9. Flow and the Fundamental Game Design Directive

Boredom is the conviction that you can’t change... the shriek of unused capacities.

—SAUL BELLOW, THE ADVENTURES OF AUGIE MARCH

Shadow of the Colossus is one of the most critically acclaimed games of all time. In it, you play as a young man who must single-handedly take down massive ancient colossi. The battles are multistaged and intense, which leads you to a real sense of accomplishment and conquest at the end. After each fight, you must travel by horse through a largely barren and unpopulated landscape in search of the next colossus. Why did the designers choose to have these quiet, peaceful rides in between heart-pounding boss fights? The answer requires you to understand how you stay engaged in games. This answer also leads you to a fundamental directive in game design.

One of the universal problems facing game designers is that you are never entirely certain of who your audience will be. If you knew precisely that your audience was Jim from Nebraska, age 31, no kids, introvert, master’s degree in philosophy, then you could base everything you made around Jim’s abilities and desires. Jim is not skilled at puzzles, so you keep them easy and to a minimum. Jim has great twitch skills, so you make jumping super difficult in your game.

Of course, the bang for your buck in making a game for one person is quite limited, so you don’t do that. Instead, you attempt to deliver enjoyable experiences to as many people as possible. And in so doing, you are forced to deliver content to people who may have contradictory skills: Jim is bad at puzzles, but Joan is excellent at them. They have to play the same game. If the game is too hard for Jim, he may give up in frustration. If the game is too easy for Joan, she may give up out of boredom. Where that beautiful middle ground exists is known as flow.

Game Flow

Psychologist Mihaly Csikszentmihalyi proposed the concept of flow to explain anecdotes of artists and other creative types becoming “lost in their work”—they became so immersed that they lost track of what was going on around them.¹ Flow is a state of focus and concentration on a task that is intrinsically rewarding. Csikszentmihalyi’s research suggested that flow transcends cultures and is one of the key components to happiness itself.

¹ Csikzentmihalyi, M. (1991). Flow: The Psychology of Optimal Experience (Vol. 41). New York: HarperPerennial.

The three conditions that must be met for a person to achieve flow are as follows:

• The person must be involved in an activity with a clear set of goals and progress.

• The task must have clear and immediate feedback so the person can adjust his actions as needed.

• The person must balance the perceived challenges of the task with his perceived skills. He must see the task as neither too easy nor too hard.

Game designer Jenova Chen noticed that Csikszentmihalyi’s features of a flow state were similar to what players of video games describe as being the conditions of fun, and he proposed his master of fine arts (MFA) thesis based on trying to elicit the flow state using game design.²

² Chen, J. (2007). “Flow in Games (and Everything Else).” Communications of the ACM, 50(4), 31–34.

Chen focused on the relationship between two elements: the challenge created by the game and the abilities of the player (Figure 9.1). Flow is a harmonious balance between these elements. As the player’s abilities increase, if the game’s challenge does not also increase, the player enters a state of boredom in which the game has become too easy. As the game’s challenge increases, if the player’s abilities do not correspondingly increase as well, the player experiences anxiety or frustration. The designer’s job is to craft an experience so that the player is always somewhere between anxiety and boredom, and they have to do this beforehand, without knowing what the player’s skills will be.

Figure 9.1 Being over- or underchallenged relative to skill causes problems.

Different audiences find flow in different areas. Hardcore players demand a higher level of challenge per their level of ability and become bored easily, whereas casual players tend to require lower levels of challenge because they become frustrated more easily. As a result, casual games most often err on the side of not providing enough challenge for players, whereas hardcore games err on the side of providing too much challenge for players. The same game can elicit flow at one point and fall out of it at another point. You may have experienced this before: Either you have played a game that took some time to “click” and was woefully frustrating at first, and then you finally settled into a rhythm, or you have played a game that was pleasantly interesting for a while, but it quickly became painfully dull.

Chen’s game flOw is a result of this research. Its main feature is a system of dynamic difficulty adjustments. It detects how the player’s abilities are increasing and adjusts the difficulty to compensate. When the player is failing, the game seamlessly adjusts to provide less challenge. When the player is succeeding, the game subtly increases the challenge. The goal is to keep players in a state in which their skills are a close match for the level of challenge.

Games have long been structured to take advantage of this phenomenon. Games generally start off with easy introductory levels and then slowly add mechanics that interact with each other to increase difficulty.

Look at World 1-1 in Super Mario Bros., for example. Imagine you are like many children of the 1980s—you have never played a video game before experiencing Super Mario Brothers. Thus, World 1-1 requires you to experience a low challenge to match your low skill level. In fact, World 1-1 has to teach you so many things right off the bat that it requires only a small set of obstacles.

At the start, the game provides you with a relatively safe environment in which you can experiment with the controls. The first screen has no enemies or interactions at all. Here, you just learn how to maneuver Mario. Shortly thereafter, you encounter your first Question Mark Block and your first enemy. You do not have to be told this is an enemy—the menacing eyebrows of the Goomba give it away. You then have to jump on or over the Goomba to pass. For many, especially those who are playing a video game for the first time, this is a challenging maneuver. For players who have been playing games like this for most of their lives, however, this is trivial. The level introduces only three other hazards: a small pit, a turtle (with his shell), and a piranha plant. For some this is too much, but for many, this teaches them the skills they need to adjust to the increasing challenges of the game.

Now take a look at the last levels of the game. By this point, the player has mastered the control dynamics of Mario to an expert level. As a result, the challenge must increase to a correspondingly high level. If every level were as easy as World 1-1, most players would get bored with it by the end. But the designers took this into account. In the later levels, players must bounce off of moving bullets, land on single-block platforms, use springs, and dodge thrown enemies.

Designers use a lot of effort to match the challenges they present to the player’s current skill level. Modern AAA studios use sophisticated techniques to find out how to avoid areas that are not conducive to flow. Microsoft’s user research arm helped Bungie produce visualizations when creating Halo 3 so they could see areas where players were spending more time or were using more ammo.³ These visualizations help designers see bottlenecks where players are getting stuck and areas where challenges are insufficient. Perhaps a jump or door was not obvious. By measuring player behavior, you can see where players are struggling (or perhaps not struggling enough).

³ Thompson, C. (2007, August 21). “Halo 3: How Microsoft Labs Invented a New Science of Play.” Retrieved June 30, 2019, from http://archive.wired.com/gaming/virtualworlds/magazine/15-09/ff_halo?currentPage=all.

By measuring player movement over time and using other types of quantitative and qualitative information, designers aim to find out where players are in flow and where they are bored or frustrated. By knowing that the flow state exists and that its boundaries are boredom and anxiety, you are aware of one of the simplest tools for generating happiness. Bored? Add more challenge. Frustrated? Remove challenge or increase player skill.

Now I will return to the question of Shadow of the Colossus. Why would the designers follow extremely challenging boss battles with long periods of challenge-free riding? If you are not familiar with that title, you can choose almost any game with a difficult boss encounter, because many share the design pattern of a difficult section immediately followed by an empty or easy section. You’ll rarely see a boss fight followed by another difficult section. Why is this? The designers are playing with flow. The flow channel is the area on the graph of a player’s flow where the player is neither frustrated nor bored. The goal is to get the player to bounce off the edges of her flow channel (Figure 9.2). After she experiences something difficult, the player needs time to breathe and readjust. Thus, she gets a period of easy play that brings her back down toward the boredom side of the flow channel. Before she gets there, though, the developers reintroduce the challenge again, even harder than before, to deal with the player’s increased mastery. It’s all flow.

Figure 9.2 Oscillating between the boundaries of the flow channel.

Game designer Richard Lemarchand posed an alternative way to view the ebb and flow of game difficulty. His position is that the amount of attention a player can devote at a specific time is limited; periods of high vigilance must be followed by periods of relative unimportance to restock reserves of attention.

Games that do not have an ulterior motive for players besides player satisfaction (for instance, simulations, art games, and games for training all have goals besides simply providing players with fun or satisfaction) should primarily be driven in all decisions toward providing players flow. Most game design heuristics revolve around getting players to a flow state. This principle is so overwhelmingly present under the surface of so much of what we label “game design” that I am willing to call it the fundamental game design directive.

Let’s consider the “why” behind other commonly held game design heuristics that you’ll find in this book and elsewhere:

• WHY SHOULD YOU PROVIDE PLAYERS WITH MEANINGFUL DECISION-MAKING OPPORTUNITIES? In the absence of these, the players have little to challenge them into a flow state.

• WHY SHOULD YOU PLAYTEST YOUR GAMES? You don’t have a “flowometer” that you can attach to your games to record the quantitative amount of flow in which a player engages. Instead, you have to test unbiased subjects in a scientific way to see if players are bored or frustrated. Until that flowometer is invented, playtesting is the most reliable way to gain that information.

• WHY SHOULD THE PLAYER HAVE CLEAR GOALS? Clear goals allow directing player behavior toward his flow state. Aimless behaviors become boring over time. Behaviors directed toward impossible goals become frustrating.

• WHY ARE RIDDLES GENERALLY POOR PUZZLES? Most riddles have no way of working toward the answer. Thus either the solver knows it right away and the challenge is trivial, or the solver has no way of determining the answer and the challenge is frustrating.

• WHY SHOULD YOU CARE ABOUT WHAT THE PLAYER PERCEIVES AS FAIR? If a player is overly frustrated, justified or not, then the player cannot be in the flow state. What is important is not whether a game is truly fair or not but whether it is perceived as fair.

• WHY IS THE “NEAR MISS” IN A GAME PERCEIVED AS EXCITING? Psychologists have studied the effect of the excitement produced by near misses, even when the player does not have control over the result (for example, slot machines).⁴ One possible explanation is that a wild miss signals to the player that she must make great adjustments to get to a winning state, but a near miss suggests that she was on the right track. A player who feels she was on the right track is likely to feel an adequate amount of challenge (as opposed to an “easy win” or a miss devastatingly off target, suggesting that a win is impossible), and thus she is more likely to be in a flow state.

⁴ Reid, R. L. (1986). “The Psychology of the Near Miss.” Journal of Gambling Behavior, 2(1), 32–39.

All these elements and more can be phrased in terms of providing players with flow. Of course, if there were some simple algorithm for adding flow for players, then I would just spell it out and the book would be complete. Getting players to flow is not easy. Not only are skills wildly different from player to player, but players can also interpret the same input in vastly different ways. Thus, we all need to have a deeper understanding for what mechanisms we can apply in what situations to produce flow.

Interest Curves

In the book The Art of Game Design, designer, professor, and juggler Jesse Schell discusses a way to illustrate the concept of a person’s engagement over the course of an entertainment experience.⁵ He named this concept the interest curve. In an interest curve, you relate the amount of engagement you have with a piece of entertainment at various points in time. You then graph the relationship between the amount of interest/engagement/fun on the y-axis and the point in time on the x-axis.

⁵ Schell, J. (2008). The Art of Game Design: A Book of Lenses. Amsterdam: Elsevier/Morgan Kaufmann.

Say the graph in Figure 9.3 represents a session in which I play League of Legends. I start out with some level of interest (1); otherwise, I wouldn’t play. The game becomes more interesting as I ease into the early game and attempt to get my first tower (2). But then something bad happens. I get defeated and my interest plummets (3). I am feeling really bad. In fact, my interest is less than when I started. I have a hard time getting it back. But eventually I catch up and get some great assists, which skyrockets my interest (4). Then my team starts carrying toward victory, which sustains my interest (5). The end of the game serves as a climax, which spikes my interest again (6). At the end, if I am more interested than when I started, I consider starting another session.

Figure 9.3 An interest curve of me playing League of Legends.

Along this curve, a number of events affect my interest: the start, attempting to destroy the first tower, defeat, a period of time in which I languish and try to catch up, a big team fight or two, and then final victory. Mapping my interest over those events creates the interest curve. By examining the interest curve, designers can better understand which events increase engagement, which events decrease engagement, and when these events happen over time.

Each game has an interest floor that, when passed, causes the player to quit the game. Without constant interesting moments to spike up the level on the graph, players will slowly fade down to this quit point. Interest increases when a player is in flow and decreases when a player is bored or frustrated. Figure 9.4 shows the interest curve for a player who cannot figure out the rules of a game. He has some early success, which spikes the graph at one point, but frustration keeps pushing him lower until he finally quits.

Figure 9.4 At minimum interest, the player gives up.

Tracking a player’s interest may seem like an academic exercise, but this is actually something that many studios record when playtesting. Some studios survey players at random points in their playtests. The players then rate how much fun they are having at that given moment on a scale of 1 to 10. By doing this with a large sample of playtesters at many points in the game experience, the designers can craft an interest curve of how different players are experiencing the level.

By crafting interest curves, designers can identify the weak points in their experiences. Spots where interest wanes are normal (and even necessary—not all moments can be a climax or players would burn out), but extended periods of low interest should obviously be avoided. If you notice that interest never peaks again after a certain event, you may wish to save that event for your game’s climax. Or if a certain puzzle causes a large drop in interest, it may not be appropriate for the audience.

What does a “good” interest curve look like (Figure 9.5)? It looks surprisingly like the dramatic structure that Aristotle lectured about over two thousand years ago (Figure 9.6), explaining the structure of a story.

Figure 9.5 A good-looking interest curve.

Figure 9.6 Aristotelian structure.

In the normal course of a drama, you have an inciting incident that causes some amount of interest, then a series of spikes of rising action. The rising action has a series of conflicts and resolutions that keep upping the ante to the point of climax, when the hero’s major problem is resolved in one way or another, which then leads to the catharsis or “new normal.” Every well-crafted drama has ups and downs, little victories and little defeats that lead up to the pivotal moment. In games, you experience the same structure. For instance, a complicated puzzle has several aha moments when you discover the mechanics and how they interrelate before the final giant aha of solving the puzzle.

Figure 9.7 shows the number of deaths per map in Half-Life 2: Episode One (which Jesse Schell also references), taken from Steam’s publicly shared data.⁶ Each series represents a different set of difficulties. The map that spikes the highest is obviously the hardest difficulty. The maps are listed in sequential order from left (beginning of the game) to right (end of the game). As you can see, the game plays in a linear manner, so players will always encounter the maps in this order.

Figure 9.7 One way to interpret this graph is that deaths are related to challenge, and challenge is the interesting aesthetic of the game that regulates flow.

⁶ Accessible in summary at www.steampowered.com/status/ep1/ep1_stats.php.

In Figure 9.8, you see the design structure repeated. Half-Life 2: Episode One features a number of mini-climaxes at the end of each act. These correspond to high death rates since, in these types of games, the most interesting areas are the ones that provide significant challenge. After these mini-climaxes, a period of rest always occurs where the death rate is low. Then the stakes are raised again and the game’s interest jumps back and forth until the climax. With this structure, Half-Life 2: Episode One’s designers keep the player in flow as long as possible by matching the actual difficulty to the player’s expected difficulty for that point in the game.

Figure 9.8 A learning curve.

Learning Curves

Another type of graphical representation of player flow is what is commonly known as a learning curve. You may have heard of this term before. The popular usage of the term is actually backward. In common parlance, a game with a “steep” learning curve is a game that is difficult to learn early in the experience. However, a learning curve is a graph of learning on the vertical axis over time on the horizontal axis, so a game with a steep learning curve has the player exercise mastery in a short period of time.

The actual use of the learning curve dovetails nicely with the discussion of flow. If a player is constantly learning, then what impact does that have on flow? What if she learns nothing after a certain point? Since a flow diagram relates the amount of challenge needed for the amount of skill, a learning curve tells you how much challenge needs to increase over time. Look at the learning curve in Figure 9.8. A player learns some initial skills but stops learning after a point. If you know this, you know that to keep the player in flow, you need to increase the difficulty during time range A, while she is learning, and keep the difficulty constant during time range B, when she is not.

Now relate what you know about the learning curve to the interest curve in Figure 9.5 if you have a game in which a lot of learning has to take place before the first climax (the commonly referred to steep learning curve), then you risk having the player reach her minimum level of interest. Teaching someone is an investment, and if that investment does not pay off in time, the player/learner abandons the activity before the payoff.

For instance, in Dwarf Fortress, the player must learn an extraordinarily large number of controls, symbols (Figure 9.9), and mechanics before he is able to understand how to actively play with the simulation. This amount of learning delays the first big payoff of the game. Not surprisingly, many players never get to that first big payoff. They hit their minimum interest level and quit playing. Because of the large amount of early learning the player experiences, he stays on the frustration side of flow. Players can handle this for a time, but eventually they quit if they cannot break out of that state.

DWARF FORTRESS. IMAGE USED WITH PERMISSION.

Figure 9.9 the glyph-based display of Dwarf Fortress is just one element of a complex interface.

If you have to design a game like Dwarf Fortress, where the player has to learn a lot before the first payoff, what can you do? Using what you’ve learned from this chapter, you can do one of three things:

• INCLUDE EARLY MINI-PAYOFFS (MINI-CLIMAXES) SO PLAYERS BECOME INTERESTED WHILE LEARNING THE GAME. This is difficult to do and requires significant planning and playtesting to see if the events that you intend to be climaxes are satisfying as such. Many games introduce mechanics slowly and let each mechanic pay off before introducing a new one. A particularly excellent use of this technique is the game Portal, which methodically teaches new mechanics in a test chamber before the big story payoffs.

• REDUCE THE AMOUNT THE PLAYER NEEDS TO LEARN FROM THE GAME BEFORE SHE CAN HIT THE FIRST BIG PAYOFF OR CLIMAX (a subtle difference from the previous strategy). This has limitations in that it changes the dynamics of your game because you are limiting the amount of mechanics you originally intended on having. Many sports games include a “casual mode” that limits the amount of mechanics in play by automating some of the more advanced options. This allows players to get their first wins quickly, yet perhaps with not as much skill as when they are using the default modes.

• IGNORE THE SITUATION AND CATER ONLY TO PEOPLE WHO TOLERATE LONGER PERIODS WITHOUT MINI-CLIMAXES (for example, people who have a lower minimum interest level). This obviously limits the reach of your game, since you focus only on the players who have already reached flow with your game as is.

Upward-sloping learning curves mean that at those points, the player is learning while playing. In the book A Theory of Fun, designer Raph Koster posits that all fun is based on learning.⁷ This would dovetail with flow theory quite nicely. If players are challenged sufficiently, then they are learning how to best use their skills, and they are in flow. If players are learning—that is, if the learning curve is trending upward—then the players always have something to challenge them. This is, of course, one half of the equation. Challenge must exist, but it also must not be too challenging. It’s not enough to just provide opportunities for learning; those opportunities must match the skills of the learner. For college-bound students, an introduction to economics class makes sense. Those students have both the theoretic (algebra, geometry) and concrete (real-world experience) tools to understand the lessons. A kindergartner, on the other hand, would have plenty to learn from the class but does not have the tools to do so.

⁷ Koster, R. (2005). A Theory of Fun for Game Design. Scottsdale, AZ: Paraglyph Press.

Individual Differences

Remember that individuals have their own personal interest and learning curves. There is no such thing as an “interest curve“ or “learning curve” that is consistent across players and time. Remember that grandma may pick up match-3 games okay, but she will have trouble sustaining interest in a first-person shooter. Your stereotypical core gamer tires of Facebook games quickly but can play the same shooter for hours and hours.

Designers tend to overestimate player skill in the short run and underestimate it in the long run. Because you have intimate knowledge of your systems, you are likely to find it difficult to imagine what it is like for players who have never seen your systems before. Always underestimate what the player will be able to understand instinctually early in the game.

This reaches an inflection point for your expert players. When you say, “Yeah, the player will always win if she gets six crystals from this lake and takes them the entire way across the world to craft them, which no one would ever do,” you are underestimating your expert players’ skill, which is almost always a mistake. On day one of release, you’ll be horrified to see videos appearing on YouTube talking about “the crystal trick.” Never underestimate how a player will be able to break your systems. Always assume that at least one player will have complete knowledge of everything about your systems and complete skills to maximize that knowledge.

Summary

• Flow is the state in between frustration and boredom that drives player motivation. Generating flow is the fundamental directive of any game designer producing games for player enjoyment.

• Games must be balanced in a way that respects each individual player’s level of skill if the designer cares for generating flow in each individual player.

• By mapping player interest over time, designers can get a visual representation of a player’s engagement over different sections of a game.

• Learning is integral to the flow process. Complicated games are neither good nor bad; instead, what needs to be managed is how they expose their systems to the player to affect their level of challenge.

• You will always know and understand your own games more than the vast majority of players. Whatever works for you must be toned down for a player who has less understanding of and experience with your systems.