[maggit]

In complex situations it is easy to come to the conclusion that you have to guess, but sometimes there is an elaborate connection you might have missed. This game is sure to call you out on any mistakes you make of this kind.

That would at least make this game more strict and perhaps even more difficult.

[reificator]

I tend to solve Minesweeper locally as much as possible, solving cordoned off sections where the answer doesn't rely on the rest of the board. A major benefit to doing it this way in traditional implementations is that if I'm forced to guess, the sooner I make an incorrect guess the sooner I move on to a solvable map.

While trying this out I've encountered several sections where I would have to guess, and that guess cannot be influenced by other unrevealed cells, such as when there's an island in a corner of the map.

In these situations I will need to guess between these two spaces, but since there are still known safe areas on the map, guessing causes me to lose the game.

I can get used to that behavior of course, but it's fairly frustrating. It'd be nice if the guessing exception rules accounted for situations like this. When there are clearings or known mine patterns that separate out discrete smaller map(s), I want to solve the smaller map(s) before I move on.

[dmurray]

Agreed, I'd like this to go one step further and allow you to guess on a square where you will inevitably have to guess anyway.

I'm not sure how hard this is to add to the SAT solver. The formal definition is something like "if for some set of maybe-squares S, no matter what the solutions are to all of the squares outside S, the set of solutions to S is the same, then allow clicking anywhere in S". But that's a combinatorial explosion: just the number of sets S to consider is a factor of 2^#{maybe-squares} . I don't know enough to say if that can be optimized into something sane so that it can be rigorously applied, but a handful of special cases for small unconnected sections of the board would cover most of it.

[reificator]

> I'd like this to go one step further and allow you to guess on a square where you will inevitably have to guess anyway.

If you have all of the possible information for that guess already, then yeah.

If there are still some unknowns that you could resolve first to get more information, then guessing should still result in a mine.

[dmurray]

Yeah, that's what I meant really, as per my attempt to formalize it.

I can't think of any situations where it makes a difference, though, other than ones where you rely on the mines-remaining counter.