This paper attempts to give a systematic account of the ways in which
games stage conflict and the strategic implications of these conflicts.
This is done by presenting an incentive perspective which builds
on the basic assumption that players prefer to win (in the broadest sense).
This assumption is shown to be common in the game design literature and
its explanatory strength, as well as its problems and limitations, will
By discussing the implications of a view of players as maximizers the
paper shows how the related concepts of conflict and strategy highlight
often underappreciated facets of game design. Games, or at least
a large number of games, may be seen as structures of incentives inside
which players attempt to optimize their outcome. In single-player games,
the player plays against a hostile (or at least challenging) environment
and in multiplayer games, players may be either pitted against each other,
positioned as allies against the environment or placed in a relationship
somewhere in between. In other words, games involve conflict or indeed
an antagonistic relationship between competing powers. While viewing players
as optimizers has obvious limitations (one being that such an approach
generally has little to say about open-ended games without clear goals),
this paper is based on the assumption that such a view is “good
enough” to merit application as it opens analytical paths which
enhances our understanding of conflict in games. Below, I will firstly
argue that the assumption that player behavior is largely determined by
the game goals is a basic one in the game design literature. Secondly,
I discuss how games may be understood within the framework of economics
and describe a series of implications hereof. In particular, the section
stresses that the level to which a game can be said to be strategic depends
on features not directly related to the game’s genre or to the number
of players involved. Finally, I briefly discuss more general issues concerning
the application of the analytical perspective outlined.
The assumption that players optimize
“It helps”, writes Greg Costikyan, “…to think
of a game’s structure as akin to an economy, or an ecosystem; a
complex, interacting system that does not dictate outcomes but guides
behavior through the need to achieve a single goal: energy, in the case
of ecosystems; money, in the case of economics; victory, in the case of
a game.” (Costikyan 2002). Costykian here makes explicit an assumption
or approximation common in game design literature: That players attempt
(or may be thought of as attempting) to optimize their outcome (e.g. their
score) in ways that are directly, if not deterministically, related to
the incentives presented by the game. This level of explicitness is rare.
More commonly, the assumption that players work to achieve the game’s
goals seems so basic that it is not even mentioned. But nevertheless,
it is often evident. Richard Rouse defines gameplay as “the degree
and nature of the interactivity that the game includes, i.e. how players
are able to interact with the game-world and how that game-world reacts
to the choices players make” (Rouse 2005). According to Rouse it
follows that “In Doom, the gameplay is running around a 3D
world at high speed and shooting its extremely hostile inhabitants, gathering
some keys along the way. In San Francisco Rush, the gameplay is
steering a car down implausible tracks while jockeying for position with
other racers”. But of course this only follows under the assumption
that players work towards the game goals. Without any behavioural assumption,
the Doom player might as well be expected to turn around in circles
or continuously fire his weapon into the nearest wall.
Rolling and Morris, in their Game Architecture and Design, consider
it useful to assume that any strategy available to a player must have
both advantages and disadvantages: “If there’s only a downside,
no one will ever use that strategy so why bother including it in the game?”
(Rollings and Morris 2004). While admitting that the player may have priorities
not exclusively linked to succeeding in the game, Rollings and Morris
maintain that features which do not further the player’s success
will soon be abandoned by players. In a related argument, Juul is explicit
that “A bad game is one where the player is unable to refine his
or her repertoire or where a dominant strategy means that there is no
reason to improve the repertoire.” (Juul 2003). This implies that
players want to succeed in the game, since a game with a dominant strategy
need not be bad if only players preferred tapping buttons to match the
game soundtrack instead of optimizing their outcome [It also implies that
players wish to succeed only given certain conditions related to game
balance. I will not go into this discussion here.]
The above examples indicate that the notion of players as outcome maximizers
is prevalent among certain influential authors who have reflected on game
design. For some, it even seems too basic to deserve mention explicitly.
Much may be said against this view of player behaviour, and we shall return
to this discussion later, but to grasp the full implications of this view
it is worth turning to economics, a field which has this assumption at
Game design and economics
As Costikyan argued, there is a close affinity between economics and game
design. So close, in fact, that game design could arguably be considered
an application of economic theory and each game session a case of experimental
economics [Most obvious in the case of games with clear goals, and far
less obvious in “process-oriented” games (Egenfeldt-Nielsen,
Smith, and Tosca In press) such as MMORPGs which do not specify clear
goals]. Classical economics considers the case of an individual attempting
to maximize her outcome in accordance with her preferences. For instance,
this person (we shall call her Alice here for convenience) might consider
whether to save for her pension, taking into account her desire to actually
retire, the annoyance of needing to put aside money each month, her perceived
risk of not living until pension age etc. Though such considerations may
be complex, the optimization principles at work are fairly simple as Alice’s
choices do not affect the variables of the equation. In a sense, the task
facing Alice is structurally comparable to that of the player of the arcade
game Moon Patrol.
Figure 1 – Moon Patrol (Irem, 1982)
The player combats an environment which does not adapt strategically in
response to the player’s actions
In Moon Patrol, the player controls a purple space buggy avoiding
rocks and holes while battling aliens attacking from above. The player,
then, optimizes her outcome against an environment which does not adapt
to counter her strategy (although it does change according to her
actions). She can concentrate on doing her best without worrying that
someone or something guesses her intentions or adapts to her playing patterns.
Returning to Alice, instead of pension funds she is now considering whether
to donate to a charity. The charity in question will only be successful
if a certain number of people donate (otherwise the income will be lost
to administration). Thus, Alice might stand to lose the money, but her
donation might help inspire others to donate as their investment becomes
safer. The outcome maximizing course of action here is dependent on the
actions and perceptions of others. This optimization problem is the domain,
not of classical economics, but of the disciplinary branch known as game
theory (for a discussion of the relationship between economic game theory
and video games see Smith Forthcoming). Game theory is the study of the
outcome of conflicts between agents who are interdependent; conflicts
in which the outcome for participants depend on the aggregate of their
choices. We can say that game theory is about strategy, understood
as choices made based on assumptions of other participant’s assumptions
Alice’s charity dilemma is comparable to video games in which someone
(other players) or something (the environment) adapts to counter the player’s
strategy whether that strategy be real or perceived. In Counter-Strike,
for instance, the outcome of one’s choices depend clearly on the
actions of the opposing players whose actions are again likely to depend
on their assumptions about the actions of their enemies. We shall
return to this issue below, but so far we can see that the player of single-player
games with non-adaptive environments is engaged in an activity analysed
by classical economics, while the player of multi-player games in which
the opposition can adapt are in a sense doing game theory.
Does it then follow that the player implied in many reflections on game
design corresponds to the Homo economicus of traditional economics? To
a high extent, that does seem to be the case. But we need to consider
what the entities are in fact trying to achieve; what they see as the
goal. In classical economics, the agent is seen as an emotionless optimizer
who attempts only to increase his or her own utility, typically in the
form of profits. This agent does not (or should not) care about how others
fare; being neither envious nor gloating since this is inconsequential
to its own profit. There are numerous problems with this view of economic
behaviour (see for instance Frank 1988) but we can see that it would correspond
to the idea that gamers merely attempt to maximize their score, even in
multiplayer games. Thus, such a gamer would not care about whether she
was beaten to the finish line in Gran Turismo 4 but would only
be focused on her achievement, her completion time (relative to her personal
record, for instance). While probably rare in its pure form, such behaviour
is not unthinkable. In fact, actual players of multiplayer games may often
place value both on defeating other player(s) and performing well
in more absolute terms. But for present purposes we may wish to consider
them distinct modes of victory and they will be referred to as “social
victory” and “personal victory” in the following.
The incentive perspective
Understanding players as goal-oriented entities and looking at games through
the lens of economics means applying an “incentive perspective”.
In the following, I will describe what such a perspective may entail [The
proposed characteristics may be said to represent a “strong”
incentive perspective. One may of course analyze incentives based on different
assumptions and topics of interest]. Firstly, the incentive perspective
is one way of looking at video games. It is an analytical stance and as
such does not claim to be more complete or correct than other perspectives.
It is blatantly rule-biased and so tends to ignore audiovisual and narrative
concerns. But it is also a formalization of core assumptions in game design
and one without which general discussions of the relationship between
game design and player behaviour become difficult.
Let us be precise about the assumptions involved. An incentive perspective
(IP) assumes that players attempt to succeed in accordance with the goals
presented by the game (that their preferences are determined by the game).
IP considers as “player” any entity with unique preferences
and this entity can consist of several individuals (e.g. a team). For
analytical purposes the game environment in single-player games may be
thought of as a player which tries to hinder the success of the (human)
player. Also for analytical purposes, IP is only concerned with in-game
interaction and does not take into account the actual level of play on
which players may interact in various ways beyond what the game itself
specifies. Each player applies a cause of action (or strategy)
and this cause of action, in relationship with the features of the game
environment and the actions of other players determines the outcome for
each player. The guiding question of IP is this: Given the options available
to other players, how will a given player (as described above) act given
the incentives structure of a game?
No particular method or approach is inherently appropriate to answer
this question. But I will suggest here that an analytical approach inspired
by game theory. This approach has three focus points: Conflict type, number
of players and player interdependence.
In Pong, the conflict between the two players resembles that of
classical two-player games like Chess, Backgammon, and Kalaha.
One player’s gain is the other player’s loss. In the language
of game theory, this conflict is “zero-sum” as the combined
gains and losses equal zero [In Pong, of course, the sum of the
final scores does not equal zero. But in the end, one player wins
while the other player loses.] In two-player zero-sum games, outcome maximizing
players will not cooperate. The game simply gives them no incentives to
do so and thus, issues of trust and deceit do not come up. If Alice is
maximizing her Pong score and believes that her opponent Bob is
being equally self-concerned, she should not believe any promise from
Bob that he will hit the ball gently on the next rebound. She needn’t
consider whether Bob is trustworthy; she can simply assume that he is
In many games, players are placed in less antagonistic relationships.
In Joust, the two players are placed in a situation where the likely
outcome of a strategy depends highly on the strategy chosen by the other
player. While there are good reasons to cooperate, there is also a temptation
to attack the other player (Smith 2005). This is characteristic of non-zero-sum
games, games in which the combined score is dependent on the aggregate
of player choices. Pong would have been non-zero-sum if the game
had also somehow rewarded players for merely keeping the ball in play
(and would have had an issue of trust if the reward for eliminating the
other player exceeded the probable gain from keeping the ball in play).
A two-player non-zero-sum game need not have trust issues at all. Another
version of Joust, let us call it Harmonic Joust,
might not award any points for killing the other player and have the challenge
set up in a way in which the loss of the other player was more of an inconvenience
than it was an advantage (the advantage being having more potential rewards
for oneself). In Harmonic Joust, the players would have no incentives
to attack one another. Rather, the players would mostly be “incidentally”
sharing the same space while fighting their own personal battles against
the game environment. Of course, this is only true to the extent to which
the player applies personal victory criteria.
Number of players
The way a conflict plays out depends not only on its sum type but also
on the number of players involved. In game theory, a game like Moon
Patrol would be described as a “game against nature”.
Here, nature is considered a player albeit one which acts without intentionality
and assumptions. This is different from single-player games in which the
environment actually responds strategically (in the sense used above)
such as Far Cry. In Far Cry, the enemies have a number of
strategies and, from the perspective of the player at least, may be said
to adapt to a perceived threat. While two-player games are often relative
easy to model and explain, interesting things happen when more players
join. The crucial change from two-player games is that it may now, even
in zero-sum games, well be in the interest of players to form coalitions.
Figure 2– FourPong (SquareFuse, 2003)
FourPong (see Figure 2) is a four-player two-ball Pong clone
in which a player takes on three computer-controlled opponents. The rebound
angle of the ball is determined by the portion of the bat on which it
strikes, making players able to control the angle to a certain degree.
The computer opponents may not be sophisticated strategists and they certainly
are not sophisticated communicators but if the game were played by four
human players there would be various ways for groups of two or three players
to ally temporarily against one or two other players. To take another
example, players (or teams) in the real-time-strategy game Age of Empires
II may be playing a game type in which there will be only one winner.
Thus, no team shares any ultimate goal with any other team but in a three-team
game two teams may well be tempted to join forces to eliminate a common
threat. This is particularly likely to happen the moment one player significantly
pulls ahead of the opposition as losing players realize that fighting
internally would mean certain defeat.
Games differ as to the level of interdependence between players. In
fact, the level of interdependence may be said to correspond to the level
to which a multiplayer game is strategic. Certain early arcade games represent
the lowest possible degree of strategicness. For instance, the game Time
Pilot offers the option of a two-player game, but in this mode players
merely take turns to play and each player’s score is displayed in
the upper corners of the screen (see Figure 3). Thus, one player’s
actions do not directly influence the possibilities of the other player;
indeed the two game spaces are kept entirely separate. The only in-game
interaction is through the displayed scores which may prompt players with
relative winning preferences to alter their play (say, towards more chancy
moves if one falls far behind).
Figure 3 – Time Pilot (Centuri, 1982)
But also multiplayer games in which the players share the same gamespace
may differ significantly in the degree to which players are strategically
interdependent. In the game Super Monkey Ball, players control
monkeys encaged in rolling balls and must compete in various disciplines.
One of these is a race in which players must try to cross the finish line
quickly and may collect various bonuses and weapons/upgrades on the way.
Now, playing against flesh-and-blood opponents may certainly matter to
a player in terms of the gaming experience but in the Super Monkey
Ball race (as in most racing games) the space for strategic choice
is slim. One may choose to focus on collecting weapons and on attempting
to push opponents over the edge of the course, but for the better part
of the race, one will be attempting to cross the finish line as quickly
as possible. The existence, and the choices, of other players within the
gamespace, while not insignificant, have only a limited influence on a
player’s likely strategy. In this sense, Super Monkey Ball
corresponds closely (but not entirely due to the weapons etc.) to a 100m
sprint in the physical world. Sprinters do influence each other to a certain
degree but essentially they are attempting to cross the finish line as
quickly as possible.
In Age of Empires II, the player relationship is crucially different.
Here, the final outcome of almost any player choice depends on the choices
of the other players. Although certain strategies are virtually always
bound to be unsuccessful (such as remaining entirely passive, building
no structures) the game has a large possibility space in which the consequence
of a given choice is dependent on the other players. To give but one example,
an early attack using mounted units is often a powerful strategy, but
not if the victim has trained the inexpensive pikemen units which counter
cavalry well. Thus, whereas Super Monkey Ball corresponded to 100m
sprint, Age of Empires II correspond to a soccer match in which
the outcome of strategic choices can only ever be predicted based on assumptions
of the strategies employed by the opposing team.
It should be noted here, that the level to which a game is strategic
does not depend on whether it belongs among what is usually called strategy
games. Tekken and Jetmen Revival (see Figure 4), for instance,
are both action games with strong strategic elements.
Figure 4– Jetmen Revival (Crew42, 2003)
Each player controls a small aircraft initially positioned at one side
of the gamespace. While seemingly a simple game, different ways of scoring
and the limited nature of player resources (health, fuel, and ammunition)
pave the way for a significant number of strategies.
However, a game can only be strategic if it is a “game of emergence”
in Jesper Juul’s terms (Juul 2002), that is a game in which variables
interact dynamically (as opposed to the game being a predetermined series
of event with predetermined outcomes). Figure 5 shows four game types
placed on a continuum from “Unstrategic” to “Highly
Figure 5 – Game types placed on continuum based on level of player
interdependence or “strategicness”. Singleplayer games with
adaptive environments are strategic to the extent that the environment
adapts or learns. Their place on the continuum is a question of the specifics
of the AI.
As mentioned above, the strategy chosen by a player is a function of the
game incentives, the actions of other players and assumptions about the
future choices of other players. In all four game types shown in the figure
above, certain courses of action may be inherently better than others.
We could imagine for instance that Moon Patrol was designed in
a way that it would always pay to drive and shoot as fast as possible
(not the case in the actual game). This would make the playing experience
a monotonous one. But there is a special feature of multiplayer games
with high player interdependence: Certain designs may render unimportant
assumptions about the choices of the player. For instance, Age of Empires
II might be designed in a way in which player A was certain that
player B would play in a certain way. This would be the case if the game
had a “dominant strategy”, a course of action which, no matter
what other players do, is always the most powerful (see also Rollings
and Morris 2004; Smith Forthcoming). Such a strategy will lead to a “strategic
equilibrium”, a state where players will not change their strategies
and (again depending on the details of the design) the game will become
highly predictable. Strategic equilibrium can be illustrated by turning
to a “spatial” model of the classical game theory situation
The Prisoner’s Dilemmas [The reader unfamiliar with the Prisoner’s
Dilemma may wish to consult Robert Axelrod’s lucid description (Axelrod
Figure 6 – The Spatial Prisoner’s Dilemma
by Serge Helfrisch (www.xs4all.nl/~helfrich/prisoner)
A strategy is marked by a colour, the essential ones being blue and red.
In this simulation, thousands of cells are playing the Prisoner’s
Dilemma against each other continuously. Each round a cell chooses the
most successful of two strategies played in its immediate surroundings
on the previous round. The screenshots are from the advanced stages of
games, which differ only nominally in the size of the advantage awarded
to a cell playing blue against a cell playing red. Only within a very
small interval does the game lead to behaviour which is “unpredictable”
and where neither strategy dominates entirely. Both on the left and on
the right we have a stable strategic equilibrium in which no player will
change his or her strategy. Although there are other variables at play,
equilibrium situations result from by far the majority of possible game
“rule sets” illustrating how difficult game balance can be
The incentive perspective is a formalization of certain assumptions found
in game design discussions, the most important one being that players
wish to win. Similar assumptions lie at the core of other fields but have
been contested and often proven problematic when used to analyze actual
human behaviour. Not rarely, for instance, have “rational choice”
approaches in the social sciences been criticised for being deeply flawed
(e.g. Green and Shapiro 1994) while the classical rational agent model
of economics has been shown not to correspond too well with experimental
results (Kagel and Roth 1995; Frank 1988). There is little doubt that
understanding people as utility maximizing entities only is a problematic
and imprecise approach. In terms of games, it is also clear that players
simply do not always devote all their resources to achieving the goals
set up by the game designers. Real players can be explorative, artistic,
have reasons for losing a certain game on purpose, set up their own goals
etc. Thus, the question is not whether the incentive perspective is capable
of explaining or predicting all facets of player behaviour, whether it
is the capital-T Truth. It cannot and it is not. Of course, for game designers
it may work well as a guiding approximation without being completely “true”
and without such an approximation constituting disregard for the various
“rebellious” uses that games are also put to. Indeed, an incentive
perspective on player behaviour may serve as a minimal theory of players,
without which constructing enjoyable games becomes difficult but which
is not capable of serving as a complete guide to game design and which
can easily be misleading if thought to be a description of actual play.
For game scholars interested in the social dynamics of gaming and of
the actual patterns in player behaviour, the general goal-orientedness
of players remains a hypothesis which is curiously understudied. At present,
we have a very limited systematic understanding of the uses to which games
are put by their players, and indeed of the ways in which players interact
while playing. In light of the often surprising results that similar questions
have provided in related fields (such as television studies) these issues
seem worthy of more rigorous attention.
Initially, this paper argued that the assumption that players prefer to
win is common in game design literature. It was noted that similar assumptions
lie at the heart of economics and ideas from this field were briefly applied
to video games to sketch the analytical implications of viewing games
as incentive structures. In particular, it was noted that the relationship
between players in multiplayer games span a large spectrum and that games
vary greatly in the extent to which players are strategically interdependent;
that games may be placed on a continuum of strategicness.
Finally, the status of the “incentive perspective” was discussed
and it was noted that what may serve as a useful approximation to game
designers still remains largely unexplored territory in terms of the scholarly
study of player behaviour and interaction. At this point, the understanding
of the activity of the player (both theoretically and empirically) seems
underdeveloped within game studies. In particular, the relationship between
game design or structure and player behaviour is virtually unexplored
but likely to offer fertile ground for future studies interested in bridging
gaps between formalist analysis and behavioural approaches.
Axelrod, Robert. 1984. The Evolution of Co-operation. London: Penguin
Costikyan, Greg. 2002. I Have No Words & I Must Design: Toward a
Critical Vocabulary for Games. Paper read at Computer Games and Digital
Cultures Conference Proceedings, at Tampere.
Egenfeldt-Nielsen, Simon, Jonas Heide Smith, and Susana Pajares Tosca.
In press. Understanding Video Games. New York: Routledge.
Frank, Robert H. 1988. Passions Within Reason - The Strategic Role
of the Emotions. New York: W. W. Norton & Company.
Green, Donald P., and Ian Shapiro. 1994. Pathologies of rational choice
theory : a critique of applications in political science. New Haven:
Yale University Press.
Juul, Jesper. 2002. The Open and the Closed: Games of Emergence and Games
of Progression. Paper read at Computer Games and Digital Cultures Conference,
———. 2003. Half-Real - Video games between real rules
and fictional worlds. PhD dissertation, Department of Digital Aesthetics
and Communication, IT University of Copenhagen, Copenhagen.
Kagel, John H., and Alvin E. Roth, eds. 1995. The Handbook of Experimental
Economics. Princeton: Princeton University Press.
Rollings, Andrew, and Dave Morris. 2004. Game Architecture and Design
- A New Edition. Boston: New Riders.
Rouse, Richard. 2005. Game Design - Theory and Practice (second edition).
Plano: Wordware Publishing.
Smith, Jonas Heide. 2005. The problem of other players - in-game collaboration
as collective action. Paper read at DIGRA 2005 - Changing Views: Worlds
in Play, at Vancouver.
———. Forthcoming. The games economists play - implications
of economic game theory for the study of computer games. Game Studies.
Video games cited
Age of Empires II: The Age of Kings (Ensemble Studios, Microsoft
Game Studios, 1999)
Counter-Strike (Valve Corporation, Sierra On-Line Inc., 2000)
Doom (id Software Inc., 1993)
Far Cry (Crytek, Ubisoft Entertainment, 2004)
FourPong (SquareFuse, 2003)
San Fransisco Rush (Atari Games, 1996)
Gran Turismo 4 (Polyphony Digital, SCEI, 2004)
Jetmen Revival (Crew42, 2003)
Joust (Williams, 1982)
Moon Patrol (Irem, 1982)
Pong (Atari, 1972)
Super Monkey Ball (Amusement Vision, SEGA Entertaintment Inc.,
Tekken (Namco Limited, 1995) Time Pilot (Centuri, 1982)