Three Types Of Patterns

The Paradigm patterns that I've designed or played so far can be roughly divided into three distinct categories, which I call "tile-based patterns", "group-based patterns", and "picture-based patterns". These terms refer to the "grain-size" of a pattern - i.e., the size of the "atomic elements" out of which the pattern seems to be made. Tile-based patterns tend to have rules that explicitly refer to individual tiles. Group-based patterns tend to contain whole groups of tiles which act as atomic units. And picture-based patterns tend to have large-scale visual elements that span the entire pattern; in extreme cases, the pattern itself seems to be one atomic unit.

Let's take a look at an example of each type of pattern. (If you want to download one of these pattern files to view it in Paradigm, right-click on the image and choose "Save Link As...")

A Tile-Based Pattern

In this pattern, there are a number of black blocks scattered across a grid. (These blocks are not game-tiles that players need to guess—they're visible when you first load the pattern.) The defining rule of this pattern is that if any tile has exactly one black-block neighbor (including diagonal neighbors), it's purple; otherwise, it's green. I categorize this as a "tile-based" pattern, because the rules of the pattern refer to individual tiles (and their individual-tile neighbors).

A Group-Based Pattern

This pattern is made up of twelve interlocking "L-shapes" in four different colors. There are an equal number of L-shapes of each color, and no shape is touching another shape of its own color, even diagonally. I categorize this as a "group-based" pattern, because the rules of the pattern refer to groups of tiles which act as atomic elements. Notice that a pattern's groups don't all have to the the same shape or size, nor do they always need to be made up of contiguously-connected tiles of the same color. A purple tile surrounded at the corners by a red, yellow, green, and blue tile would be a perfectly legitimate "atomic unit" on which to base a pattern.

A Picture-Based Pattern

This pattern is, of course, a happy face. I categorize this as a "picture-based pattern", because the entire pattern feels like one atomic unit. Although the pattern does contain groups of gray tiles, these groups don't feel like atomic units in the same way that the L-shapes do in the previous pattern.

Fuzzy Boundaries

Notice that there are fuzzy boundaries between these three categories. A pattern might be predominantly tile-based, with a few higher-level groups thrown in. Or, it might consist of groups which are so large that they make the pattern feel like one big picture. The key concept here is that every pattern has a kind of "grain-size" which can roughly be categorized as small, medium, or large.

Which Patterns Are Best?

Although I don't want to discourage experimentation with any of the three pattern-types, I will say that, although I've enjoyed most of the picture-based patterns I've played, my favorite patterns have almost all been group-based, and I've only marginally enjoyed the tile-based patterns I've played. I've been trying to figure out why this is the case. Here's what I've come up with so far.

First of all, group-based patterns tend to offer the best "reward schedule". When playing a group-based pattern, you usually figure something out very quickly—and figuring things out is what makes the game fun. If you were to play the group-based pattern displayed above, it would not take you long to figure out that the pattern consists of "L-shapes". And, as silly as it may sound, it feels good to figure that out.

Of course, there's more to figure out after you've identified the L-shape aspect—and that means you'll continue to have fun as the game progresses. You'll begin to look for higher-level patterns in the arrangement of those L-shapes. When you uncover a new red tile, you'll "know" that there must be two more red tiles nearby, but you won't necessarily know exactly where. This provides a good mix of knowledge and uncertainty. You won't feel totally lost, but you've still got plenty of things to figure out.

Notice that, even if you knew all of the pattern's rules, you still wouldn't necessarily be able to run the entire thing. I believe that this is a strength of group-based patterns, not a weakness. Solving a group-based pattern requires a healthy mix of induction, deduction, and odds-calculation. By applying all of the rules that you've been able to glean, you may be able to make some surprisingly accurate guesses. For instance, if you've figured out that there are exactly three L-shapes of each color, you may be able to deduce exactly where a certain uncovered shape must lie.

Contrast all of this with the experience you're likely to have while playing the tile-based pattern. You probably won't figure anything out about that pattern right away. In fact, there's really only one thing to figure out, and until you do, you'll probably feel relatively lost. Furthermore, it will be difficult not to view the pattern as a collection of purple and green groups, which is a red-herring in this case. This pattern doesn't provide a very satisfying reward schedule; in fact, it's quite possible that very few players will figure anything out about it, even after seeing the whole thing. I've never tried this pattern on anyone, but I suspect that it wouldn't be very much fun to play.

Picture-based patterns lie somewhere between these two extremes. They don't always provide as many small rewards as the group-based patterns seem to, but you usually feel like you're getting somewhere with them, so they can still be quite fun to play. Additionally, these patterns can sometimes provide you with a powerful "aha!" moment, when you suddenly feel that you know what the rest of the pattern must look like.

One problem with picture-based patterns is that they often incorporate some form of symmetry. Symmetry in patterns tends to create what I call the "unhappy-seat" phenomenon. One player guesses incorrectly; the next player can now use the new piece of revealed information to get a few easy points, due to the pattern's symmetry. The next player is now in the unhappy-seat, with no good guesses. If he or she guesses incorrectly, the next player gets a few easy points, and then the player after that is in the unhappy-seat... and so on. Of course, any time a player in the unhappy-seat makes a good guess, the tables are turned. This dynamic can be fun or annoying, depending on the mood of the players. Picture-based patterns with symmetrical elements can be fun, but I tend to prefer the ones that have no obvious symmetry.

Conclusions

My intent here is not to discourage experimentation with tile-based patterns. In fact, I'm fascinated by the challenge of finding some styles of tile-based patterns that really work. For instance, I haven't even begun to explore the world of one-dimensional sequences—patterns which are essentially one long string of tiles. This might turn out to be a fertile genre.

Nevertheless, one of the conclusions of this essay is that, if you're just starting to experiment with pattern design, group-based patterns are probably the easiest to start with. Just come up with some atomic elements that you want to build your pattern out of, and figure out some interesting, patterned ways of putting them together.

Regardless of what style of pattern you're trying to design, I recommend that you always keep the "little rewards" concept in mind. If your pattern has the potential to generate multiple little moments of discovery throughout the game, you've probably got a winner.