Dutch Defense: What Each Rating Faces After 1.d4 f5, and the 2.Bg5 Gap

The Dutch Defense answers 1.d4 with 1...f5, claiming e4 at once and accepting a weakened king in return. Guides sort it into three systems and tell you it is rare enough that opponents arrive unprepared. What they rarely say is what White players at your level actually play against it, and how far those choices sit from the engine's. We ran the maia3-5m checkpoint, see how Maia chess actually works, on the position after 1.d4 f5 and on a position two plies deeper, at seven rating bands from 1000 to 2200, and graded every legal move with Stockfish 18. 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 +0.50 means White is better by half a pawn. The dataset tables come from Stockfish at depth 20. The line checks are separate runs of ten seconds per line that reached depths between 23 and 26, so expect a position to differ by several hundredths between the two methods.

What the Dutch is, and the three systems

The opening is defined by 1.d4 f5. Wikipedia files it under ECO codes A80 to A99, and credits Elias Stein with recommending it as the best reply to 1.d4 in his 1789 book. Tarrasch rejected it as unsound in 1931. As of October 2026, chess.com calls it somewhat offbeat but ever-present, and says it never had a period of exceptional popularity. As of October 2026, Wikipedia lists Alekhine, Larsen, Najdorf, Simon Williams and Nakamura among players who used it with success, and says Carlsen used the Stonewall to beat Anand and Caruana.

Black has three standard setups, and the choice mostly decides what kind of game you get.

• Stonewall. Black plays ...d5 and ...c6 behind the f5 pawn and ...e6, locking e4 so a knight can settle there and drive a kingside attack. A typical order is 1.d4 f5 2.c4 Nf6 3.g3 e6 4.Bg2 d5 5.Nf3 c6 6.O-O Bd6. Tartakower's 1922 win over Maroczy is the classic example. The cost is a bad light squared bishop and a hole on e5.
• Leningrad. Black fianchettoes with ...g6 and ...Bg7 and plays ...d6, which frees the light squared bishop. The structure resembles a King's Indian, see the King's Indian Defense page. The key position is 1.d4 f5 2.g3 Nf6 3.Bg2 g6 4.Nf3 Bg7 5.O-O O-O 6.c4 d6. As of October 2026, chess.com says the Leningrad is more common, though it adds that perhaps no variation is more associated with the Dutch than the Stonewall.
• Classical. No fianchetto and no early ...d5. Black keeps pawns on d6, e6 and f5 with the bishop on e7. It is the most flexible and the least forcing.

We checked one tabiya of each with Stockfish, White to move. The Stonewall position above scores +0.46, with 7.Ne1 best. The Leningrad position scores +0.54, with 7.Nc3 best and 7.d5 next at +0.53. The Classical position 1.d4 f5 2.c4 Nf6 3.g3 e6 4.Bg2 Be7 5.Nf3 O-O 6.O-O d6 scores +0.60 with 7.b3. So White keeps about half a pawn in all three, and no system is clearly the soundest. The Stonewall is the lowest of the three, by a few hundredths.

The first decision: what White plays after 1...f5

This is the position at the heart of the dataset. Stockfish ranks 2.Bg5 first at +0.68, then 2.c4, 2.Nf3 and 2.g3 together at +0.55, 2.Nc3 at +0.47, 2.Bf4 at +0.46 and 2.e3 at +0.43. The weakest of the main candidates is 2.e4 at +0.10. So apart from the 2.e4 gambit, the engine puts every common second move between +0.43 and +0.68. 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 story is the top row. The engine's best move is found by 1.3 percent at 1000 and only 8.3 percent at 2200. Nobody at any band is predicted to play it as their first choice. It climbs with rating but never gets near the club favourites.

The likeliest move is 2.Bf4 at 1000 and 1200, at 25.7 and 27.1 percent. From 1400 up, 2.c4 takes over: 30.9 percent at 1400, a peak of 40.3 at 1800, then back to 30.2 at 2200 as 2.Nf3, 2.e4 and 2.Nc3 gain ground. 2.c4 is the traditional move order and heads into the Classical, Stonewall and Leningrad families. 2.Bf4 heads for London System structures instead, see the London System page, and it fades from 25.7 percent at 1000 to 10.7 at 2000.

The gap is small, and we want to be plain about that. The distance between the engine's choice and the likeliest human choice is 22 centipawns at 1000 and 1200 and 13 from 1400 up. That is a fifth of a pawn at worst. Missing 2.Bg5 costs White almost nothing. What the data shows is behaviour, not blunders: humans sit on the main road, and the engine's sharper move is a path they rarely take.

The Hopton Attack, 2.Bg5

2.Bg5 is called the Hopton Attack, though we could not source who Hopton was. The idea on chess.com is to eliminate Black's knight when it goes to f6, and Wikibooks calls it a very aggressive way to attack where Black can equalize later. Our numbers say otherwise, but only by a little. After 2.Bg5 Stockfish gives White +0.63 at depth 24 with Black's best replies as follows.

• 2...h6 is best at +0.63. The line 3.Bh4 g5 4.e3 Bg7 5.Bg3 c5 6.c3 is what the engine prefers.
• 2...g6, which Wikipedia calls the most popular reply, is next at +0.68. After 3.Nf3 White keeps +0.63, and 3.e3 and 3.Nc3 are at +0.59 and +0.60.
• 2...d6 is the third option at +0.77.
• 2...Nf6 3.Bxf6 exf6 doubles Black's pawns, and White is at +0.76 with 4.e3 or 4.c4. Wikibooks says White is happy to make this trade, and our check agrees.

Then there is the trap everyone quotes. After 2...h6 3.Bh4 g5 4.Bg3, Black plays the greedy 4...f4, White plays 5.e3, and 5...fxg3?? loses on the spot to 6.Qh5, which our check confirms is checkmate. Even if Black does not take the bishop, 4...f4 is bad: the position after 5.e3 scores +2.03 for White at depth 23, against +0.63 after 2...h6. Instead of 4.Bg3, chess.com notes 4.e4 is also playable, and the engine's own line above prefers the quieter 4.e3.

The Hopton Attack is also easy to meet without knowing the trap. Just do not play ...f4 when the g3 bishop is hanging on the diagonal. Maia3 puts 2.Bg5 at 1.3 percent at 1000, so a Black player at that level will almost never need this section. By 2200 it is predicted at 8.3 percent and the trap knowledge is worth having.

The club moves: 2.Bf4 and 2.c4

2.Bf4 is the favourite at the lowest two bands, and it is a London System move. Wikipedia lists 2.Bf4, or 2.Nf3 followed by 3.Bf4, as the London System. After 2.Bf4 our check gives White +0.53 at depth 26, with three Black replies in range: 2...e6 at +0.53, 2...Nf6 at +0.57 and 2...d6 at +0.64. The engine puts all three replies within eleven hundredths of a pawn.

2.c4 is where most rated players go. The second dataset position follows the likeliest line at 1400, 2.c4 Nf6. There Stockfish ranks 3.Nc3 first at +0.56, with 3.g3 and 3.Nf3 at +0.55 and 3.e3 at +0.50. Maia3 agrees with the engine strongly.

This is the opposite of the first table. Here the engine's best move is also the likeliest at every band, a gap of zero, and it climbs from 49.6 percent at 1000 to 77.2 at 2000. It dips to 71.4 at 2200 because 3.g3, the mainline fianchetto, rises to 10.9. The move that falls fastest is 3.e3, from 12.5 to 1.4 percent. The one to avoid is 3.Bg5, which scores only +0.22, but it sits at 5.2 to 6.5 percent through 1600 and keeps a small share to the end. Everything else stays near half a pawn.

So the pattern across both positions is consistent. Players find the structural moves, c4 and Nc3, quickly, and the engine rates them at or near its best. Where the engine disagrees is in the piece moves: its Bg5 on move two is almost never chosen, and the bishop to g5 on move three is a mild error that a few players still make.

The other anti-Dutch lines

Three more choices come up in sources and in our data.

• 2.Nc3. Maia3 gives it 10.4 percent at 1000, 9.0 at 2200. Our check scores it +0.47 in the dataset, and the longer check finds Black's best replies are 2...Nf6 at +0.43 and 2...d5 at +0.51. One reply to avoid is 2...Nc6, which gives White +1.06. The standard continuation 2...Nf6 3.Bg5 is the Raphael Variation, per Wikipedia.
• 2.e4, the Staunton Gambit. White gives a pawn for quick development. Christiansen and Silman say it offers White equality at best, and chess.com says it used to be thought to lead to a very strong attack and no longer scores well. Our check agrees: 2.e4 fxe4 3.Nc3 Nf6 4.Bg5 g6 gives White only +0.10. One Black alternative, 4...Nc6, gives +0.41 and 4...c6 +0.44. Maia3 predicts 2.e4 only from 2000, at 7.1 percent, and 10.9 at 2200. Wikibooks also calls it by far the most popular choice, which contradicts both Wikipedia and our data, so we ignore that claim.
• 2.g3. The main modern move according to Wikipedia, and the Leningrad and Stonewall move order we showed above. It scores the same +0.55 as 2.c4 and 2.Nf3 in our data but was outside the top five at every band.

There are more offbeat tries, the Krejcik Gambit 2.g4 and the Mayet idea 2.h3, but we have no data on them and they are rare.

Which system to choose as Black

Your choice should depend on what you want from a game. If you are 1000 to 1400, you will most often meet 2.Bf4 or 2.c4, and the answer to both is a flexible setup. The Stonewall is the easiest to learn because the structure never changes: pawns on c6, d5, e6 and f5, a knight heading to e4, and a plan on the kingside. Its price is a locked bishop. The Leningrad gives more dynamic play and a free bishop, but you must know your King's Indian ideas, since White often plays d5 to control e6.

If you are 1800 and up, the engine's second move for White is found a little more often, 3.9 percent at 1800 and 8.3 at 2200 for Bg5, and you will see 2.e4 and 2.Nc3 as well. The Classical setup gives you the most room to transpose. Whatever you choose, accept that White stays ahead by roughly 0.5 in the sample lines. The result you want is not equality. It is an unbalanced game where White needs to play accurately.

If you play White against the Dutch, the data says what you already suspect. 2.c4 followed by 3.Nc3 is the most commonly chosen path and also one of the engine's best, and 2.Bg5 is the sharper alternative the engine likes a shade more. At the club level the gap between them is small. Our chess openings by rating dataset uses the same method for other openings, and for learning the ideas rather than memorising moves, why you should stop memorizing openings and learn plans fits the Dutch, since Black's plan is the same in every system: control e4, then attack the kingside. If you like a sharper counter to 1.d4, the Benko Gambit page works the same way.

Is the Dutch a bad opening?

Forum opinion is split. One poster calls it a bad opening, another calls it refutable but tricky, and a third says that once you ran into a prepared opponent there was little chance of a decent outcome. Book assessments are kinder: Wikibooks marks the Stonewall and Leningrad lines after 2.c4 Nf6 3.g3 as equal. Our engine numbers sit between. White keeps +0.46 to +0.60 in every tabiya we checked, and +0.68 after the best second move. That is a clear but modest edge, well short of the kind of lead that decides a game between humans. The honest verdict is that the Dutch is objectively worse than the main replies to 1.d4 and still practical, because the positions are unbalanced, because many White players at your level do not know the theory, and because White's tempting attacking tries, like 2.Bg5, are not much better than the quiet ones. Chess.com lists the cons plainly: the early ...f5 exposes the king and a well prepared opponent causes difficulties.

One caution on popularity. Chess.com's 2006 article called the Dutch unpopular in top level play, while Wikipedia credits Carlsen's Stonewall wins. Both are dated and perishable, so we do not lean on either.

Where other pages are better

This page is narrow on purpose. The dataset holds two positions, the one after 1.d4 f5 and the one after 1.d4 f5 2.c4 Nf6, and nothing past move three. The line checks are quick local runs and not a substitute for an opening database. We have no game results by rating and did not open the Lichess explorer. Wikipedia is the better reference for history and the full ECO map, and chess.com is better for browsing the named variations. None of the pages we opened splits White's replies to 1...f5 by rating, or shows how rarely 2.Bg5 is found, which is the gap this page fills.

What to take from it

The Dutch Defense gives White a small, steady edge, and the engine's favourite second move, 2.Bg5, is the one humans skip: 1.3 percent at 1000, 8.3 at 2200. The club choices are 2.Bf4 at the bottom of the ladder and 2.c4 from 1400, and they cost White only 13 to 22 centipawns against the engine's best. After 2.c4 Nf6, 3.Nc3 is both best and most popular, up to 77.2 percent at 2000. As Black, pick the Stonewall for a fixed plan or the Leningrad for activity, and learn the one trap that matters. 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.