Let's Dance: Pattern Recognition in Games

Introduction

Inspired by Raph Koster’s book A Theory of Fun for Game Design, I want to explore the process of pattern recognition in games.

In machine learning area, pattern recognition focuses on the recognition of patterns and regularities in data. There are two stages in pattern recognition: training and matching. Considering the interactivity of games, I prefer to add one more stage into this process – feedback. Now that the concept has established, let us figure out how it is applied in gaming.

Fair action games are just like rhythm games

The simplest example of pattern recognition comes from the rhythm action genre such as osu! or jubeat, in which you need to press the key corresponded with the indicator showing on the screen at the perfect time. Indicators will tell you which key to press or hold clearly so there is no difficulty in recognizing pattern or matching stage, whereas the challenge in rhythm games is about the reaction juncture, how quickly and precisely you can perform to follow with the music, or as I mentioned above, the feedback stage.

Here is a more complicated instance in Souls Series. When I try to fight a boss in Dark Souls, I find it like dancing: each boss has a distinct tune in it, all I have to do is to observe, dodge and slash back in a particular rhythm. Unlike other fast-paced action game, Souls Series have stamina restriction, and you can’t strike with many combos. Many players suffer from this hardcore gameplay not because they can’t act fast enough, but for the ambiguity of enemy’s action. There are many counterintuitive behaviors in boss’ action design and even in level design. You may think you have seen through boss’ moves, but that’s a death trap. You died many times to learn the pattern, and then you can predict boss’ next action correctly and make a suitable reaction. This is called training and matching, as known as pattern recognition in its original meaning.

In NBA 2K Series, in addition to character’s own ability, shooting percentage will also be influenced by the timing when the player pushes and releases his shooting key. As we all know that we should hold the ball upwards and throw it out when the ball reaches the highest point, it is a good metaphor of real-world shooting, a meaningful map between games and reality. In order to decrease the difficulty of finding best shooting timing, the shot meter is added to later games of the series, just like the indicator in rhythm games.

Pattern recognition on macroscopic scale

Those discussed above are actions on a microcosmic scale, whereas when we look at a larger scale, pattern recognition existed in a different way called tactics or strategy.

Why would we send an SCV to scout opponent’s base at the very beginning of a Starcraft match? To collect valuable data through which we can infer enemy’s development routine and employ pertinent tactics to win the game. The existence of fog of war ensures such information absence, and an experienced player will know the importance of hiding a sub base or even deceiving and misleading the rival in a wrong direction. Of course, you need to play a lot to gain that experiences and be familiar with common patterns, it’s called training through practices. In most RTS games, there is a greater war over local intense conflicts, a war centering on information discrepancy, which needs more global perspective rather than micromanagement.

What is the chess manual for? Except for basic rule examination, there are many more patterns in it, patterns that have ever emerged through prolonged history since chess was invented. I’m certain some of these patterns are never considered by the rule maker, and there are more patterns waiting to be discovered. As simple rules deriving diversified gameplays (patterns), it is called elegant game design.

There is a genre of strategy-based games called 4X, its name itself is a sketchy description of a general pattern in this genre: exploration, expansion, exploitation and extermination.

Out of games, we can take cartoon or anime into consideration, which is widely accepted as a children-oriented media, while movie and TV show is mostly for adults. It is a prejudice with no doubt, but nevertheless, there must be some reasonable explanations about this. If we use pattern recognition view into this issue, we can find that comprehending cartoon graphics is way easier than movie graphics because cartoon omits many unnecessary details and is usually endowed with more stylization, it is essentially an abstraction of the real picture.

Conclusions

Unlike any traditional media, games have two more stages in the pattern recognition process: training and feedback. Different games emphasize different stages and challenge players in particular ways. Only through this interactive media can we experience a complete procedure of pattern recognition, and whichever the stage is, it has its own fun.

As intelligent creatures, we have used to recognizing patterns in everyday life and are restricted in many aspects, but we can always be whoever we want in games. Just like Dutch historian Johan Huizinga once said, “civilization arises and unfolds in and as play”.

Appendix

The author isn't native English speaker so I apologize for any grammar mistake that exists in this article. If someone interests in its original language version, here is the link.