The Designer's Notebook: Tuning Puzzle Games for Non-Puzzle Gamers

I'm not very good at certain kinds of puzzles. It has to do with combinatorial thinking: I'm poor at chess, nine men's morris, or any other kind of game or puzzle in which you have to look several moves into the future. But there are a few puzzle games that I really like, and I began to wonder what makes them different. What was it about those particular games that let even a middling player like me enjoy them? How do you go about making a puzzle game attractive to people who aren't very good at puzzles?

You might think, "what's the point of even trying?" But the answer is pretty obvious: the more people who like your game, the better it will sell. The trick is to find a way to appeal to two different audiences, the serious puzzle gamer and, uh, the rest of us. I know it can be done, because it works for me with the right design and judiciously-chosen features.

Not all puzzles require you to look ahead. Usually the reason that you have to look ahead is that each move strongly influences the state of the puzzle and the availability of subsequent moves. Rubik's Cube is a classic look-ahead puzzle: every move changes not just one, but eight "sub-cubes" (the one in the center of the face that is turning is unaffected -- it only rotates in place). The classic sliding-tile puzzle is another, though not nearly as difficult as Rubik's Cube: moving one tile enables up to three other tiles to move into the gap it left, but no others. A jigsaw puzzle, on the other hand, is not a look-ahead puzzle. You can attach pieces together in any order, without any constraints over what you choose to do next. Lemmings and The Incredible Machine were both great computerized puzzle games that didn't require a lot of anticipation, so I enjoyed them immensely. They didn't give me nearly as much trouble as strong look-ahead puzzles like Rubik's Cube or Sokoban, so I won't be discussing them here.

Solitaire card games are essentially puzzles as well, although because they have randomized starting conditions they are not always solvable. For our purposes, I want to look at two: Freecell, which you probably know because it comes free with Windows, and a variant called 4 Towers. There are some small but critical differences in the rules that I believe make 4 Towers more fun for a non-look-ahead guy like me.

If you already know the rules of Freecell, skip this paragraph. The object of the game is to move all the cards from the board into one of four home cells at the upper right, one home cell for each suit, starting with the ace of that suit and working your way up to the king. The cards are initially laid out at random in a series of 8 columns, and only the cards at the bottom of each column may be moved. In order to uncover cards that are obscured, the card at the bottom of one column may be moved onto the bottom of another column provided that it is one lower in rank than the card it is going onto (i.e. a 6 may only be placed on a 7) and of the opposite color (red cards may only be placed on black ones). To help out, there are also four free cells at the upper left, which may be used as placeholders. They can hold exactly one card each and there are no restrictions on moving a card to a free cell. However, when a card moves out of a free cell, it must either move to a home cell or back onto a column, obeying the rank and suit rules already described. In addition, an empty column may also be used, either to hold a single card or to start a new column of cards. Any card may be placed into an empty column, so empty columns are very valuable.

If you compare Freecell to 4 Towers, you'll see that the home cells in 4 Towers are divided up on the left and right, with the free cells in the middle; but that's a purely cosmetic issue; it doesn't affect the gameplay. The real differences in 4 Towers are as follows:

On balance, I would say that Freecell is somewhat easier to win than 4 Towers; that exact-suit requirement in 4 Towers makes it pretty tough. So why does a non-look-ahead guy like me prefer playing 4 Towers to Freecell? Because of one more thing, something that's not so much a rule as a feature of the game's implementation in software. Freecell allows you exactly one "undo" move in the course of a game: if you make a mistake, you can correct it - once. If you make any other mistakes, you have to start over from the beginning. The game is easier in general, but less forgiving of error. 4 Towers, on the other hand, maintains a complete list of all the moves you have made back to the beginning of the game; you can undo, and redo, as much as you like.

This makes a huge difference. It allows the player to experiment, to try alternatives secure in the knowledge that if they don't work, she can try something else instead. It permits an empirical, investigative approach rather than a purely strategic one. Freecell requires you to do all your experimenting in your head, like a chess player. That's exactly the thing I'm not very good at. The rules are harder in 4 Towers, but the game gives me better tools to meet the challenge with.

Undo/redo is actually part of a larger issue, the question of reversibility. With a Rubik's Cube or a sliding tile puzzle, every move is reversible. Therefore, an undo feature is clearly not a violation of either the letter or the spirit of the rules. In solitaire card games, however, the moves are usually not reversible because they're restricted to particular conditions. In those cases an unlimited undo feature is outside the letter of the rules - but it can help to counteract particularly tough rules, as it does in 4 Towers.

Some people would probably say that being able to undo and redo as much as you like is just cheating - it certainly would be in chess. But that's part of the difference between a puzzle and a game. The object of a puzzle is to arrive at a solution from a particular set of initial conditions. You're not playing against another person, even a simulated one. Undoing against an opponent is cheating because the opponent is deprived of the opportunity to take advantage of your mistakes. Taking advantage of mistakes is part of how you win multi-player games, especially in a zero-sum game of perfect information like chess, in which there's no element of luck. But there's no fundamental reason to penalize the player for mistakes in a puzzle game; there's nobody else who should benefit.

If you want to offer a particularly tough challenge in your puzzle game -- "nightmare mode," to use Doom terminology -- then by all means turn off undo/redo in that mode. But if you never make it available at all, then you'll lose players like me. Undo/redo is an important part of making such games accessible to non-hardcore puzzle-players. That's why I play 4 Towers rather than Freecell. For you, as the designer, the choice is pretty simple: do you want my money, or not?

(The puzzle-master Scott Kim tells me that he feels undo and redo are rather tech-y features that will really only be familiar or comfortable to a tech-y audience, and that a mass market game needs to accomplish the same psychological end without them. This is the subject of some of his current research. He may be right there; since I'm using myself as a baseline, undo and redo are fine with me, but they probably aren't the ideal solution. All I know is they make the difference between me playing, or not playing, some games.)

Another feature that makes a look-ahead puzzle more enjoyable is save points. In many games you start off doing a certain amount of "housekeeping" -- easy, obvious moves -- before really attacking the heart of the problem. Sokoban, the game about sliding crates in a crowded warehouse, is a good example. If you get into an irretrievable mess (which you can do with just one slip of the finger), it's a real nuisance to have to start the level over from scratch. Most versions of Sokoban allow undoing, but some also allow the player to set a save point so he can restart quickly from a partially-complete state. Unlike action games, in which (for some people) part of the challenge involves doing a certain number of actions in a row without a mistake, with a turn-based game like Sokoban there's really no point in making the player repeat easy and obvious moves.

I have one last recommendation for look-ahead puzzles, and that is to avoid time limits. By all means keep track of the time spent if you like, and assign a score on the basis of the speed at which the player solves the game (Tetravex and Minesweeper, which come free with Windows, both do). But don't halt the game after a limited amount of time and make the player start over - not if you want to keep us non-look-ahead players.

Finally -- and this applies to any sort of puzzle game, whether it relies on look-ahead or not -- it's important to choose a compelling metaphor and to execute it attractively. Serious chess players can have just as much fun with a $2 plastic travel chess set as with a $300 inlaid board and pewter statues for pieces. To them, it's all in the mind anyway, so the presentation is utterly irrelevant. But for us non-look-ahead types, part of our enjoyment is derived from the aesthetics of the experience. Puzzle games are seldom known for their flashy graphics in the Final Fantasy sense, and it's certainly possible to go overboard with annoying sounds and distracting lighting effects. But just because it's a puzzle game - perhaps even a very simple one - doesn't mean you can ignore the niceties. The core gameplay of Lemmings didn't require all those different environments or those beautifully tuned tiny animations… but they were an essential part of making that game the smash-hit that it was.