Beat the Spam: Nh3

I think white having to learn 5 or different spam variations that black has choice of which to play of is not such much of a problem. white c pawn removed, white f pawn removed, white g pawn removed, black d pawn removed, and black e pawn removed are all similar or better to how bad it is for black of in normal atomic I think. This would be 6 choices that could pick which white would have to try. Black would only have to pick one of these lines and could prepare in depth, while white if he did want to try to find a forced win would have to learn much more for all 6. White having prep advantage I do not think it would be of such an issue. Additionally, I think it is much more likely that there is a draw of in one of these positions than in standard atomic. Certain other pawns such as, black f pawn, maybe is a little worse of for black than normal atomic, but still there are some very interesting ideas.

Even if removal of a pawn ended up only to cause black problems from a point of view of theory, which I think this is very much unlikely, black could pick to not remove a pawn at all and be no worse off than standard atomic. Infact, black would still be better than currently since white players not would be able to get quite as much practice of spam lines since other players would would likely remove pawns leading to white having to spend more games of in other lines. There is literally no downside of to this rule change other than that it is would need to be a change of to the code of lichess, and would certainly help to make for a more balanced game of in practice, even if not it did do such (again this is unlikely) of in theory.

Not even such hard it would be to code this into lichess for the gui. At the start of the game black could pick a pawn to remove by clicking a pawn, or pick not to remove a pawn by clicking a king. Maybe this could also be clicking any other piece than a pawn. It would not need of to be complex. It would, I think, be very easy to code into lichess, and would be an easy transition for players to make because still games of normal atomic easily could be played. It could even fit the theme of explosions by saying that black "blew up" a pawn (although maybe it is should be reworded since it not would follow normal capture rules since surrounding squares not would explode) which would be a nice addition.

Atomic960 has many clear downsides compared of to normal atomic, but I fail to realize any downsides to this change of the rules compared of with normal atomic. If you do not think this rule change would be good for atomic overall, please say why. The only people it is would harm are those who only play one forced win or forced draw line with white over and over. I am personally fine with making this aspect of the game harder for people.

I like the idea in #51, however this will never be implemented, it is much pain for little gain for the developers.

On the other hand I wonder why a6 is not played in the place of b5 in the standard 1. Nh3 setup. Nc3 does not force b5 at all, you can argue that Na3 forces b5, because if a6 then there are Nc4 ideas, but even those can be taken care of, if black is aware of this threat. As I reviewed the database, a6 was indeed played in many cases and I could not find any clear refutation of this move in the games.

>it is much pain for little gain for the developers.
I do not think that this is true. It is little gain, yes, but not I do think it would be such painful compared with many other changes proposed such as adding a 960 option of for all variants (I also still would like this btw, but it would be a huge amount of work) because all that is would have to change is a single call or two at the start of atomic games that adjusts the fen based on black's click. More could be added to make this look more pretty, yes, but that is all that are should be needed.

>however this will never be implemented
Sadly I do think this is probably true. I do think that if there was overwhelming support of for this idea from the atomic community, and if there was someone willing to code this feature, lichess might be okay with implementing it. However, I think that all 3 of these things would need to happen for this feature to be changed. I think it is possible however, but agree that not it is likely of to happen without at least one coder and one community leader actively proposing of it to lichess.


I beg to differ. This is at least as difficult as adding 960, because it is essentially starting the game from a custom position. If you can start a game with the FEN including a missing pawn, then you can start a game just as easily from a 960 position. Also you have to add a non standard move input to the board, as removing a piece is not part of the board currently ( may be it could be implemented as dragging the piece out of the board ). This is adding complexity beyond imagination, and even if someone coded it, developers would think twice to take the risks this introduces to the code base.

@lishadowapps I disagree, adding 960 to all variants requires lots of changes all over the site GUI, including new ratings (how the user page will look?), picking the variant (a check to make 960 or what?). As another downside of 960, some of 960 positions may be more forced wins than standard one, and strong people will definitely learn some of these forced wins.

In contrast, changing the rules of one variant is more like some work under the carpet, even if it includes a new type of move, but it is not complexity beyond imagination, as Crazyhouse shows. For what Xeransis asks I'd recommend double-click on the pawn mechanics, and that is much easier than Crazyhouse pulling from the pocket move. Just replacing Atomic with Atomic 960 would be even easier. But when we talk about any replacings other than additions, there will always be people who are strongly against that, and that is more of a social problem, than a programming one.

Knowing thibault, I'd say this will not be implemented in any foreseeable future, just because there is little gain from this for "average players". In fact, they will probably get mad in case of replacing, because they cannot spam their favourite line now. Xeransis talks from the viewpoint of top players, but that's not how things work on this site :(

You say that adding an option that does make black more playable in this way not would be appreciated by average players, but not I do agree. How many times have you seen a post from a <2000 player who does ask how to defend against 1. Nf3 f6 2. e3 as black? How many lower level players there are who you think quit because it because they feel black is too bad? I think there are many low level players who I think don't play because they believe atomic as black is too hard in such lines. They quit before understanding that really most lines are not truly such bad for black except maybe at the absolute highest level. Even the ones that do keep playing often complain in chat of tournaments about how hard it is to play black. This rule would at least give a benefit to being black, of all levels. Currently there is only a benefit for to play white, since him is better, but this would give some benefit to being the black player as well, so that not it is simply a game that must be suffered through. It is possible that some players would prefer playing black to playing white (even though theoretically still the advantage is with white at high levels) with such a rule, which not I do think would be a bad thing at all! I think a survey would be the proper way of to tell. Additionally, I do not see why Thibault would be such against it if a strong majority of atomic players, especially those who play most of the games of atomic, agree that it is better, and if someone else does the work to add in the coding. If it is would take very very little or no work from Thibault, I do not see why him would want to reject something if not it does require very much work from him either way, or does not divert resources from other lichess projects.

I think that the best way would be to see if there is are interest from the active community of atomic, such as in a thread of this forum to see if anyone does have opposition to such an idea, and to see why they think such a rule is would be bad. Perhaps I will write up one and post it here. After a week or so, if the community thinks such a rule would be a clear improvement with little or no trade-offs, and only if the community is in a very clear agreement over such a rule, it would be best to send a pm to Thibault asking exactly what would have to happen for him to be okay with such a change. I feel that there must be at least some set of things that if they were to happen he would be okay of with such a rule change. Especially I think this would be the case if it does not change very much about the game negatively. One advantage I like of this rule over others is that there is very little change in the skills required of to be good at this atomic because still it would be an opening theory heavy game with most structures and patterns remaining the same (this is one of the big dislikes I have of 960), and that it is 100% possible to play in the previous way as well, meaning opening database could be kept. Then after hearing what would be required of the atomic community to cause such a change to happen, it can be seen whether these things can be accomplished.

@Xeransis Maybe you're right, I'm just too tired of the fact that all lichess new features for the latest several years look useless for me (like team tournaments), while some things I find useful like Swiss tournaments and bughouse are largely ignored despite community desire to have it and implement it. I am ready to support you in advocating for this or similar feature that will make Atomic both newbie-friendly and more playable at the highest level.

I had another idea that was to oblige white play first 3 moves with pawns, while black is only forced to make first 2 moves with pawns (or white makes 2, black makes 1). This should (though I'm not sure it will) help greatly as most damage in the opening by white is caused by the knights, while in this variant it is black who can make the first knight move. This has an advantage of not adding any implementation difficulties with new kind of moves. As a downside, I'm pretty sure that black is still considerably worse in such a game, and it may also reduce opening variety.

IIRC, @onubense had an idea that the kings can legally disappear simultaneously (for example, by taking one another), making the game a draw. This will make many opening lines more drawish if black is able to connect the kings, but it also has downsides of white still having large advantage (maybe even winning in 2N opening) and rewriting the beautiful endgame theory that we have now.

If anything, I think that prior to opening a forum thread with asking for a concrete Atomic update we should first discuss best ideas among ourselves (though we should really ask thibault if he can consider such a change).

As an additional support of ideas like that, I promise to donate $250 to lichess for implementing a modification of Atomic that will be reasonably community endorsed (probably tested by strong players on prior to release aside from forum debate) and more newbie-friendly than classical Atomic + a separate $100 on the prize fund of a big inaugural tournament for such a variant.

The problem of with such ideas as onubense's suggestion is that I think it will make too much easy for a draw in many lines, and still not work in others like 1. Nf3 f6 2. Nc3 as you have said. The suggestion of having to move certain pawns first is interesting, but it would kill all current theory and make a clearly different variant. Here this suggestion is would be an expansion that would not get rid of any theory, just add more. The only problem I think of my idea is I do not know if it is would be enough for to draw. If all lines are still lost, it would be an improvement, but I would very much like it to be enough of an improvement that at least it could be argued of theory that atomic is a draw before something such drastic is done. Currently that is not the case for 1. Nf3 f6 2. Nc3. I think I will analyze all starting positions with stockfish to see if it can be making the case for it to be theoretically better for black than standard atomic as well as for in practical games. This will give a rough idea about how bad for black the new positions are. If this idea is does work, with black having options of to have a position less than +2.0 I think it would be best solution in my opinion. I thought this did work with d pawn removed, but position in after d pawn removal 1. d4 (clearly this is the only good move) was +2.34 with the following lines:

<--[17:14:39.233]:info depth 47 seldepth 83 multipv 1 score cp -234 nodes 46237125599 nps 1795224 hashfull 1000 tbhits 0 time 25755622 pv g8h6 f2f3 c8h3 b1c3 c7c6 g2h3 e7e6 c1g5 f7f6 g5f4 e6e5 c3d5 c6d5 c2c3 d8a5 e2e3 a5a4 d1c2 b8c6 f1b5 a4c2 g1e2 e8c8 d4d5 h6g4 f4h6 g7h6 f3g4 h8g8 h1g1 g8g2 g1g2 c8b8 e2g3 f6f5 a1c1 f8h6 b5c6 f5f4 d5d6 f4g3 d6d7 h6g5 c1d1 g5h4 e1e2 e5e4 d1d5 b7b5 d5f5 d8f8 c3c4 h4g5 e2f1 b5c4 f5f7 f8f7 b2b4 a7a6 b4b5
<--[17:14:39.233]:info depth 46 seldepth 72 multipv 2 score cp -239 nodes 46237125599 nps 1795224 hashfull 1000 tbhits 0 time 25755622 pv b8c6 c1g5 g8f6 a2a3 e7e5 e2e3 f8b4 a3b4 e8g8 c2c3 g8h8 f2f3 h7h6 h2h4 c8g4 g1h3 a7a5 f1b5 e5d4 d1d7 c6b4 e1g1 b4d3 b5d3 d8d7 b1a3 b7b5 e3e4 b5b4 e4e5 f6d5 e5e6 d5e3 g5e3 f8d8 f1d1 d8d2 d1d2 a8d8 a1d1 f7e6 d1d7 d8d7 h3g5 g4f5 g5f7 h8g8 g2g4 f5d3 g4g5 h6g5 f3f4 g8f8 f4f5 b4a3 h4h5 a5a4 g1f2 d3e2 h5h6 g7h6
<--[17:14:39.233]:info depth 46 seldepth 71 multipv 3 score cp -259 nodes 46237125599 nps 1795224 hashfull 1000 tbhits 0 time 25755622 pv g8f6 c1g5 b8c6 a2a3 e7e5 e2e3 f8b4 a3b4 e8g8 c2c3 h7h6 f2f3 g8h8 h2h4 c8g4 g1h3 a7a5 f1b5 e5d4 d1d7 c6b4 e1g1 b4d3 b5d3 d8d7 e3e4 f8d8 b1a3 d8d3 f1d1 b7b5 d1d3 g4d7 a1e1 a5a4 e4e5 f6d5 g5e7 g7g5 e7f6 d5f6 h4g5 a8g8 g2g3 d7e6 e1d1 b5b4 h3f4 g8d8 f4e6 d8d2 d1d2 h6h5 f3f4 h5h4 g3h4 b4a3 e5e6 f7e6 f4f5 h8g8
<--[17:14:39.233]:info depth 46 seldepth 77 multipv 4 score cp -327 nodes 46237125599 nps 1795224 hashfull 1000 tbhits 0 time 25755622 pv e7e6 g1f3 b8c6 f3e5 c6e5 e2e4 g8f6 h2h3 c7c6 b1a3 b7b5 g2g3 f8d6 f1e2 h7h6 f2f4 f6h5 e1g1 e8g8 g1h1 a8b8 b2b3 a7a5 a1b1 f7f6 a3c4 b5c4 b3b4 c8a6 c2c4 a5a4 c1d2 a4a3 b4b5 d6b4 d2b4 c6c5 b5a6 b8b2 b1b2 d8d7 f4f5 d7a7 f5e6 f6f5 d1a4 f5f4 a4d7 a7d7 e4e5 c5d4

I feel I should at least have idea of how good this positions actually are before going further with this idea, but if it did look like it did work, it would be better than other suggestions I have heard I think.

Putting out a solution I first heard from ulath, and tried out over the board with him, is to swap the starting positions of all bishops and knights. (RBNQKNBR)

- Symmetric starting position
- SF eval close to zero
- Castling still the same
- Pattern recognition for major piece movement still largely unchanged
- Opening choices still quite varied and balanced (from testing)
- Middlegame/endgame principles unchanged

- Not the regular chess starting position (but if you're considering throwing that out the window anyway...)
- Opening theory needs to be rewritten (not a bad thing?)

Removing White's g-pawn (negating some of Stockfish's favorite Nf3 strategies) lead to a 100 cp position after depth 26:

info depth 26 seldepth 55 multipv 1 score cp 100 nodes 272606055 nps 2021715 hashfull 440 tbhits 0 time 134839 pv g1f3 f7f6 h1g1 g7g6 b1c3 c7c6 e2e3 d7d5 f3e5 f6e5 d1f3 c8f5 f1d3 e7e5 g1g6 d8f6 d3g6 e8d8 g6h7 f6h4 f3g3 d8d7 c3a4 h4h2 d2d3 b7b5 f2f3 f8e7 a4c5 e7c5 e3e4 b8a6 c2c3 d5e4 c1h6 a8g8 h6g7 e5e4 d3e4