 first at +0.09, then 2...dxc4 at +0.29, 2...e6 at +0.34 and 2...c6 at +0.35. Those four are within about a quarter of a pawn of each other. The next best, 2...Be6, is already at +0.77. So Black has four good answers, and the question is which one players choose.

Rating1000120014001600180020002200
2...dxc4 (+0.29)42.4%44.7%43.6%39.6%32.0%21.1%11.9%
2...d4 (+0.09)9.1%12.7%16.6%20.1%22.9%23.4%21.4%
2...c6 (+0.35)10.0%9.8%10.0%11.1%14.7%22.1%30.0%
2...e6 (+0.34)10.6%10.0%10.1%11.0%13.1%17.4%24.6%
2...Nf6 (+0.92)17.9%13.9%11.4%10.3%10.2%11.0%9.5%

The shape is clear. From 1000 to 1800 the likeliest reply is 2...dxc4, at 42.4 percent at 1000 and still 32.0 percent at 1800. The engine's top move 2...d4 is predicted at 9.1 percent at 1000 and climbs to 23.4 percent at 2000, where it narrowly leads. At 2200 the lead moves to the solid 2...c6 at 30.0 percent, with 2...e6 second at 24.6 percent and 2...dxc4 down to 11.9 percent. Stronger players stop grabbing the pawn and hold d5 instead.

Is taking on c4 a mistake? Not much of one. The gap between 2...dxc4 and the best move is 0.20 pawns. Wikipedia says White regains the pawn with a slight edge, and chess.com calls 2...dxc4 generally discouraged. Our data sides with Wikipedia: a 0.20 gap does not discourage anyone, and the players who take are not losing much by it. The one reply that does cost something is 2...Nf6 at +0.92, which is the second most likely move at 1000 (17.9 percent) and fades with rating.

The Reti is also rare below master level. As of October 2026, the Chessiverse tables, built from Lichess blitz and rapid games, show 7.1 million games with 1.Nf3 d5 2.c4, and the line is 0.05 percent of games at 1000. At club level you will rarely meet it across the board.

After 2...dxc4 3.e3: the findability picture

On the most likely human path, 2...dxc4 is followed by 3.e3, the move White plays to regain the pawn with Bxc4. A local check says the pawn comes back comfortably: 3.e3 scores +0.40, 3.e4 +0.34 and 3.Nc3 +0.25. In the dataset, 2...dxc4 3.e3 is also the likeliest path at 1400, at 43.6 percent and then 37.9 percent.

Now Black has to decide whether to give the pawn back or hold it. The top engine moves are all about the same: 3...c5 at +0.27, 3...e6 at +0.28, 3...a6 at +0.29 and 3...Nf6 at +0.30. A player who finds any of those is fine. So the question is not whether Black finds one precise move. It is whether Black avoids the one tempting error.

Rating1000120014001600180020002200
3...b5 (+0.73)26.0%30.0%28.5%24.9%21.2%18.5%14.9%
3...Nf6 (+0.30)13.7%14.0%16.2%19.7%23.6%27.8%33.2%
3...Nc6 (+0.62)14.4%12.5%12.7%13.2%13.1%11.9%10.1%
3...Be6 (+0.34)11.9%14.1%12.3%9.0%not top fivenot top five6.9%
3...e6 (+0.28)not top fivenot top fivenot top fivenot top five8.9%7.8%6.7%

From 1000 to 1600, the single likeliest move is 3...b5, at 26.0, 30.0, 28.5 and 24.9 percent. It tries to keep the extra pawn, and the engine rates it +0.73 for White against +0.27 for the best move, a loss of 0.46 pawns. From 1800, 3...Nf6 at +0.30 takes over, and it is the favourite at 33.2 percent by 2200. The engine's literal best move, 3...c5, is predicted at only 0.9 percent at 1000, 1.3 percent at 1600 and 6.1 percent at 2200. That low number is easy to misread. Missing 3...c5 costs nothing, because 3...e6, 3...a6 and 3...Nf6 are within three centipawns of it. The rating-linked slip is 3...b5, and it fades from 30.0 percent at 1200 to 14.9 percent at 2200.

Refuting 3...b5

Wikipedia says holding the pawn with 3...b5 is dubious and that 4.a4 is superior for White, with no further detail. Local Stockfish at about depth 22 puts 4.a4 at +0.80 for White, the best move by a wide margin. The idea is simple. The a4 pawn attacks the b5 pawn, and Black cannot support it without weakening something. The engine line runs 4.a4 b4 5.Bxc4 e6 6.O-O Nf6 7.d3 Be7, and White is better with a lead in development.

The common human reply is 4...c6 5.axb5 cxb5. Here the engine's top move is 6.Nc3 at +1.55 for White, ahead of 6.Na3 at +1.27 and 6.Nd4 at +1.06. In the main line 6.Nc3 Qb6 7.b3 b4 8.Na4 Qb7 9.bxc4 Nf6 10.d4 e6, White has taken back the pawn and Black's queenside is loose. That is a bigger edge than the +0.73 the dataset gives 3...b5 itself, so 4...c6 makes Black's problems worse, not better.

A chess.com forum user recommends 6.b3 as the strongest move here, claiming Black must give back the pawn. That is a single user post, not an authority, and the engine does not agree that it is best. After 6.b3 the best Black reply is 6...Nd7 at +0.83 for White, with 6...Nf6 at +0.98 and 6...e6 at +1.03. Those are fine scores, but 6.Nc3 at +1.55 is nearly three quarters of a pawn better. We did not test the forum's 6...Be6 line.

The other lesson is a warning for White. After 3...b5, the natural looking 4.b3 lets Black off. In a local check, 4.b3 cxb3 5.axb3 c6 is about equal, with the evaluation at 0.05 in Black's favour. Do not mix up the move order. Against 3...b5, play 4.a4.

Other third moves for Black, and White's answers

If Black avoids ...b5, White does not have a forced edge. After 3...Be6, local Stockfish prefers 4.Ng5 at +0.32 for White, with 4.Nc3 and 4.Na3 at +0.30. After 3...Nf6 4.Bxc4, the engine's first line is 4...a6 at +0.42 for White, with 4...e6 at +0.39 and 4...c5 at +0.41, all within a few centipawns. All of these are small advantages for White, the kind you get from a bishop on c4 and a tempo or two. Wikipedia compares the whole structure to the Queen's Gambit Accepted, and the comparison is fair: White gets the pawn back and plays for a slightly freer game. If you want that structure with a pawn on d4 at once, compare it with the Queen's Gambit page.

Wikipedia also lists 3.Na3 and 3.Qa4+ as common alternatives to 3.e3. Chess.com says the Na3 plan works because White regains the pawn with Nxc4. We did not run the Qa4+ idea, and no source we read claims a trap in it, so we leave it there.

2...d4: the reversed Blumenfeld

The engine's top reply is the advance 2...d4, and the best human players choose it often enough that you should have an answer. Sources disagree about what to play. Wikipedia gives 3.e3 or the sharper 3.b4, which resembles a Benko Gambit with colours reversed, because Black has no pawn on c5. Chess.com recommends 3.g3, planning 4.Bg2, and as of October 2026 calls 2...d4 Black's most successful reply. Its suggested Black setups are 3...c5 or 3...Nc6. Chessiverse names a sideline, the Reversed Blumenfeld Gambit: 3.e3 c5 4.b4 dxe3 5.fxe3 cxb4 6.d4, which, by Chessiverse's estimate as of October 2026, scores about 52 percent for White.

Our local check ranks the three third moves differently. 3.b4 is first at +0.01 for White, which is equal. 3.e3 comes out at 0.09 in Black's favour, and 3.g3 at 0.24 in Black's favour. So Wikipedia's sharper choice is the engine's choice, and the quieter 3.g3 is the one that gives Black a small edge. If you play the Reti, 3.b4 deserves a place in your repertoire. The same 2...d4 structure appeared at the top: in Game 11 of the 2024 World Championship, Gukesh played 1.Nf3 d5 2.c4 against Ding Liren, who answered 2...d4. Gukesh replied 3.b4, the engine's choice, and after 3...c5 4.e3 the game reached a reversed Blumenfeld. Gukesh won to lead 6 to 5. Commentator Kidambi called 9.d3 dubious and preferred 9.c5, and Gukesh said he mixed up his lines, so treat the game as proof the opening is playable at the top, not as a model of the best play.

The solid replies: 2...e6 and 2...c6

Black can also keep d5 and transpose. Wikipedia says 2...e6 3.d4 reaches a Queen's Gambit Declined, and 2...e6 3.g3 Nf6 is the Neo-Catalan, with 4.Bg2 Be7 5.d4 a Closed Catalan or 4...dxc4. After 2...c6, 3.d4 is a Slav, and 3.e3 Nf6 4.d4 is a Slow Slav, with 4.Nc3 e6 5.d4 the Semi-Slav.

The engine confirms that these are small edges, not problems. After 2...e6, 3.d4 is best at +0.33 for White, with 3.g3 at +0.26. After 2...c6, 3.d4 is best at +0.34, then 3.e3 at +0.32. These are also the replies that stronger players choose. At 2200, 2...c6 and 2...e6 together account for 54.6 percent of predicted moves. So if you are White and you reach 2000 or 2200, expect a queen's pawn structure by transposition, not a theoretical gambit. Learning the Slav and Catalan plans from the White side is worth more than another opening line.

Who should play it, by rating

Our data has no game results by rating for the Reti, so this is what the model and the sources suggest, not a result table.

  • Around 1000 to 1400, as White. The surprise is real. 2...dxc4 is the favourite at 42.4 to 44.7 percent, and 3...b5 follows at 26.0 to 30.0 percent. Learn 3.e3 and the refutation 4.a4, and do not play 4.b3.
  • Around 1600 to 1800, as White. 2...d4 climbs to 20.1 and 22.9 percent. Have 3.b4 ready, since the engine scores it equal and it leaves Black the harder choices.
  • 2000 and above. 2...dxc4 falls to 21.1 percent and then 11.9 percent at 2200. You are playing a structure from the Slav or the Catalan, so know those plans.
  • As Black, at any rating. Pick one of the four equal replies. If you take on c4, answer 3.e3 with 3...c5, 3...e6, 3...a6 or 3...Nf6, not 3...b5.

For the same idea against a different first move, the Nimzo-Larsen Attack page covers the other flank opening with a fianchettoed bishop, and the Grob Opening page covers a flank try that is objectively worse. For how all of this fits with other openings by strength, see chess openings by rating.

Where other resources are better

This page is narrow on purpose. The dataset holds two positions, and Maia3 5m is a small model of human choice, so its percentages are predictions. Our line checks are quick local runs, not a repertoire book. Wikipedia is the best place for the history and the named lines, and it is where we found the Kmoch and Tartakower naming. Chess.com's opening page is a good plain tour. Chessiverse has real Lichess results by rating band, which we have not repeated here. We did not obtain a full game score for Reti against Capablanca, Bogoljubov or Rubinstein. We also found no source that says Kramnik or Carlsen plays it, so we make no claim either way. Several pages in the search results were unreadable, and we did not open the Lichess explorer. None of the pages we did read ties each reply to an engine score and to the rating curve, which is the gap this one fills.

What to take from it

The Reti can mean 1.Nf3 alone or 1.Nf3 d5 2.c4, and this page is about the second. After 2.c4, the engine's best reply 2...d4 is only +0.09 for White, but it is predicted at 9.1 percent at 1000. Players take on c4 instead, at 42.4 percent at 1000 and 32.0 percent at 1800, and that costs a mere 0.20 pawns. The real slip arrives on move 3: after 2...dxc4 3.e3, 3...b5 is the favourite from 1000 to 1600, it rates +0.73 against +0.27 for the best move, and 4.a4 punishes it. Three other moves are within three centipawns of the best one, so missing 3...c5 does not matter. Keep every percentage for what it is: the prediction of a model, not a count of real games.

Frequently asked

What is the Reti Opening?
It depends on the source. Wikipedia says that in modern usage the Reti means 1.Nf3 d5 2.c4, with White fianchettoing a bishop and avoiding an early d4, and it files that line as ECO A09. It calls 1.Nf3 on its own the Zukertort Opening. As of October 2026, chess.com calls 1.Nf3 the Reti and 1.Nf3 d5 2.c4 the Reti Gambit. This page covers 1.Nf3 d5 2.c4.
Is the Reti Opening good?
Yes, it is sound. In our Stockfish data, after 2.c4 Black's best reply 2...d4 leaves White ahead by only 0.09 pawns, so it is close to equal, and the three other main replies (2...dxc4, 2...e6, 2...c6) are within 0.26 pawns of it. It is a way to reach a flexible position, not a way to win material by force. Wikipedia calls it hypermodern, because White does not occupy the centre with pawns.
Is the Reti Gambit a real gambit?
Only loosely. Some sites, chess.com among them, call 1.Nf3 d5 2.c4 the Reti Gambit. But White can regain the c4 pawn after 2...dxc4. Our local Stockfish check at about depth 20 gives 3.e3 at +0.40 for White and 3.e4 at +0.34, so the pawn comes back comfortably. Treat the name as a label, not as a promise of compensation.
What should White do after 2...dxc4?
Play 3.e3 or 3.e4 and take the pawn back. Wikipedia gives both with a slight edge for White, and our local check agrees: 3.e3 scores +0.40 and 3.e4 +0.34. Wikipedia also lists 3.Na3 and 3.Qa4+ as common, and chess.com adds that White regains with 3.Na3 and 4.Nxc4. In our check 3.Nc3 scored +0.25.
How do you refute 3...b5?
With 4.a4. Wikipedia calls 3...b5 dubious and 4.a4 superior for White, and local Stockfish at about depth 22 puts 4.a4 at +0.80. After 4.a4 c6 5.axb5 cxb5 the top move is 6.Nc3 at +1.55, then 6.Na3 at +1.27. Avoid 4.b3: after 4.b3 cxb3 5.axb3 c6 the position is about equal.
What is the best reply to the Reti?
By the engine, 2...d4 at +0.09 for White, with 2...dxc4 at +0.29, 2...e6 at +0.34 and 2...c6 at +0.35. As of October 2026, chess.com calls 2...d4 popular and Black's most successful reply. Humans below 2000 prefer 2...dxc4. Players around 2200 prefer 2...c6 at 30.0 percent, and 2...e6 at 24.6 percent.
Did Gukesh play the Reti?
Yes. Gukesh played 1.Nf3 d5 2.c4 in Game 11 of the 2024 World Championship against Ding Liren, who answered 2...d4, a reversed Blumenfeld Gambit. Gukesh won and took a 6 to 5 lead. Commentator Kidambi called 9.d3 dubious, preferring 9.c5, and Gukesh said he mixed up his lines. We could not find a source for any other world champion using it in a title match.

Want to face 2.c4 against a model that answers like your rating?

Chessdrive runs the real maia3-5m checkpoint on-device as your opponent at any rating you set from 600 to 2600, so you can play the Reti from either side against a model that picks its moves the way your band does, then see where the game turned with a separate Stockfish review. Available now on iOS.

Keep reading

All guides

Updated