Hexcells is a puzzle game take on Minesweeper. Unfortunately, it’s been decades since I’ve played Minesweeper, so I can’t really comment authoritatively on that game, but it has you gradually unveiling more information about the game. This distinguishes Minesweeper from Nikoli/Conceptis-style paper puzzles where all the information is there at the start, you just have to make deductions from it. And, in Minesweeper, you have to guess instead of depending on deductions, you regularly get to situations where there’s not enough information to be able to mark / clear any of the spaces with confidence.

Hexcells keeps the gradual unveiling approach, but changes the rules so that you can find a solution just by reasoning. Which requires some changes! To start off, it’s a hex grid; I’m sure that affects things, I just don’t have a great feel for how. It’s actually possible that Minesweeper on a hex grid would make it too easy to solve puzzles, because another change in Hexcells is that, for some cleared spaces, they don’t show you the number of adjacent marked spaces, so you have less information that you might have. Also, there are two other variants of cleared spaces: some of them say that the adjacent marked spaces are consecutive, others say that the adjacent marked spaces aren’t consecutive.

But some of the marked spaces get numbers as well. These don’t show the number of adjacent marked spaces: instead, they show the number of marked spaces within a radius of two. And then there are numbers in some places at the edge of the grid, showing the number of marked spaces in a particular line. (Again with consecutive / non-consecutive variants, though the details of what “consecutive” means are a little different.) Also, the grids aren’t full grids, there are parts of the grid removed completely. And, finally, there’s a total count of the number of marked spaces on the grid.


That is a lot of rules. I’m sure each one is there for a reason, but they don’t feel particularly elegant to me. And some, honestly, feel a little ridiculous: that total count, for example, is almost never relevant until the very end, where it sometimes lets you fill in the last two or three spaces because the remaining count is either zero or matches the number of undetermined spaces.

Also, there’s another design decision that’s related to the need to gradually uncover information: how do you handle speculation? Sometimes, when thinking out a pencil puzzle, I like to try to work out a chain of reasoning on the puzzle, and then erase (or rather undo, since I’m almost always on my iPad) once I find a contradiction. But, if you want to support that way of solving, then you run into problems with uncovering new information, because that brings a non-deductive element into the game.

Hexcells decides to not let you speculate at all: if you click and get it wrong, it’ll refuse to make the move and an error count will go up, while if you click and get it right, then that move is there permanently. I found that latter decision to be actively annoying: sometimes I’d tap in the wrong place, or realize right after tapping that I’d done something that wasn’t justified by facts on the ground, but then the effects were there on the grid. So I either had to come to peace with guessing or try to pretend that I hadn’t actually gotten information about that one hex on the grid or restart the entire puzzle; I don’t like any of those solutions.

And you can only go so far with deduction, because you have to be able to do all the deduction in your head, so the game can’t really stretch you. And, because of all the rules, everything feels ad-hoc anyways, so I don’t feel like there are clever deductions to be made. Which doesn’t mean that the puzzles are trivial, some of them do actually require a bit of thought, but nothing like the best pencil puzzles.

Or at least that’s what I thought until I started playing the game’s randomly generated puzzles, and found that I get completely stuck in, say, one out of 25 out of them. Maybe that’s a sign that I’m being dense, maybe that’s a sign that it’s not actually reliably generating puzzles that it’s possible to solve without guessing, but maybe there are subtler strategies lurking there that I just haven’t found?


I still basically enjoy Hexcells: it’s soothing, and its random puzzle generator is pretty good. But it’s also way too ad-hoc to be a great puzzle type, and the specific implementation choice of not allowing undos or speculation is one I don’t like even granted the rules.

And, in the genre of “puzzles that are like Minesweeper but are solvable”, there is a much better puzzle type, the one that Conceptis calls Fill-a-Pix. It’s just the single Minesweeper rule except with all of the numbers that you’re going to get being visible at the start (so, in particular, the numbers are only there on a small portion of the squares): one rule, but one that gives rise to a lot of tricky deduction.

Conceptis actually does two versions of Fill-a-Pix puzzles: the Basic puzzles are trivially solvable by finding a number where enough of the adjacent squares are known to let you fill in all the remaining adjacent squares as marked or unmarked; that mode is soothing, it’s just a hunt for where to go next. Whereas the Advanced puzzles require you to make more subtle deductions; those can get very tricky indeed, on large grids I’ll spend hours on a single puzzle sometimes.

So I wish Hexcells had leaned more in that direction: keep the progressive unfolding and/or the hex grid, but find a way to use a much smaller rule set to force you to make deeper deductions.

Post Revisions:

This post has not been revised since publication.