Few genres of game are so closely linked with very deliberate design like puzzle games. Associated with intricately connected mechanics and components, arranged in such a way as to facilitate a single, solution.
It’s worth noting that the concept of a “puzzle” in games really can be stretched across a spectrum. On one end are open-ended challenges, which are problems for which the possibility space of the answer is extremely open. A camp of Bokoblins in The Legend of Zelda: Breath of the Wild is an open-ended challenge where the player can use any valid approach to solve it. All the way on the other end are pure puzzles, where there is a single state of the puzzle which qualifies as success. I call these pure puzzles because this category comprises the non-game things we traditionally consider “puzzles”: there’s only one correct configuration for a jigsaw puzzle, and you can’t just put any words you want into the New York Times crossword.
Turn-based strategy games, meanwhile, exist somewhere in the middle of this spectrum. Generally speaking, the number of atomic moves available to the player in such a game are limited (ex. instead of movement being a single fluid motion, it’s usually divided into large units like spaces, which limits the number of distinct options available). This means that there are only so many valid combinations of moves that will yield success.
However, many turn-based strategy games incorporate some amount of random chance, most famously in XCOM‘s “to hit” probabilities. This inclusion creates distinction between turn-based strategy games and pure puzzles, since random chance means that the exact same set of moves can yield both success or failure, depending on the way the dice land, whereas the solution to a pure puzzle is always its solution. If you put the correct words into a New York Times crossword, there’s not some chance that you’ll be wrong still.
I mention all of this because Into The Breach, the latest game by Subset Games, is commonly referred to as a turn-based strategy game, but I’d argue there’s something to be learned by thinking of Into The Breach as crossing the boundary into becoming a pure puzzle game.
If you’re unfamiliar, Into The Breach is a grid-based, turn-based game in which the player controls a small squad of giant mechs attempting to protect the civilian populace from a race of giant insects called the Vek. Unlike many similar games, Into The Breach does not have “to hit” chances, and enemies always broadcast the exact move they are going to perform next turn. Into The Breach is not a game of perfect information (buildings have a certain percent chance to resist destruction upon being hit, and while you can see where enemies will spawn next turn, you can’t see what they are). However, it’s much more open about information than other games which are mechanically similar.
In fact, the designers themselves have noted that this seems to push their game into the territory of being a puzzle game. “It’s very fair to say that Into the Breach is a puzzle game wrapped up in a strategy game,” said game co-creator Matthew Davis in an interview with PC Gamer. “But we just kind of stumbled into that design as we went into it.” Intentional or not, Into The Breach‘s deterministic gameplay meant that the AI of this game needed to be built with incredible care.
The AI in non-pure puzzle games, after all, is equally as much as beneficiary of random chance as the player. The need to ensure an game’s AI is beatable is, at least somewhat, lessened by the presence of random elements, as what could otherwise be an unbeatable AI will occasionally be pulled down by random chance, as will a defeated player occasionally be lifted up by random chance. In this way, an AI can exist in somewhat of a range around “perfect difficulty”, whatever that means for the game at hand, because random chance can pull an imperfect AI towards that sweet spot.
Into The Breach‘s AI gets no such help. Since there is little to no random chance in the game, that means that every turn, the game’s AI must place every active enemy in such a place, and get them to perform such an attack, that the player, with their current position, abilities, and mech statuses, can complete the turn with minimal permanent damage.
While it is, I guess, possible that the game’s AI moves the Vek across the board without taking all of these factors into account, having played a lot of rounds of Into The Breach, I can say that the game so consistently strikes a balance of difficulty that if it doesn’t take all of these things into account, it’s a damn impressive fake-out.
Instead, it seems more likely that the game’s AI algorithm is looking at the player’ mech abilities, placement, the location of key objectives, Vek health abilities, the map itself, and probably yet more things. It is taking all of these things, performing some math stuff on them, and generating a puzzle, one for which there is at least one solution, which isn’t too easy to derive (or else it won’t be a satisfying challenge), nor too difficult to get (or people will get frustrated). Striking the balance, without having random chance in the wings to give your difficulty balance a nudge in the right direction, with such a small set of options available to the player to solve the problem, is hard.
So, I’d be interested in really seeing the AI algorithm driving Into the Breach, because what it does is really something special. Subset Games have essentially generated an engine which generates pure puzzles on a regular basis, an algorithm which spits out fairly complex and interesting puzzles with a set of pieces, some of which are outside of the algorithm’s control (the AI gets to pick what Vek spawn, but not what the player brings in to the fight). I hope this game inspires other designers to adopt this sort of “dynamic puzzle” mindset, because clearly, the idea of trying to design every moment of gameplay as a puzzle to be solved can occasionally produce some amazing games.
Now, excuse me, I’m going to go back to trying to finish my 2-island run with the Blitzkrieg.