[In this in-depth article, originally printed in Game Developer magazine earlier this year, Microsoft Game Studios user research expert Phillips looks at a simple but vital problem - why people stop playing games, and what feedback we can give them to encourage them to continue.]

Most games are challenging by design. Winning every time isn't fun, but neither is always losing. The typical user experience is somewhere between these two, with most players experiencing some degree of failure. Players will lose races, blow up, fall to their deaths, get lost, or fail in thousands of different ways. Some will persevere and continue to play, while others will get discouraged and give up.

As we game developers seek to expand our audiences, some traditional methods of keeping players engaged are becoming less effective (according to Microsoft's databanks, which I use as the main source throughout this article). Fortunately, we can do some relatively simple things to motivate players.

Keeping players motivated is difficult. The most popular solution is to manipulate the game's difficulty using tutorials, dynamic difficulty adjustment, player-selected difficulty settings, feedback systems, userfriendly controls, and in-game hints. The goal is to strike the right balance between difficulty and player ability, thereby always keeping the player within arm's reach of a new achievement.

Despite these attempts to balance difficulty for a wide range of people, the players will still experience failure. More importantly, many of these folks will stop playing because of these failures. It's rare for people to leave a restaurant because they don't like the food, and it's not too common for people to walk out of a movie because it's bad -- but game players do put down the controller and leave the game all the time. What's worse, when game players have a negative experience, they are likely to tell their friends, family, and community.

When someone quits a game prematurely, we haven't just lost a player; we've created a detractor.

Quitter Talk

How serious an issue is quitting? It's worse than you might have guessed.

Table 1. The percent completion is found by dividing each player's Gamerscore by the total possible Gamerscore for the title; those numbers are then averaged.

Table 1 shows the average Gamerscore completion for each of the top 13 Xbox Live games for 2008. The data was drawn from about 14,000 players. As you can see, even the games with the highest achievement completion rates (Fable II and Call of Duty 4) had players who, on average, attained less than half the possible Gamerscore.

This particular sample tends to be more hardcore than the average player, and I would expect the actual completion rates for the entire population to be lower than the numbers recorded here.

Of course, the Gamerscore tells only part of the story. Players could finish a game and do little else, resulting in a low Gamerscore but high completion rates. However, most games award achievement points for completing the single-player campaign.

Table 2. The bar graphs show how many players earned a campaign completion achievement -- in other words, finished the game -- for the titles listed.

Table 2 shows how many players finished a sample of the games listed in Table 1 as determined by whether they earned a campaign completion achievement (on any difficulty). For even the most popular games on Xbox Live last year, about 30 percent of players didn't play to the end.

Players don't finish games for many reasons, but no matter what explanations arise, it's also likely that a significant number of players stopped out of frustration and that is what we will discuss here.

What leads some people to persevere after experiencing failure and others to give up? Why do some people anticipate eventual success where others only see continued failure?

There are probably many answers to that question, some of which are out of the game designer's control. However, there are at least two things we can and should do. The first has to do with how we word feedback to players, and the second is related to the goals we provide.

The Psychology of Feedback

Research on motivation, primarily in education, suggests that an important factor for explaining how people respond to failure is their perception of why the failure occurred. Those who believe that their failure is the result of stable factors, such as native ability or intelligence, which they cannot easily change, are most likely to give up or not even try.

However, people who believe that a failure is the result of unstable factors that they can change through effort or strategy are more likely to believe they can overcome initial setbacks. The determining factor is the person's mindset about his or her ability.

The same holds true for video games. Players who believe they can learn and master the game persevere, while those who think they lack a particular game-playing ability, or that some other stable factor lies between them and success, are likely to quit.

Fortunately, this is susceptible to change. There are things we game developers can do to encourage a mindset that anticipates success rather than failure. But before we get to that, consider these two studies, both of which illustrate the simple and subtle means through which we can shape players' perception of ability.

M. L. Kamin and Carol S. Dweck (see References) conducted a study in which they had students take a difficult test. After the test, they praised half the students for being smart (the "ability" group) and the other half for their effort or strategy (the "learning" group). The participants were then given the choice of two new tasks to complete: a simple one at which they were likely to succeed but learn little, or a difficult task that would be more interesting but would likely result in mistakes. Most of the ability group chose the simple task, while the learning group tended toward the more difficult task.

As far as psychology experiments go, this was a very simple manipulation. The researchers merely changed a few words in their feedback, which produced significant changes in the students' attitudes.

In another experiment, two Stanford University researchers manipulated the attitudes of participants before a task. Craig Anderson and Dennis Jennings (see References) told half their subjects, prior to having them take a test, that their success on the test was likely dependent on innate ability -- either they had the ability to perform well, or they didn't.

The other half of the subjects were told that doing well was a matter of determining the right strategy -- anyone can do it, but it takes effort.

However, Anderson and Jennings designed the test so that everyone would initially fail (does this sound like a game you've played?). After taking the test once, the subjects were asked how they thought they would do on another, similar test. Those who were led to believe that success depended on strategy and effort were more likely to expect future success. Those who believed that success was a matter of ability did not.

Interestingly, this manipulation had an even more dramatic effect. The subjects with the strategic mindset were more likely to have monitored their own methods for completing the tasks so that they were able to modify them in subsequent attempts. That is, they were able to learn from their experiences. The participants with the ability mindset did not monitor their strategies and therefore did not learn as much as the other group from their experiences. In short, the manipulation affected both the participants' anticipation of success (or lack thereof) as well as what they learned when taking the test. 

Perception Management

Intuitively, these results make sense. If people believe that success is dependent on ability, then no matter how much effort they expend on the task, they believe they are going to fail again. When people say things like, "I can't cook," or "I can't draw," or "I'm no good at first-person shooters," they don't typically then sign up for a cooking class or begin carrying a sketchbook everywhere or practice playing Halo. "Why bother trying to improve if you don't have the innate talent?"

What is less intuitive, and what we need to leverage as game developers, is our ability to manage expectations and mindsets.

To do this, we have to change our mindsets as well. I strongly suspect that most designers spend very little, if any, time considering what the player experience should be in the 10 seconds between a failure event, such as dying or losing a race, and the moment when play resumes -- or worse, the moment the player quits. (There are some notable exceptions, including Team Fortress 2 and Call of Duty, which I discuss in the following section.)

Given the opportunities to keep players involved and motivated, it's unfortunate that game developers rarely take advantage of these moments. In fact, they might be the most important 10 seconds of your game. While we spend weeks creating a few seconds of a cut scene and hours perfecting a texture, we spend very little time considering and implementing appropriate feedback at those very moments when a player decides whether to continue playing (See also research summarized in "GDC: Top 10 Video Game Research Findings," by Jill Duffy.)

Let's take a quick look at some of the player experiences surrounding defeat. Historically, these moments have been brutal. The "game over" screens for most arcade games were terse, bordering on insulting: "You Lose," "Game Over," "You're Dead." How far have we come since then? Not very. Most games fade to black, switch cameras to provide a view of the corpse, or simply pop the player back to a save point.

But we still see ghosts from the arcades. Some of these are intentionally reminiscent of the past, though most are just thoughtless designs. Language such as "You Suck" (yes, I've seen it), "Failed," and "Game Over" encourages players to put down the controller and do something else.

I'm not suggesting we swap in touchy-feely or overly encouraging language. In fact, there's a fine line between providing appropriate feedback and being patronizing. What we should be doing is focusing on the player's actions and emphasizing improvement.

Examples of Feedback

There are several good examples of feedback in games that teach strategy. The most common are the in-game hints players get at appropriate moments, such as after a death. The Call of Duty series has been doing this for a while, leaving a message for players when a they die from a grenade in Modern Combat, for example, or are stabbed in World at War.

These messages ask players to focus on developing strategies to avoid dying in similar situations in the future. (Although in practice, these indicators can be frustrating as well, if the player sees a grenade death but feels it was unfair -- the message then becomes insult to injury.)

Less common are feedback systems that inform players about their improvement at the game. One great example happens in Team Fortress 2. When players die, a message informs them about how they did relative to previous attempts, for example: "On the bright side... You came close to your record for time alive as a Scout in that round." The message goes on to indicate how long the player lived that round and what his previous personal best was. This is excellent feedback.

We also may want to shift our thinking about tutorials. Tutorials should not be considered the 10 minutes of instruction players get when they first start playing, or the part of the game we develop at the last minute after we finish making the "real" game.

Tutorials (perhaps we should stop using that word, too, as instruction needn't simply be text-based information) are the game and should occur over the course of the entire experience. Players are constantly learning to play, right until the end, and we need to provide relevant and informative feedback to them.

When we think about the game this way, we force ourselves to think about what players need to know at each point in the game, when to deliver that information, and how to track the data we need to provide this feedback. Sure, the bulk of what the player needs to know to get up and running happens early, but most good games require players to learn and adapt throughout. We should be doing our part to feed players the information and encouragement they need to keep up with these changes (though the "how" of this could take up an entire article on its own).

Creating Goals

The second thing we need to get better at is creating goals for players. Research has shown that the goals created for people -- by teachers, bosses, parents, and game designers -- go a long way toward shaping their mindsets about prospects for success and how they respond to setbacks. Most relevant to game developers are two types of goals: performance goals and learning goals.

Performance goals. Performance goals (or outcome goals) represent the most prevalent goal type in video games. There are three defining characteristics of performance goals. 

First, either people achieve the goal or they don't. There is no middle ground. Examples of performance goals include finishing a level in a platformer, getting the "Overkill" achievement in Halo 3, or finishing a race in Forza Motorsport in under two minutes. Examples of performance goals outside games are grades in school, or medals at the Olympics. There is no reward for progress toward the goal; you don't get half a driver's license for denting only one side of the car.

Second, the criteria for success are typically not defined by the goal-seekers.

Third, performance goals are usually complex activities that encompass a variety of smaller component skills. Passing a driver's license test requires many different skills, such as understanding the rules of the road, parallel parking, and braking safely.

What's wrong with performance goals? While performance goals are pervasive in school, work, and games, research on learning and motivation has shown that they often produce perceptions of lack of ability as well as decreased motivation. This is particularly true in cases where rewards or praise are contingent on successful completion. Further, negative feelings resulting from failed performance goals are more likely when a person's perception of their ability is already low, as may be the case with novice game players.

Consider this example. A child gets an A on his math test (a performance goal) and his parents tell him how smart he is. Maybe he even gets a reward. These are good parents. It is a popular belief that rewarding and praising abilities in situations like this is good parenting.

However, this kind of feedback can also have negative results. If the child's parents have consistently rewarded him for his ability, and because his parents made their praise contingent on a performance outcome (success on tests), it may backfire in situations where performance is poor. He will view his failures, like his successes, as a measure or indicator of ability, and failure equals lack of ability.

Now consider a player who believes he has low game-playing ability. A performance goal, such as completing a level in a shooter, will lower his motivation to continue trying, if he fails repeatedly.

When the goal focuses on ability, and the individual believes he does not have that ability, motivation and performance suffer. To extrapolate from the research further, he is less likely to focus on strategies for improvement if he views success as being contingent on a skill or ability he doesn't have. 

I don't think we should remove performance goals from games. A lot of players enjoy these types of challenges, and most games are structured around activities such as levels, rounds, races, and so on. However, we should consider incorporating other types of goals into games, too, specifically those focused on learning.

Learning goals. In many ways, learning goals are the opposite of performance goals. While performance goals focus on ability, learning goals focus on effort. It's not so much about doing as it is about trying. Improvement and progress toward the goal is as important as success.

It's important to explain that learning goals are not simply smaller or more frequent performance goals. Rather, they involve a philosophical shift in thinking about how we reward player progress.

To illustrate the differences between these goals, consider the performance goal of finishing a level in a FPS. At a low level, a player typically has to cross some boundary that triggers the level completion event, or maybe has to reduce a boss' hit points to zero. These events either happen or they don't. Additionally, the player is usually rewarded -- the story advances, the player gets a new weapon, and so on.

However, despite whether they complete the level, most players will improve their abilities over the course of playing. Some will finish faster and experience fewer frustrations, some will take longer, and some will eventually give up, but most will show signs of improvement.

This is good stuff to call out. It stands to reason that adding learning goals -- which focus on the skills and abilities that, when improved, make it possible for players to achieve performance goals -- would enhance players' appreciation of their own abilities. All this takes is a little more time focusing on the journey, versus the destination.

Learning goals make people try harder, take more risks, spend more time on a task, become less discouraged when facing setbacks, and, in the end, succeed more frequently (also see the sidebar, above). Doesn't that sound like the kind of player we should be cultivating?

Measure for Measure

One likely reason we don't often incorporate learning goals is that implementing them into a game is more difficult and requires more thought than traditional performance goals. It requires breaking from molds and doing something new. It's much easier to pop up a "level completed" message, a story cinematic, or an "achievement unlocked" notification after the player hits a predefined milestone in the game than it is to integrate learning goals that reflect the improvements players make.

Only recently have games been tracking player data in a way that could support learning goals, which could also be a contributing factor. But most likely, we have simply been stuck following conventional wisdom about how we reward players and provide feedback.

There are ways to start implementing learning goals in your games. One of the easiest (and most likely to have a significant effect on player motivation) is to tell the player how he has improved. While this is not a goal per se, it provides the player the information he requires to track his progress and set his own goals, and also provides the foundations upon which you can build actual learning goals.

First, break the game down into component skills. What skills does the player need to be successful? Does he need to do double-jumps? Does he need to master aiming? Does he need to figure out how to counter an attack?

Learning goals should focus on behaviors or skills that, when combined, give the player tools to complete more complex activities. Of course, these also need to be skills or strategies that your game can track. For example, if the player needs to understand how to play with stealth, it might be impossible to track [understands the stealth system], but you could track [was hit by enemy] or [used crouch].

Feedback can help teach players more effective strategies.

Then display progress on these component skills to the player. Rather than listing how many times x or y event happened, communicate metrics that relate to improvement, much like the example cited previously from Team Fortress. The obvious places to display progress information to players are 1) at the end of a level, 2) when they pause or quit the game, and 3) when they die.

Better yet, display a progress chart that players can access whenever they want. One example from Gears of War 2 are the messages that appear as a player nears a new achievement.

Another example from Gears of War 2 is the "war journal" which keeps track of the player's current campaign status. There's no reason we couldn't put similar messages in other games to keep players informed about their progress in mastering basic skills.

Of course, people have their own motivations and mindsets that they bring to games. Some people have a learning mindset and are likely to focus on getting better at a game. Others prefer goal-based achievements and do in fact feel motivated by them. In both cases, players are likely to have some preexisting beliefs about their gameplaying abilities. However, the type of goals presented and the feedback they receive during both success and failure can have a significant effect on how they respond to those setbacks.

Through better feedback and goal-setting, we can encourage a mindset of competence, reduce frustration, and encourage players to play longer, try harder, and feel more confident about future gameplay challenges. 

Resources

Anderson, C. A. & Jennings, D. L. (1980). "When experiences of failure promote expectations of success: The impact of attribution failure to ineffective strategies," Journal of Personality, 48, 393-407.

Ames, C. & Archer, J. (1981). "Competitive versus individualistic goal structures: The salience of past performance information for causal attributions and affect," Journal of Educational Psychology, 73, 411-418.

Butler, R. (1987). "Task-involving and ego-involving properties of evaluation: Effects of different feedback conditions on motivational perceptions, interest, and performance," Journal of Educational Psychology, 79, 474-482.

Clifford, M. M. (1986a). "The comparative effects of strategy and effort attributions," British Journal of Educational Psychology, 56, 75-83.

Clifford, M. M. (1986b). "The effects of ability, strategy, and effort attributions for educational, business, and athletic failure," British Journal of Educational Psychology, 56, 169-179.

Elliott, E. S., & Dweck, C. S. (1988). "Goals: An approach to motivation and achievement," Journal of Personality and Social Psychology, 54, 5-12.

Kamins, M. L., & Dweck, C. S. (1999). "Person versus process praise and criticism: Implications for contingent self-worth and coping," Developmental Psychology, 35, 835-847.

Seijts, G. H., & Latham, G. P. (2006). "Learning goals or performance goals: Is it the journey or the destination?" Ivey Business Journal, 70, 1-6.