Nimzo-Indian Defense: What Each Rating Plays After 3...Bb4, and the 5.e4 Slip

The Nimzo-Indian Defense is the opening where Black pins the knight before White has built a centre and then decides what to trade. After 1.d4 Nf6 2.c4 e6 3.Nc3 Bb4, White cannot play e4 at once, and the whole game is how White gets rid of the pin and what it costs. The pages we opened list the named variations and say little about which of them people at your level will actually face. We ran the maia3-5m checkpoint, see how Maia chess actually works, on two positions at seven rating bands from 1000 to 2200 and graded every legal move with Stockfish 18 at depth 20. Percentages on this page are Maia3's predicted probability that a player at that rating chooses the move. They are model predictions of human play, not game counts or win rates.

Every evaluation is from White's side, so +0.24 means White is slightly better by about a quarter of a pawn. The two dataset positions are the one after 3...Bb4 with White to move, and the one after 4.Bd2 O-O with White to move. The line checks are separate local runs of ten seconds per line at depths 23 to 26, which we call our local Stockfish check, so expect a position to differ by several hundredths between methods.

What the Nimzo-Indian is, and where it came from

The defining moves are 1.d4 Nf6 2.c4 e6 3.Nc3 Bb4, which covers ECO E20 to E59. Move orders such as 1.c4 e6 2.Nc3 Nf6 3.d4 Bb4 reach the same position. The bishop pins the c3 knight, which stops 4.e4 and removes a key defender of the centre. Black often gives up the bishop pair with ...Bxc3 to double White's c-pawns, and develops quickly without committing the pawn structure.

It is named after Aron Nimzowitsch, the hypermodern player who popularised it at master level in the early 20th century. He did not invent it: Wikipedia lists Englisch against Blackburne, London 1883, as the earliest recorded game, and chess.com says he was not the first master to play it but was perhaps the most important in spreading it. It first appeared in world championship play in the 1929 Alekhine against Bogoljubov match.

The first decision: what White plays after 3...Bb4

This is the position at the heart of the first table. Stockfish ranks 4.Qc2 first at +0.24, with 4.Nf3 and 4.e3 both at +0.21, so the top three sit within three hundredths of a pawn. 4.Bd2 is +0.09, 4.a3 is -0.02, and 4.Bg5 is -0.16. Here is what Maia3 predicts people play.

In the table, "out" means the move was outside Maia's top five at that band. It does not mean zero probability.

The headline is the ladder. At 1000 the likeliest move is 4.Bd2 at 38.5 percent, nearly three times the 14.1 percent for the next move, 4.Nf3. It stays first at 1200 and 1400, at 31.8 and 23.5 percent, then slips to third at 1600 and fourth at 1800. 4.Bg5 takes over at 1600 and 1800, at 21.9 and 21.0 percent. The engine's own favourite, 4.Qc2, is only 4.2 percent at 1000 and does not become the likeliest move until 2000, at 23.3 percent, rising to 28.0 at 2200.

Read this as a story about ideas, not accuracy. 4.Bd2 prepares to recapture on c3 with the bishop, avoiding doubled pawns, and costs about 0.15 of a pawn against the best move, so nobody at 1000 is making an error. 4.Bg5 at -0.16 is the move with the biggest cost in the table, about 0.40 of a pawn off the engine's top choice, and it is the likeliest move at 1600 and 1800. That is a middling gap, and Black's reply in our check supports it: after 4.Bg5, Black's best is 4...c5 at -0.18 (White POV, so Black is slightly better), with 4...h6 at -0.07 and 4...b6 at +0.14.

The Rubinstein tells a different story. 4.e3 is 6.6 percent at 1000 and out of the top five from 1200 to 1800, then climbs back to 11.7 percent at 2000 and 19.7 at 2200, where it is second to 4.Qc2. At the top of our range, the two principled moves split the field. See also chess openings by rating.

The Rubinstein System: 4.e3

The main line runs 4...O-O 5.Bd3 d5 6.Nf3 c5 7.O-O, and Black's three standard replies are 4...O-O, 4...c5 and 4...b6. 4...d5 allows the Botvinnik Variation with 5.a3, and 4...Nc6 is the Taimanov. Our check puts the three main replies in a narrow band.

• 4...O-O. The engine's top move for White is 5.Nf3 at +0.24, then 5.Bd2 at +0.18 and 5.Bd3 at +0.10, all at depth 25.
• 4...c5. White's best is 5.Ne2 at +0.30, with 5.Nf3 at +0.24 and 5.Bd3 at +0.21.
• 4...b6, the Fischer Variation. White's best is 5.Ne2 at +0.37, with 5.Bd3 at +0.31 and 5.Nf3 at +0.21. This is the highest of the three, though still well under half a pawn. 4...b6 is also Nimzowitsch's own favourite, in accordance with the spirit of the opening, and it aims for ...Ba6 after ...d5.

The Classical Variation: 4.Qc2

The Capablanca line lets White recapture on c3 with the queen, so the pawn structure survives and Black cannot double the pawns. The main line is 4...O-O 5.a3 Bxc3+ 6.Qxc3 b6. The drawback is that the queen often moves twice and the kingside lags. Revived in the 1990s, it is, as of October 2026, described by Wikipedia as just as popular as the Rubinstein.

• 4...O-O. White's best is 5.e4 at +0.35 at depth 26, ahead of 5.Nf3 and 5.a3, both +0.14. Compare this with the 5.e4 slip below: with the queen on c2, the pawn push works.
• 4...d5. White's best is 5.cxd5 at +0.19, then 5.a3 at +0.07 and 5.e3 at +0.01.
• 4...c5. White's best is 5.dxc5 at +0.27, then 5.a3 at +0.07 and 5.e3 at -0.01.

As of October 2026, a Chessiverse table for this exact position (not the whole Nimzo) covers 2,484,888 Lichess blitz and rapid games. 4...O-O is Black's top reply at every band, and White scores 55.3 percent at 400 and 46.8 at 2500. That is a different population from the master-level figures, so we do not merge them.

The late position: where 5.e4 loses a pawn

The second dataset position follows the single most popular human continuation at 1400, 4.Bd2 O-O, with White to move. Stockfish has 5.e3 first at +0.14, with 5.Nf3 at +0.13, 5.Rc1 at +0.09 and 5.Qc2 at +0.04. The move to note is 5.e4, which scores -1.09 in the dataset. It looks natural, because White seems ready to recapture on c3 with the bishop, but the e4 pawn has lost its defender and it drops a full pawn.

This is the headline of the page. 5.a3 is the likeliest move from 1000 to 1600, at 28.1 percent at 1000 and 33.2 at 1400, then 5.Nf3 takes over from 1800 and reaches 44.5 percent at 2200. 5.a3 is nearly equal at -0.01, about 0.15 of a pawn from the best, so it is a fine move. The gap is in 5.e4: it is 17.6 percent at 1000, more common there than the engine's own first choice 5.e3 at 14.8, then 14.9, 12.2, 9.5 and 6.6 percent at 1800, and it leaves the top five by 2000. At 1000, roughly one White player in six who reaches this position makes it.

Our local check at depth 26 confirms the cost. After 4.Bd2 O-O 5.e4, Black's best is 5...Bxc3, and after 6.bxc3 Nxe4 Black has won the e4 pawn, with White at -1.08. 5...e5 is +0.00 and 5...d5 is +0.10, so those other replies are fine for White, but the capture is the move to find. If you play Black at 1400 or below and White pushes e4 at once, expect it in roughly one in eight to one in six of the games that reach this position, and know the answer: take on c3, then take on e4.

If you play White, 5.a3 is the safe move. After 5.a3 Bxc3 6.Bxc3, our check gives Black 6...Ne4 as best, at +0.03, with 6...b6 at +0.15 and 6...d5 at +0.25. That is the quiet, equal line, and 5.a3 is the choice of about a third of 1400-rated players in this position, while 5.e4 is the clearest rating-linked slip in the data.

The other fourth moves: Samisch, Kasparov and the rest

Three more choices come up.

• The Samisch, 4.a3 Bxc3+ 5.bxc3. White gets a strong centre and kingside play, and Black blockades with ...c5, then hits c4 with ...b6, ...Ba6, ...Nc6-a5 and ...Rc8. Botvinnik used it five times against Tal in 1960, all draws, and once in the 1961 rematch, a win. As of October 2026, it is rare at the top. Our check, with Black to move, gives 5...c5 at -0.07 (White POV), and 5...Nc6 and 5...b6 both at -0.02. Black is fine in all three.
• The Leningrad, 4.Bg5. Black answers 4...h6 or 4...c5, and after 5.Bh4 c5 6.d5 the main line is 6...d6 7.e3 Bxc3+ 8.bxc3 e5. 6...b5 was popular in the 1970s after Tal beat Spassky at Tallinn 1973. The numbers are the ones above: 4...c5 at -0.18, 4...h6 at -0.07.
• The Kasparov Variation, 4.Nf3. Kasparov played it six times against Karpov in the 1985 world championship, with three wins and three draws. 4...b6 is best in our check at +0.21, with 4...O-O at +0.20, while 4...d5, which reaches the Ragozin, is the least precise at +0.36.

White can also avoid the opening with 3.Nf3, after which 3...Bb4+ is the Bogo-Indian. Our dataset starts after 3.Nc3, so it says nothing about how often this happens.

What the numbers say about how to play it

Put the two tables together and the practical advice follows. At 1000 to 1400, you meet 4.Bd2 more than any other move, followed by 5.a3 or 5.Nf3, and your most likely bad news is 5.e4, which is good news for you. At 1600 and 1800, 4.Bg5 and 4.Nf3 become the main roads, so the Leningrad ...c5 plan is worth knowing. At 2000 and 2200, 4.Qc2 leads, 4.e3 climbs to second by 2200, and the work is the Classical and Rubinstein structures above.

Black's plan changes with the system. In the Classical, open the game fast to exploit the lead in development. In the Rubinstein, 4...c5 pressures d4. In the Samisch, blockade with ...c5 and attack c4. That is the argument in why you should stop memorizing openings and learn plans.

If you want a more forcing Indian setup against 1.d4, the King's Indian Defense page and the Grunfeld Defense page use the same method for the other main Indian defences. If you would rather avoid the pin and play for a solid centre, the Queen's Gambit page is the other main road.

Is the Nimzo-Indian a good opening?

Yes, in the narrow sense that Black is fine in every line we checked. The largest White edge in any line we engine checked was +0.37 (White slightly better), after 4.e3 b6 5.Ne2, and several lines fall to +0.07 or lower, with the Samisch slightly favouring Black. None is a refutation, and none of these numbers is a winning edge for either side. At master level, chess.com reports, as of October 2026, that White wins 34 to 37 percent of games with its top nine fourth moves, which are overall scores, not rating splits.

The real cost is the repertoire. In the 2012 chess.com thread, one poster wrote that the Nimzo is great but a bit drawish, another that the variations are not intuitive if you are not already a Nimzo player, and a third that you must learn another defence in case White does not play Nc3. Those posts are from 2012, so treat them as opinion, not evidence. A backup against 3.Nf3 (Bogo-Indian, Queen's Indian or Queen's Gambit Declined) is standard advice, and Sielecki's Everyman repertoire pairs them.

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, not a record of real games, so its percentages are predictions and can be wrong at the margins. The line checks are quick local runs, not a substitute for an opening database or a repertoire course. We have no game results by rating, so we cannot say how Black scores at your level, and our dataset starts after 3.Nc3, so it says nothing about how often White avoids the Nimzo. Wikipedia is the better reference for the full list of named variations and history and chess.com is better for master-level scores. None of the pages we opened splits White's choices after 3...Bb4 by rating, which is the gap this one fills.

What to take from it

The engine sees 4.Qc2 at +0.24 and 4.Nf3 and 4.e3 at +0.21 as effectively tied, so the interesting story is system choice: 4.Bd2 leads from 1000 to 1400, starting at 38.5 percent, 4.Bg5 leads at 1600 and 1800, and 4.Qc2 leads only at 2000 and 2200. After 4.Bd2 O-O, 5.e3 and 5.Nf3 are the engine's choices, 5.a3 is the popular quiet one, and 5.e4 at -1.09 is played 17.6 percent of the time at 1000 and 6.6 at 1800. As Black, take on c3 and then on e4 if White pushes, and aim at the centre with ...c5 or ...d5. Maia3 is a model of human choice from the maia3-5m checkpoint, not a count of real games, and every percentage above should be read that way.