Atomic Opening Help

What is the best response against 1. Nf3 f6 2. e3 e6 3. Nd4?

I've tried 3...c6 but that runs into 4.Nc3. 4...a6 nad 4...b5 lose to 5.Nf5, while 4...g6 or 4...Bb4 run into 5. N(d/c)b5

Please help me haha

The 3... f5 lines where black sacs lots of material for lots of initiative and quite a few traps in quite a few different lines is probably black's best try there imo.

3... c6 4. Nc3 Bb4 5. Nb5 g6 also isn't terrible for black even though it's annoying to play. Black is at least able to keep making moves there and has some plans to attack white. 3... c6 4. Nc3 Bb4 5. Nb5 g6 is probably a draw if white blunders with 6. Nxa7 for example where a sample line could go something like 

3... c6 has another problem however. 1. Nf3 f6 2. e3 e6 3. Nd4 c6 4. Nb5 cxb5 5. Qh5+ g6 6. Qb5 Nc6 7. Qb6 axb6 8. Bb5 Kf7 9. Bxd7 Rxa2 10. b4. Here black is down a piece for two pawns. The endgame is complex and black has decent drawing chances via pawntization, but if white plays things conservatively black should have no winning chances and is probably winning objectively in a much simpler manner and with less risk than after 4. Nc3.

Instead I would recommend you play what most top players play: 1. Nf3 f6 2. e3 d5. Here black does lose two pawns after 1. Nf3 f6 2. e3 d5 3. Ng5 fxg5 4. Qh5+ g6 5. Qe5 Be6 6. Qxc7 b5, but black gets very good attacking chances. Black's kingside knight will be very hard to prevent from entering white's position, as attempts by white to stop it with f3 will allow black to move his knight forwards anyways in most cases, as black's play with a rook on the f file is usually faster than white's in these positions. This line is probably still lost, but it's far from clear, and there are a seemingly endless amount of tricks black can play here. It's what almost all top players play against 1. Nf3 f6 2. e3. The opening explorer on lichess shows that black actually scores better at all but the highest level after 1. Nf3 f6 2. e3 d5 3. Ng5 fxg5 4. Qh5+ g6 5. Qe5 Be6 6. Qxc7 b5, and even at the highest level in this position white only wins 51% of games with black still being able to win 39%. This position gives good chances to both sides and really favors the more creative player imo.

I've lost every game after 1. Nf3 f6 2. e3 e6 3. Nd4 c6 4. Nb5 cxb5 5. Qh5+ g6 6. Qb5 Nc6 7. Qb6 axb6 8. Bb5 Kf7 9. Bxd7 Rxa2 10. b4.

I will give the 2...d5 line a try. 1. Nf3 f6 2. e3 d5 3. Ng5 fxg5 4. Qh5+ g6 5. Qe5 Be6 6. Qxc7 b5 looks very playable. Thanks for your help!

You certainly shouldn't be losing every game. I suspect you just don't know the drawing ideas. For some examples of what the drawing ideas in 1. Nf3 f6 2. e3 e6 3. Nd4 c6 4. Nb5 cxb5 5. Qh5+ g6 6. Qb5 Nc6 7. Qb6 axb6 8. Bb5 Kf7 9. Bxd7 Rxa2 10. b4 I suggest you look at these high level games from the 2019 Atomic World Champion, Onubense:

https://lichess.org/sLmq0ZgW

https://lichess.org/WPVaScr4

https://lichess.org/VKQm7bxA

Again, black is worse, but no one has shown that it's for sure completely lost yet. Black is playing for a draw instead of a win, but he certainly has good drawing chances if white is competent.

Keep in mind that stockfish just sees white up a bishop and therefore thinks that white is winning in positions that we say are "pawnitized". Pawnitization occurs when one side is up a bishop in an endgame but can't make effective use of it because there is no way to create a passed pawn as all pawns are on the opposite color of the bishop. There are exceptions of course, and these types of positions are hard to set up, but this idea is absolutely critical to playing these lines as black and makes this line somewhat playable instead of simply being lost.

Atomic has much bigger drawing margins and much more complex endings than most people give it credit for.

Those 3 games are great examples. You are right that I didn't understand the drawing ideas. In game 3 I can understand how 16...Rb2 draws, and how in game 1 the white bishop wasn't able to sac on any of the pawns anymore. However can't white just trade his bishop for black's bishop/knight, fix a pair of pawns, create 2 passed pawns and win? After black sacs his last minor piece white will queen and win the endgame. Even if White can't fix a pair of pawns I feel like a position like this (taken from game 2)

is won for white, since neither pawn can ever be taken. 

I most likely just don't understand how black can avoid positions like this, these endgames take a different mind to analyze.

Edit: Actually, maybe the diagramed position is still drawn. Even with two queens white can't force black away from his king. Thanks to the extra bishop he won't ever be in zugzwang (maybe).

The position shown above is a draw as black's bishop will be able to take when one of the pawns promotes leaving material as k vs kq with connected kings. And if white moves his king anywhere black can just follow, making sure to keep his king close enough that the kings could connect if white spent time to promote. This is a draw. Black actually messed this up in the actual game due to time pressure, but that's more or less irrelevant since that position is dead drawn. Here's a list of endgame evaluations for various material situations if you want it: https://chronatog.com/2019/12/atomic-chess-pawnless-endgame-mate-studies/

Generally though black should work his hardest to ensure that white cannot ever take with his bishop in those lines without incurring a draw in some other manner. It's very hard for humans in practice, and engines don't understand these endings at all, and those lines offer next to no winning chances for black, but they've never truly been shown to be completely busted. So if you really want to play that way you can, but again, I would certainly recommend 2... d5 or even 2... e6 3. Nd4 f5 over it.

In regards to catask's mention of 2... e6 3. Nd4 f5, I still think that 2... d5 is a better choice. It teaches players how to play from a material disadvantage very well in my opinion and will continue working even in 3+2 games, which is what most people here are playing, and against strong opposition. 2... e6 3. Nd4 f5 is probably the second best way to play imo, and good enough to work well if your opponent hasn't studied it, or if it's a bullet time control, but I don't think it holds up nearly as well when higher quality of play is demanded, nor do I think it's as good for a player's growth.

That being said catask is a stronger player than me, and is certainly much better at bullet, even if he hasn't researched lines as much as me. He is certainly correct that 2... e6 3. Nd4 c6 is hellish and miserable to play for black though, even though it has been shown to be maybe viable if you really want to go that route for some reason.

only way to win here is to time someone out

I was idle on chess.com for years, in the meantime I became a relatively strong player and somewhat of a theoreticion in atomic chess at an other site, which I won't name.

I just discovered that chess.com introduced atomic variant and gave it a try.

I'm surprised how much players here are theoretically unaware. There was tons of theory developed for atomic, and I don't see these known lines played, not even from strong players. I have met a very rude player who accused me of cheating, where I was just playing a well known theoretical line.

Also players here resign very quickly.

In general my observation is that chess.com players are more positional. I'm a poor bullet player, but here my bullet rating is not terribly different from my rapid, which is a surprise, because I'm used to my rapid being 200-300 points higher than my bullet.

Just to give you an example:

https://www.chess.com/variants/atomic/game/17687220/21/3

Players here fall to this known opening trap countless times. A 2000 player to play like this... Surprising.

I'll add that after 1.Nf3 f6 2.e3 e6 3.Nd4, the move 3...f5 is due to yokke (from FICS many years ago). As White, after yokke would play 3...f5 against me, I got a good game with 4.Nf3! Also, we used to play 3...Nh6 and barely hold on for a draw (but engines now think it is bad). (Rekursiv)

 I think atomic players here prefer to have fun by thinking of moves by themselves instead of memorizing theories. If you just want to become a stronger player, this site really doesn't suit you.

ok

Many thanks to all who contributed to this page, There is much to ponder here,

The comment by 123a567 made me think about what I want from playing atomic chess. To become a strong player one needs to memorise a whole series of opening traps, how to set them, and how to avoid them. It reminds me of how I beat beginners in three moves, which is not satisfying. By learning complex opening traps one can beat players of my level (around 1950) using memorised sequences, much like beating 1500-rated players. I wonder whether I want to do this. The fun part of chess is being engaged in a battle of wits where neither side can predict the outcome. In atomic chess there are dramatic turn-arounds and deadly threats inherent in innocent-looking positions. By using long memorised lines one avoids engaging in a real battle between two minds. If the other person does not know the refutation then one wins the game, but one does so without being creative, without even thinking. Instead, it is the losing player who is called upon to be creative to find the refutation in real time.

Another thought:

On the other hand, learning to play well in the sense of being able to make plans and generate tactics relies on learning patterns. So a degree of memorisation is necessary. I guess there is a line in the sand between memorising sequences of moves and recognising patterns. For example, it is enough to know that a queen can mate by herself provided the kings are not next to each other. One does not need to memorise the mating sequence.

play Nh3

ok

or play Nf3

Nf3 f6 g4 g5 h4

sounds good

but i have a defense for it too