Atomic chess engine

Hello.

Recently I've noticed that now Stockfish plays atomic. I've decided to play few games vs it. Stockfish plays very good in midgame and endgame, but there are some strange things about openings.

First of all, Stockfish level 6-7 as black blunders in openings very frequently. Something like "1. Nf3 e6".

Secondly, I don't really understand Stockfish's openings book (as black). Just one example in opening I have played many times, Stockfish plays as black:

1. Nf3 f6 2. e3 e6 3. Nd4 c6 4. Nb5 cxb5 5. Qh5+ g6 6. Qb5 Nc6 7. Qb6 axb6??
en.lichess.org/eENO5DF6/white#14
Stockfish thinks it's a good idea to blow up white queen with a pawn, but I don't think so. Openings explorer says this position is much better for white: 79/5/16 for 661 samples at average rating 2k+

After these moves both sides has not so much choice and game continues this way:
8. Bb5 Kf7 9. Bxd7 Rxa2 10. b4
en.lichess.org/eENO5DF6/white#19

This position is relatively simple and white has 2 pawns for knight. Not bad for atomic chess.

I'm not 100% sure it's won for white. But I have played this opening vs Stockfish Level 8 many times and I often won despite the fact I'm a weak in atomic. I think there are 2 variants: 1) This opening is very bad for black or 2) Stockfish permanently blunders in later game after this opening.

In both variants atomic Stockfish is far from perfect.

Here is a small proof: 12 games (w/o takebacks) won vs Stockfish 8 after this opening.

en.lichess.org/eENO5DF6/white
en.lichess.org/lLaXeYwI/white
en.lichess.org/9q08vjCU/white
en.lichess.org/gJxb12m9/white
en.lichess.org/HZbC0UoE/white
en.lichess.org/YeRbZ8sf/white
en.lichess.org/7sduPcPU/white
en.lichess.org/gJxcaKSO/white
en.lichess.org/eArCHqHt/white
en.lichess.org/5MJD1keV/white
en.lichess.org/3wwlYx9E/white
en.lichess.org/HFMSCyxW/white

What do you think? Is this opening bad or are there black's mistakes in later game?

I almost forgot: last but not least.

Stockfish has a very strange position estimation in endgame: it says white has 20 pawns advantages in KQQvKN position (kings are close to each other), but it seems to be a draw.
The strange thing is "20 pawns advantage" estimation in 5-figures endgame without ideas how to win the game.

Do note stockfish is only using 1 second per move.

not that i understand atomic chess

@1

That opening is favorable for white.

At level 8, Sf only searches at 0.4 sec per move and a max depth of 12. So with correct opening preparation, Sf is no match.

Somehow Lichess should equipped Sf atomic with shallow book, just to avoid the well known inferior positions in the opening.

@2

Could you post the FEN of the kqq v kn position?

here is the game with KQQ v KN positions
en.lichess.org/B092AjWH/white#105

I was wrong, 10 pawns advantage instead of 20, but it doesn't matter. There's no advantage, it's a draw

OK, position KQvK with kings next to each other is estimated as +8.9 pawns. But it's a draw: black king would always follow white king

8/8/8/8/8/3Qk3/4K3/8 w - - 2 74

@ixdon2 #6

This has been reported before and I assume it's being worked on.

I also support the idea of using the more accurate atomic clock for OTB chess.

This is an old problem. On Toadofsky's website I even managed to defeat Stockfish Level 8 due to its bad opening book.

SF is strong after the opening and weak in the opening.

@5

kqq v kn

1Q6/k2n4/QK6/8/8/8/8/8 b - - 0 53

I suspect this is a win.

If you want we can play this out, you handle black and I will take white. I will try to prove that this is a win for kqq.