# How to solve atomic chess?

Since losing chess is almost solved, I believe that the next step towards solving chess should be atomic chess, which has less forced moves than losing chess but still a lot more than regular chess.

What are the practical steps we should take towards the solution of atomic chess? I seriously want to help solving it in my leisure time as well as learn more about game solving during this process..

• I don't think this will help us towards solving chess but it's a good goal. I also think it will be exponentially more difficult than loser's to solve because the forcing sequences are more rare. Aug 6, 2015 at 0:48
• By definition, the only way to truly solve a game is to enumerate all and every possible possibility. Aug 6, 2015 at 0:50
• @Cleveland I believe atomic might be a draw although it may still be a white victory. Most human games, including those played by high Elo ranking do not end in a draw though top ranking players draw more often. Aug 6, 2015 at 1:25
• I don't think solving atomic chess is easy. You must generate all possibilities in the game. You can't store all of it in your computer. This will take you years and it's a hard work. Don't underestimate the complexity. You'll most likely give up. Please reconsider. Aug 6, 2015 at 1:32
• @JiK strongly solve all possibilities Aug 6, 2015 at 7:57

Solving atomic chess is probably only feasible if it is a white win, like loser chess is.

To try to solve atomic chess I would probably write/acquire a strong atomic engine first.

Then you build a tree starting from the starting position. If it's white to move you chose only one move, if it is black to move you try all possible moves. The engine evalution will guide you in which white move to chose and which black moves should be investigated deepest.

Basically I would try to build a tree that has leaves with a certain minimum positive evaluation. This way you always have a form of very weak solution of atomic chess, i.e. a tree that tells you how to play with white to get an advantage of at least +x.yz.

You'll try to improve x.yz by analysing those parts of the tree where the leaves have the lowest evaluation. If x.yz gets high enough you can try to work positions out until mate (probably you'll need some tablebases at that point).

If you start this kind of project you probably quickly get an impression, whether getting a significant white advantage is generally easy or difficult, which might be a pointer to whether it is ultimately solvable or not.

This is pretty similar to building a repertoire, so I would probably quick start the tree with an existing atomic chess repertoire.

Of course odds are that your tree will never converge towards a strong solution, but in that case at least you've created a really nice atomic repertoire.