The quarterbacking problem or “alpha gamer problem” is a well-known phenomenon in tabletop cooperative games in which one or more alpha player(s) dominate the decision-making of the game against the consent of the other players. This is a problem when the behavior is antithetical to either the designer’s intended experience of the game or the experience that the players want from the game. For related reasons, designer Alex Jaffe (2019) calls this problem “cursed.” To author and lifelong board gamer Eric Thurm (2019), the frustration of quarterbacking is that it removes the cooperation from cooperative games (emphasis mine):
“[Quarterbacking] drains away some of the thrilling benefits of cooperative games, which frequently posit a model for how to achieve something like genuine solidarity. Competitive games often replicate zero-sum, or at least difficult-to-navigate, systems, areas of life where you’re conditioned to take risks — real estate, financial markets, battlefields. Collaborative games can do the same thing, but in ways that suggest the possibility of shared struggle.”
Quarterbacking is doubly problematic for transformational games (e.g. educational and serious games) whose goal is to provoke extraludic (i.e. extending beyond the game) transformative experiences, such as gaining knowledge or experiencing “endo-transformative reflection” (i.e. reflections which result in transformations behaviorally or conceptually within the game). This paper provides a comprehensive review of all known solutions to the quarterbacking problem with examples and critiques of each design paradigm. Next, I propose Roleplaying as an underexplored design paradigm which can mechanically and socially address quarterbacking in a way that is compatible with enabling transformative play experiences. This paradigm is demonstrated through two games, one educational and one solely for entertainment, which apply the Roleplaying design pattern. Finally, I offer a brief case study to demonstrate how one could apply this taxonomy to design analysis.
Is quarterbacking a player problem or a design problem? Anecdotally, I and others have observed that players believe it to be a player problem, while designers see it as a design problem. On the surface, it seems like an issue with the players for two reasons. Firstly, not all groups encounter this dilemma. Secondly, the problem can be traced to a specific player’s behavior! Yet, consider this analogy: suppose your word processing software deletes your file if you press a specific but uncommon key, one which you have a habit of pressing accidentally. Is it the fault of the program if you delete your file? You’re the one who presses the key; not everyone who uses this software will accidentally delete their files — is this not your problem?
No, it’s not. It’s the responsibility of the software to prevent you from causing yourself accidental harm. It’s the responsibility of the designer to prevent players from creating negative experiences for themselves. Quarterbacking is a design problem.
More specifically, quarterbacking is a problem with consent. Telling other players what to do isn’t actually a problem if the other players want to be told. There can exist healthy player relationships of asymmetric power and knowledge, but only when all players consent to those roles. The problem of quarterbacking is not the seizing of decision-making, but the non-consensual abduction of it.
Sociologist Erving Goffman (1961) posits a model of “interaction tension” — a discrepancy between the expectation of interactional norms and the present social interaction. Evidence for this model in video game culture was found last year by Sebastian Deterding (2019) through interviews with German gamers who confirmed a sense of tension and especially disinvolvement when their spontaneous experiences misaligned with their expected social norms, such as the display of emotions during video gaming experiences. However, when norms and experiences do align, players report a sense of ease, which Goffman calls “euphoric”. In this way, the alignment (or misalignment) of social norms and expectations with the player’s reality influences the valence of the experience; it’s only a positive play experience if the player feels their social expectations are being met. Therefore, it is possible that part of the problem of quarterbacking is an interaction tension between the expectation of cooperative play and its spontaneous seizure, leading the tense players to disengage at the loss of the ease they would have felt from a more cooperative environment.
So how do we design for consent and euphoric ease? The easiest way, and indeed how it has been handled historically (as shown below), is removing the possibility of this dynamic entirely. By putting all players on equal footing, this asymmetric power dynamic is no longer a viable option of play.
The other half of the design space is tinkering with consent itself: allowing quarterbacking to happen but only in consensual ways. Though the full exploration of this design space is outside the scope of this work, the proposed space of Roleplaying as a new solution is a first step in getting us there. Through roleplay, we can unpack different relationships of varying amounts of power and knowledge, blurring the lines between explicit game rules and implicit social contracts, and in doing so get closer to this concept of designing for consent.
The Design Space
In addition to consent, there are several other axes of design worth considering. Veteran designer Raph Koster (2018) describes some of these, which he breaks down into two major categories: forms of parallelism and asymmetry (teams, roles, individual agency, and success metrics), and the sharing of resources and liabilities (risk, or the cost of trust, information and the information horizon, rules of exchange and ownership, and time). Since Koster is examining this space with respect to trust mechanics, he also notes that the quarterbacking problem occurs primarily in games with a moderate level of trust: games with minimal asymmetric mechanics, a mix of cooperative shifting roles and oppositional teams, overlapping but distinct roles, abilities to help others, and perfect information except against opponents. Koster recommends mitigating the quarterbacking problem with increased time pressure or a reduced information horizon.
Koster’s analysis alludes to the fact that cooperation and competition in games is a spectrum. Semi-cooperative genres include team games, dynamic partnerships, and partial partnerships. Yet, the more competition a game has, the less likely it will be to struggle with the quarterbacking problem, since quarterbacking is based on an assumption of shared goals. Allen and Appelcline (2018) define four genres of cooperative games: true or “pure” co-op, hunter games, overlord games, and traitor games. In these latter three types of cooperation, not everyone is on the same team, and quarterbacking is deterred by the presence of enemies in the midst. However, quarterbacking often remains an issue for the majority of players (who are on the same team) in hunter, overlord, and traitor games, which have a one- (or few-) versus-many construct, since the “many” are effectively a subgroup of true co-op players. Yet, the existence of an oppositional team is often enough to deter overt quarterbacking. It is therefore only “pure” co-op which strongly suffers from quarterbacking, but through my taxonomy of solutions I will discuss how the remaining genres avoid this issue (namely, by breaking the assumption of shared goals).
The Need for Learning: A Unique Problem to Cooperative Educational Games
Like cooperative games, multiplayer educational games (and other transformational games) also often include working together to solve a problem. However, in educational games, the goal is not only to solve the problem but for each player to learn how the problem is solved. For this reason, quarterbacking is very detrimental to these games because it can prevent more novice players from gaining the learning outcomes intended by the game’s design. Moreover, the need for learning in a cooperative game implies the encouragement of mentoring. That is, the designer’s intent for a cooperative educational game is often that the players will work together and learn together, teaching each other as they learn. In this way, cooperative play becomes an opportunity for dialogic peer mentorship. Therefore, solutions to the quarterbacking problem for educational games are limited in how they can use existing solutions, since most paradigms which restrict quarterbacking, as shown in this article, also restrict mentoring by preventing or discouraging table talk and the open sharing of ideas.
Indeed, Allen and Appelcline (2018) claim that the fundamental nature of cooperative games is to allow cooperation and then limit it, either by restricting information, communication, assistance, or competency. However, this largely prevents the act of mentorship, which benefits from the ability to scaffold, guide, and assist learners as needed. In this article, I argue that the dynamic of roleplaying is an underexplored design pattern uniquely suited for addressing both quarterbacking and the need for learning in cooperative and educational games.
Known Solutions to the Quarterbacking Problem
In this section, I analyze the existing design patterns which address or alleviate the quarterbacking problem. This taxonomy was formed through a combination of literature review and close reading, or “close play” analysis of popular and critically notable cooperative games. The play itself done in proxy through online videos of the games being played in order to observe a wider range of experiences, augmented by the synthesis of documented player experiences.
“In many ways, [the quarterbacking problem] is the central problem of cooperative game design.” (Allen & Appelcline, 2018)
Before beginning, I should note that I am not the first to enumerate solutions to the quarterbacking problem. In addition to online blog posts and forums describing the issue (e.g. Alexander, 2013) from both designers and players, Allen and Appelcline (2018) proposed a list of solutions as shown in Table 1.
Table 1. Overview of Current Taxonomy Compared to Solutions to Quarterbacking from Allen & Appelcline (2018).
|Current Taxonomy||Mapping to Allen & Appelcline|
|Information Restriction||Limit communication|
|Information Restriction||Limit communication|
|Hidden Information||Hide the big picture|
| Restricted Information
|Force an individual to make decisions|
|Complexity||Create real options|
|Uniquity||Keep the player busy|
|Social Norms||Convince players that the controlling player doesn’t know best|
|Ownership||Vary the players; Create singular tasks|
There are four supercategories of mechanical solutions: Information Restriction, Complexity, Uniquity, and Imperfect Cooperation. A fifth supercategory, Player Psychology, describes psychosocial dynamics which affect quarterbacking but are not (yet) designed first-order by specific mechanics. Note that games which consider and address this design problem often use a mixed approach, since these patterns are not mutually exclusive and work well in synergy. Further, bear in mind that this taxonomy does not describe the entire possibility space of quarterbacking solutions, only those which we have seen in existing games so far.
In Celia Pearce’s The Interactive Book, game information is separated into four categories:
- Information known to all players
- Information known to only one player
- Information known to the game only
- Randomly generated information
The first category can be considered perfect information. Games that use the information restriction pattern rely on categories 2-4, also called imperfect information. However, as Salen and Zimmerman observe in Rules of Play, information known to the game only and randomly generated information are often very similar. I argue that these two categories are equivalent with respect to the quarterbacking problem, because what they afford is hiding information from all players, and so I call this Incomplete Information. Category 2, information known to only one player, could be generalized to information known to some players but not all, and I call this Hidden Information. Finally, perhaps the most interesting case is a variant of hidden information which is treated as typologically distinct for the purpose of quarterbacking: Restricted Information Sharing. In this category, information is hidden but limited in how it can be shared. Notably, Allen and Appelcline (2018) consider limiting communication as a distinctly separate approach from limiting information; however, because both serve the same purpose (restricting the control of information), I consider these to belong to the same supercategory.
In the Incomplete Information pattern, the game withholds some information from all players. This is primarily achieved through chance of various forms. Appelcline (2003) defines two categories of chance in games: randomness (such as dice) and arbitrariness (such as cards). In Knizia’s Lord of the Rings (2000), much of the strategy of the game depends on the arbitrary cards that are dealt and the tiles that are drawn. In Betrayal at House on the Hill (Avalon Hill, 2004), there is randomness to the encounter, board, events, items, and who the defector is.
The other type of Incomplete Information besides randomness is unknown information. In Sherlock Holmes: Consulting Detective (Sleuth Publications, 1981), players work together to solve a mystery which none of them have encountered before, creating a shared puzzle for everybody.
In theory, Incomplete Information offers a paradigm of a game that cannot be quarterbacked because no one is able to know the information. In practice, this can break in one of two ways.
For Sherlock Holmes: Consulting Detective, so long as no one has played the case before, everyone stands on relatively equal footing, at least with respect to game knowledge. Yet, this means that there are only so many cases that can be played before the game is no longer playable. This solution works for a while until it renders the game finished (at least for those players).
Games with randomization, on the other hand, can still be quarterbacked with probabilistic strategies. The quarterback can demonstrate their knowledge of the random distribution, and in this way the paradigm begins to resemble perfect information again. The only exception is complete randomness for which there is no strategy, leaving you with a game of pure luck. This is not necessarily bad, but it circumvents the quarterbacking problem, which applies only to games with skill-based decision-making, by re-defining the game outside the scope of the problem.
It is perhaps worth differentiating types of information which can be incomplete (or hidden). For example, there are statistical probabilities (e.g. dice, cards), cognitive schemas (e.g. connecting clues in a mystery), and social knowledge (e.g. understanding how your teammates feel, what they would do in a situation). In practice, the handling of each type of information greatly affects the likelihood of quarterbacking.
Consider, for example, Mysterium (Nevskiy & Sidorenko, 2015). In this game, one player plays as a “reverse overlord,”, a benevolent asymmetric role in which they are cooperating with the rest of the players through restricted means. The goal of the game is for this player to convey information about a murder to the other players using only cards depicting surreal art. Unlike incomplete mathematical information, which any quarterback could logically reason through, the information is inherently psychosocial: “what cards would [the reverse overlord] choose to convey this message? Why would they pick these cards to mean this?” In this way, completing the puzzle demands a social collaboration, since each player claims Uniquity to their relationship with the reverse overlord’s psyche. A quarterback does not have equally incomplete information to everyone else because there is some value to what each player has to offer, i.e., their social, perceptual, and cognitive opinions.
Incomplete Information comes in two forms: a game which has no quarterbacking but a finite length, or a game which is harder to quarterback because the strategy depends in part on chance. The former is useful for a niche of games, the latter is a more generally applicable solution, albeit only a partial one.
Contrary to Incomplete Information, where no player holds the information, Hidden Information is the pattern where one or more players hold information withheld from the rest of the group. There are many fewer examples of this pattern in isolation, since it’s usually used in conjunction with other patterns, such as Emergent Complexity, Restricted Information Sharing, Asymmetric Objectives, or Secret Competition.
Allen and Appelcline (2018) describe Hidden Information in more detail, breaking the pattern down into the three ways players can struggle with hidden information: misremembering, misunderstanding, and miscommunicating. They suggest that both hidden information and confusing (i.e., naturally forgettable) information should be meaningful, complex, generic (non-unique, harder to remember), and vaguely trackable.
As an example, consider Dead of Winter (Plaid Hat Games, 2014), a game where each player has a hand of cards hidden from other players. Yet, because each player has their own objectives to fulfill, this information remains hidden for fear of betrayal. In Sentinels of the Multiverse (Greater Than Games, 2011), players also have a hand of cards. However, what keeps this information hidden is not competition but complexity. Each deck is strongly unique, and although it’s possible to show your hand and discuss the options (which can lead to quarterbacking), doing so regularly would slow the game down near to a halt, making the information seemingly hidden through the dynamics of Execution Complexity and Ownership.
Hidden Information as a pattern does not often exist on its own without a reason to hide the information. Items you gain in Forbidden Island (Leacock, 2010) become public informationbecause that knowledge is useful to share; nothing prevents the hidden information from being revealed to the benefit of the group. The three common reasons to hide information are fear (that other players are against you, i.e. when cooperation is not perfect), complexity (discussing this information would take too long), or the game itself prevents sharing this information (see Restricted Information Sharing).
Restricted Information Sharing
The third way to restrict information is to do so explicitly. This pattern always appears with Hidden Information but adds explicit rules preventing certain information from being shared. Allen and Appelcline (2018) enumerate the six mechanics seen so far which create Restricted Information Sharing: revelation limitation (i.e. preventing all information from being accessible at once), detail limitation, unreliable communication, resource cost, time cost, and closed communication (i.e. no communication for a limited time). They recommend that co-op games employ Restricted Information Sharing and suggest that communication restrictions be meaningful, clear and unambiguous, unassailable (i.e. players can’t bypass this rule through undesigned means), don’t interfere with strategic discussion, and ideally emerge organically (such as through Imperfect Cooperation or Social Norms).
The most popular example is perhaps Hanabi (Bauza, 2010), in which players hold a hand of cards away from themselves, being able to see only other players’ hands, and can spend resources to give hints about who has what cards. In …and then, we held hands (Chircop & Massa, 2015), players can speak about anything except the game, preventing strategic communication. Similarly, The Grizzled (Riffaud & Rodriguez, 2015) prevents players from discussing the contents of their hands or how they plan on playing. In Magic Maze (Lapp, 2017), players cannot speak at all, relying only on communicating through the mechanics of the game to work together.
Of all the patterns to reduce quarterbacking, Restricted Information Sharing is often the most drastic approach. Depending on implementation, this pattern can limit or entirely prevent table talk, which can be important for casual enjoyment, since board games are such a social activity. Yet, from the success of these examples we know that it’s possible to use this pattern effectively and still have a great game. Its effectiveness, though, can come at the cost of building the game around that rule: the game is now about the fact that you can’t share the information and you want to, rather than focusing on another aspect of the game.
The other pitfall most often seen in using this pattern is ambiguity and assailability, especially when limiting only the details of information. Players want to win, so they look for the unclear edges of the rules. “Perhaps I can’t say this, but can I say this as a hint?” Ultimately, this ambiguity results in each group having their own interpretation of the rules, loosely approximating the designers’ intent. Few games are truly successful in balancing between information restriction being both hard enough as to be unassailable and soft enough as to not disturb the social enjoyment of table talk and strategizing.
Whereas Information Restriction attempts to limit the quarterback’s ability to dominate, Complexity attempts to limit the practicality of doing so. There are two kinds of complexity: Execution Complexity and Emergent Complexity.
Complexity also addresses a related cooperative game design problem, which Allen and Appelcline (2018) call “the obvious solution,” or the dynamic of having a single correct solution to many gameplay challenges, i.e. an optimal move. Their solutions to this problem include both Execution Complexity and Emergent Complexity, as well as Hidden Information, Incomplete Information, and Ownership (in this case, personal choices or challenges).
Some cooperative games have so much tactical depth that even if the quarterback tries to command the decision-making process, there is enough freedom in the execution of the strategy that the other players can make meaningful choices. That is, to try to control every player’s actions would be infeasible due to the number of moving parts and small decisions. Notable examples include Arkham Horror (Launius & Wilson, 2005), the Forbidden series, Sentinels of the Multiverse, and Spirit Island (Reuss, 2017b).
Because each player’s turn consists of several micro-actions, each of which is difficult to argue as the optimal move, the quarterback struggles to make decisions for others without slowing down the game significantly. In this way, Execution Complexity creates a hindrance rather than outright disabling quarterbacking. The Execution Complexity pattern offers the benefit of reducing quarterbacking, yet it does not guarantee its impossibility.
The other way to create complexity (and thus disrupt quarterbacking) is to have complexity emerge from dynamic systems. Legacy mechanics such as those in Pandemic Legacy (Daviau & Leacock, 2015) and Risk Legacy (Daviau & Dupuis, 2011) ensure that each game is unique, creating complications for the potential quarterback. In a way, this is similar to the Incomplete Information pattern, because the legacy mechanics create a dynamic system which is not fully understood by any player.
The other way complexity can arise is through emergent strategies, such as in Spirit Island. In this game, your character gains upgrades and abilities over time (further complicated by chance). Each game session will create a unique situation, thus replicating the tactical depth of an Execution Complexity pattern.
Because Emergent Complexity is a hybrid of Incomplete Information, like Sherlock Holmes: Consulting Detective, and Execution Complexity, like Arkham Horror, its pros and cons reflect the combination of these two patterns: with legacy games, the design limits how often it can be played fresh; on the other hand, emergent strategy games can discourage quarterbacking but not eliminate it.
Uniquity-based designs attempt to prevent quarterbacking through physical constraints rather than mental or social ones. This is not about the uniqueness of the player characters, also known as asymmetric roles, this is about the uniqueness of the players themselves. The Uniquity pattern relies on the principle that each player is unique and necessary, that the game requires a number of physical bodies to operate and the quarterback is simply unable to be more than one of those bodies.
The Real-Time Play paradigm is simple: the game has real time constraints; there is little to no time to discuss strategies or quarterback, because every moment spent talking is a moment taken away from the execution of that strategy. Examples include FUSE (Klenko, 2015), Space Alert (Chvátil, 2008), Space Cadets (Engelstein, Engelstein, & Engelstein, 2012), XCOM: The Board Game (Lang, 2015), Escape: The Curse of the Temple (Østby, 2012), 5-Minute Dungeon (Reid, 2017), and Magic Maze.
Often, these games give each player unique abilities or unique objectives, further convincing the player that they are unique and necessary to the game (see Ownership). Some games, like Space Alert, flow between timed and untimed play. This positions the quarterback more like a coach who can suggest plans before the play but ultimately has no power to control the team’s execution.
Like Restricted Information Sharing, Real-Time Play is a strong approach which has the power to completely eliminate quarterbacking at the cost of significantly impacting table talk. During the timed parts of gameplay, table talk (if it’s even allowed, see Magic Maze) is focused entirely on scraping together as much strategy and awareness as possible during the hectic round. This creates a particular dynamic. One cannot casually leave the game table to use the restroom or get a drink. Yet, because the timed sessions are short, this is an acceptable proposition from the game.
If Real-Time Play is the strong version of Uniquity, Upkeep is the weak version. Often described by the board game community as “fiddly bits” (e.g. Cooper, 2010), Upkeep is the need to operate the moving parts of the game as it’s played. Trackers must be tracked, counters counted, tokens stacked, resources replenished, and so on. The game’s physical components must be continually moved around as the game progresses through phases and rounds and someone has to do that, typically someone who’s not currently taking a turn. Upkeep is seen most in games like Arkham Horror, Sentinels of the Multiverse, and even some real-time games like XCOM: The Board Game, in which each player’s asymmetric role includes managing their own fiddly bits fast enough to keep up with the timer. These are complex games with many components in which each player takes a complicated turn (or, in real-time games, turns are simultaneous) while the others maintain the rest of the game state.
Overall, Upkeep is not a very effective way to address quarterbacking, because there is no guarantee from the Upkeep pattern itself that quarterbacking will be at all reduced. Games which use Upkeep and Real-Time Play together are already real-time, so quarterbacking is already prevented in that way. Games which use Upkeep and aren’t timed can still be quarterbacked: nothing prevents the quarterback from continuing to disrupt the game. Quarterbacking can happen before, after, or while managing fiddly bits, or the maintenance can be handled by other players. In cooperative games which successfully mitigate quarterbacking and use the Upkeep pattern, it is typically in combination with other design paradigms.
The fourth set of patterns to prevent quarterbacking in cooperative games is a lack of cooperation. By adding the possibility that not all players are working toward the same goal, the quarterback is silenced for fear of betrayal or sharing information with a competitor. Allen and Appelcline (2018) describe perfect cooperation as a “meta-problem” which can be solved with other solutions, namely introducing confusion or chaos, tempting the greedy (i.e. with anti-cooperative incentives), or embracing corruption (i.e. with traitors). There are two kinds of imperfect cooperation: Asymmetric Objectives, in which players share the same high-level goal but different sub-objectives, and Secret Competition, in which most players share the same high-level goal but a hidden minority among them may not.
Asymmetric Objectives are also called meta-cooperation, competitive cooperation, or semi-cooperation (though the latter two can also refer to Secret Competition), and constitute the weaker of the two paradigms for Imperfect Cooperation. In games with Asymmetric Objectives, players have personal goals which, though they don’t typically affect the team’s higher-level win condition, may manipulate the team’s strategic execution, often at the detriment to the team as a whole.
In Dead of Winter, for example, each player has a personal, secret victory condition, such as ending the game with certain cards in your hand or with certain tokens on the board. The group has the public goal of fending off a horde of zombies, but each individual player seeks to also complete their secret victory condition. In Castle Panic (De Witt, 2009), although the team wins or loses together, only the player with the most victory points is declared the Master Slayer.
This pattern offers the benefit of naturally limiting quarterbacking because every player has their own strategy in mind, at least with respect to how their personal goals relate to the main objective of the team. In the ideal outcome, players work together while subtly inching the team in the direction of their ulterior motives. In practice, however, a quarterback can still declare what the optimal move might be and call out any player straying from that plan as a renegade. Moreover, extra victory conditions such as the “champion player” mechanic are often mostly ignored during gameplay, making this pattern weak as an anti-cooperation incentive.
The stronger paradigm of Imperfect Cooperation, and arguably the strongest paradigm of all known solutions, is Secret Competition, also known as traitor mechanics, defectors, or betrayals. In these games, the game explicitly acknowledges that at least one player is working against the rest of the team. Often, this means that a minority (often 10-30%) of players are the traitors while the majority of players are the alliance. Examples include The Resistance (Eskridge, 2009), Avalon (Eskridge, 2012), Shadows Over Camelot (Cathala & Laget, 2005), and Battlestar Galactica (Konieczka, 2008). Two notably unique examples are Betrayal at House on the Hill, where the defector is both randomized and revealed midway through the game, and Dead of Winter, in which the game may or may not include a defector, and whether it does is unknown to the players.
Secret Competition solves a cooperative game design problem by making the game uncooperative. This creates an intriguing design pattern that evolves quarterbacking into a social challenge of deceitful strategizing. Although this paradigm strongly affects the traditional cooperative system, it is extremely effective as demonstrated by the popularity of this genre of semi-cooperation.
Finally, I include a category in this taxonomy for capturing psychosocial dynamics. Yet, these patterns are special in that they have been only alluded to as known solutions of the quarterbacking problem. They are not (yet) imposed by game rules directly, but rather designed second-order through mechanics and cultures which suggest a particular mode of interaction. Rulebooks may discourage quarterbacking on an informal level, but they are not enforced by the rules in any way. For this reason, Player Psychology should currently be considered separate from the rest of the taxonomy of paradigms for designing against quarterbacking, because it is not currently designed first-order. The categories of this approach are: Social Norms, Ownership, and Roleplaying, the last being proposed in this article as an underexplored territory of design.
This dimension of player psychology captures the cultural influences of social dynamics as they affect quarterbacking. Namely, these are the house rules, personal tenets, social contracts, and community expectations which encourage players to voluntarily restrain themselves or enable them to consent or dissent to quarterbacking. Also referred to as “table manners,” social norms are perhaps the strongest but least tangible factor influencing quarterbacking, since game mechanics have yet been unable to directly control these norms. Allen and Appelcline (2018) discuss how this design pattern may theoretically unfold, which is by “providing real and immediate opportunities for an opinion to be wrong,” thus enabling players to learn that the quarterback doesn’t know best and that they can safely ignore the quarterback’s opinions or dissent to being controlled.
Although designing mechanics to explicitly affect social norms is an underdeveloped design philosophy, recent games are becoming more aware of the influence of social norms. Most notably, perhaps, is the recent Spirit Island, which not only succeeds in preventing quarterbacking but was actively designed to do so.
This facet of player psychology captures the notion that players are resistant to quarterbacking when they feel a sense of ownership in the game. Most often, this is achieved through the use of unique player character roles, via unique powers or resources, personal avatars, or personal tasks which enable the players to feel self-sufficient. Spirit Island is again a good example here, using both personal avatars and unique powers which expand and progress over the course of gameplay (which adds both Execution Complexity and Emergent Complexity), augmenting the player’s customization of their avatar and providing a strong sense of ownership over their character’s actions.
The last relevant aspect of player psychology is Roleplaying, defined here as the act of communicating and decision-making as a character rather than as a player, often to a non-optimal gameplay outcome. This paradigm is related to Ownership in that the player must feel a sense of control over their character to communicate or make decisions as that character, but unlike most implementations of Ownership, the character need not be unique to imagine oneself as that role. Despite this paradigm being the core of tabletop roleplaying games, it is rarely seen in true co-op board games — the player character is, almost always, less a character and more an extension of the player. Yet, there are several game design and learning theories (discussed below) which suggest that roleplaying can have a positive effect both as a learning tool and as a social framing, which could help limit quarterbacking to moments of consensuality.
Theories on Roleplaying
The focus on roleplaying stems from research on its cognitive and social benefits. The first benefits to roleplaying are empathy and cognitive perspective taking, key mechanisms of social interaction. Researchers have found that exchanging social positions (such as through repeated and varied roleplay) enhances perspective taking in cooperative problem solving tasks.
Taking on a role can also aid in understanding the challenges of that role. Schaffer (2006) describes the act of incorporating an epistemic frame as a mechanism of joining a community of practice and developing an island of expertise. In this way, players may learn to understand real-world problems by roleplaying through the lens of an expert who encounters and addresses these problems. This extended commitment to a new identity is required for deep learning according to Gee (2005), who describes two kinds of avatars: designed avatars and blank slate avatars. Many roleplaying games implement the latter, but by letting the player project into and inhabit a designed avatar, they are more closely practicing realistic engagement, learning as part of that role. Practicing this role-taking encourages a “projective stance” to the real world. Taking a projective stance, Gee (2007) claims, leads to great mastery and control, and has three steps: recognizing the present affordances, understanding the role of a character who benefits from these affordances, and projecting into that role to interact with the affordances.
As a toy example, Jonathan Blow’s The Witness (2016) is about seeing line puzzles in the world the way Jonathan Blow himself sees these patterns. By the end of the game, the player is enabled to explore the real world and see the affordances of lines in the world, taking on the role of Jonathan Blow to “solve” these puzzles.
As a more practical example, consider Brenda Romero’s Train (2009), a game about complicity. In this game, you play as Nazi transporting Jews to concentration camps. When this realization occurs, players experience a critical reflection on their actions, enabling them in the future to reflect on how a person can become complicit in an experience and, perhaps, how they can recognize complicity in their own lives. In this way, the roleplaying of Train enables the player to take a projective stance on recognizing complicity and taking on the roles of both complicit and critically questioning persons.
In summary, roleplaying can increase empathy, perspective taking, mastery, control, and learning, making it a potentially effective tool in educational games and cooperative tasks. Roleplaying can therefore be a preventative measure to quarterbacking in this way: by encouraging the players to take on the identities of avatars who reject non-consensual quarterbacking, the game’s design can develop — first-order — a set of social dynamics nuanced enough to prevent non-consensual quarterbacking while allowing consensual quarterbacking.
Roleplaying as a Solution for Cooperative Educational Games
Through Roleplaying, players can choose to withhold information, not because the game explicitly prevents it, but because it would violate their roles established by the game’s narrative and intended dynamics. Just as easily, players can choose to provide information.
In this way, Roleplaying addresses educational needs for learning and mentoring through a kind of voluntary restraint. The players can construct a relationship in which the mentor seeks to guide their mentee, not through command but through personalized pedagogy and dialogue.
To demonstrate how a game might use Roleplaying in this way, I have developed two examples, one educational game and one entertainment game, which exemplify the Roleplaying paradigm. I discuss their design in the section below.
Game Examples of Roleplaying
In this section, I introduce two designs I have created to explore roleplaying as a paradigm. For a non-educational example, consider the proposed design of My Realm Will Be Yours, a game for two players who play as a king and his daughter. In the beginning of the game, the king holds a set of randomly generated information about the kingdom and wants to convey this information to his daughter, but fears giving too much information lest she become too powerful and rebel. The princess, on the other hand, presses him for information so that she can competently rule in his absence during the final phase of the game. In addition to the Roleplaying paradigm, this game uses Hidden Information (only the king knows how to rule the kingdom) and Asymmetric Objectives (the princess would be happy to know everything, but the king wants her to know only some things).
As an educational game example, consider Cirque du Socrates, a proposed roleplaying game for studying flashcards. In this game, some players are psychic philosophers putting on a show by calling members of the audience (other players) to participate in a mind transfer. The goal of the game is to get the participants to learn the contents of the flashcards but without the philosophers outright telling them the information, a rule which is policed by the other philosophers who don’t want to be upstaged. In addition to the Roleplaying pattern, which is used to uphold the rules of Restricted Information Sharing, this game uses the “champion winner” mechanic from Asymmetric Objectives. Although the philosophers are working together to put on a good show, each wants to be the best performer (and likewise the participants want to be the fan-favorite audience member).
Although these two games don’t span the entire design space of what Roleplaying can do to restrict quarterbacking, they demonstrate how Roleplaying can be used in both entertainment and educational contexts.
Does Roleplaying “Fix” Quarterbacking?
Although I argue in this article that Roleplaying is a helpful tool for combatting quarterbacking, and especially helpful for the context of educational games, it is just that — a tool — one among several in a designer’s toolkit. Roleplaying is no magic bullet, nor is it useful for every game. However, it fills a niche previously underexplored within the space of consensual asymmetric knowledge dynamics: it encourages healthy dialogic mentorships and gives room for learning and teaching at the gaming table. In this way, roleplaying adds more true cooperation to co-op games.
But for readers looking for concrete examples or takeaways, here are a couple of roleplaying variants you can try with games that might benefit from them.
Forbidden Desert, The Helpful Navigator
For this, and the other games in the Forbidden series, try this variant rule if you tend to have a quarterback at your table: Have the quarterback play as the Navigator, whose base ability is moving another player three tiles (or two for Forbidden Island) as one of the Navigator’s actions.
In this variant, instead of the Navigator directly moving others, they take on the role of the team’s tactical leader, speaking as a mentoring officer. Using one action, the Navigator offers 2-3 choices to another player by describing either what target to head toward and/or the high-level reasoning behind the strategy of each option. The Navigator is encouraged to be thorough in any reasoning they give, describing the pros and cons of each choice. However, they should also be neutral in this judgment, not outwardly favoring any choice, since the Navigator wants to empower his allies to understand the tactical decisions of their own accord. The navigated player then picks one of the options. If the details were left incomplete by the Navigator, the player adds their own description of the choice; i.e., suggesting the reasoning if the Navigator didn’t, or suggesting exactly how to move if the Navigator gave only high-level strategy choices. Then the navigated player moves a number of tiles toward their chosen goal: 1 tile (or 0 for Forbidden Island) +1 if the Navigator gave 3 choices instead of 2, +1 if the player’s choice matched what the Navigator would have done (and the Navigator’s choice was not obvious from their description), +1 if the navigated player provides details that matched what the Navigator would have said about the choice (and these details were not obvious from the Navigator’s description). For an easier game, +1 more tile.
In this way, true cooperation and mentoring can bolster the team’s efforts, while ineffective command will hinder the Navigator’s own ability. The choice-response paradigm provides a handshake of consent in the navigated player allowing the Navigator to guide them while maintaining their own autonomy. Notice, though, that the focus on this variant is on the alpha gamer, while the next example focuses instead on the novice player.
Spirit Island, The Young Spirit
But wait, I hear you say, Spirit Island does so much to reduce quarterbacking already! And yet, sometimes it still isn’t enough. On the first game of Spirit Island I played, the most novice gamer at the table still felt controlled and overwhelmed by not knowing how to deal with the game’s complexity. If this happens to you, try this variant rule:
In this variant, each player takes on the role of their spirit. They are encouraged to use a silly voice to embody their character: perhaps a deep voice for the Vital Strength of the Earth, a raspy voice for Shadows Flicker Like Flame, or a warbly voice for River Surges in Sunlight. In this roleplay, the novice player’s spirit is the youngest of the spirits, and the elder spirits are teaching them about the ways of the island. To this effect, turns happen less simultaneously than traditionally played as each spirit talks through their actions. The elder spirits describe what they are doing and why, elaborating in narrative detail what effects this has on the colonizing invaders and the indigenous Dahan. Whenever an elder spirit plays a power card, however, instead of the card going to their discard pile, it goes into the novice player’s hand: the elder spirit shows the younger spirit how to perform this power and gives it to the younger spirit. In return, the younger spirit discards a card from their hand, which may be the card just given to them, to the elder spirit’s discard pile. The young spirit is not ready to handle so many new powers, and thus decides to temporarily abandon one from their practice.
As stated earlier, this variant focuses on empowering the novice player rather than hindering the quarterback, although both approaches reduce quarterbacking. Notice, too, how consent is used, albeit differently. In the Forbidden Desert example, the handshake of consent takes place at each action the Navigator uses to guide other players. In this example, on the other hand, consent happens prior to the start of the game. By opting into this variant, the novice player is consenting to the role of being taught, and the other players are consenting to the role of the teacher. This reframing from forced quarterbacking to consensual mentorship is the core of the Roleplaying paradigm.
Roleplaying as Part of a Growing Taxonomy
Consent is a critical part of healthy play, and roleplaying as a design pattern is but one step toward designing for consent. To reiterate, this taxonomy does not represent all possible design spaces, but rather the ones which have been thoroughly explored over the last two decades.
Applying This Taxonomy: A Case Study of Pandemic and Spirit Island
In just the last decade and a half, designers have learned a great deal about designing cooperative games, from the classic Pandemic (Leacock, 2008) to the sudden hit of Spirit Island (Reuss, 2017b). To demonstrate how one might apply the taxonomy provided in this article, let’s examine these two popular games through the lens of this framework.
Pandemic makes use of Incomplete Information (randomization), both for what abilities you have available as well as where disaster strikes. Yet, Pandemic is well-known for its quarterbacking problem. Pandemic, and games of the same format, are easily quarterbacked because they are playable by a single player. As author and designer Justin Alexander (2013) puts it, “this is really just Solitaire, but you alternate turns.” The ability for these games to be played solo tends toward what Koster (2018) calls “playing alone together.”
The introduction of Pandemic: Legacy, which adds Emergent Complexity to the game, makes an attempt to address this problem with mixed results, according to some players. Spirit Island, on the other hand, is widely praised for addressing the quarterbacking problem. How does it achieve this? In addition to similar randomization strategies of Pandemic and the Emergent Complexity of Pandemic: Legacy (as player characters level up over the course of the game), Spirit Island introduces simultaneous play. While this isn’t technically Real-Time Play because you aren’t being timed, this form of personal Upkeep means that everyone can attend to their own turn’s actions and leave other players to think about their own strategies. There’s also an element of Execution Complexity, since there are multiple ways to deal with any given threat.
The Future of Quarterbacking
Quarterbacking has been a well-known problem in cooperative games for more than a decade now. It is not only an issue with cooperative board games, but many other cooperative tasks as well — escape rooms, tabletop RPGs, and many other group learning or group problem solving activities suffer from the dynamics of non-consensual control. Larps for example struggle with player control from both the game masters (GMs) and other players. In “secrets and powers” and “blockbuster” larps, GMs pre-write the game’s story in a way that must strike a balance between player freedom and game structure, often leading to non-consensual player control via “plot monopolization” and “plot trains.” Despite existing scholarship on player motivations (e.g., McDiarmid, 2011), little has been explored on how the game design itself — such as logic puzzles and hierarchical factions — emergently create quarterbacking problems. In recognizing this pattern in cooperative board games, will we see a rise in designing for new social dynamics? Will we begin to see designs in which the players collaborate with the system to maintain healthy social dynamics? Will collaborative tasks, even games, begin to carry codes of conduct to implicitly or explicitly help the players establish consensual relationships? I only hope that, through this article, I have widened the awareness of issues with consent in asymmetric power relationships and provided enough vocabulary to continue this discussion.
Quarterbacking is a design problem for every cooperative game to consider. Educational cooperative games bring an additional design constraint: all players need to learn the material for the game experience to be a success. Moreover, quarterbacking is a problem with consent, since asymmetric decision-making is only a problem when it goes against the players’ desired experience. In this paper, I provided a taxonomy of the existing design solutions to the quarterbacking problem separated into four major mechanical categories: Information Restriction, Complexity, Uniquity, and Imperfect Cooperation, with the fifth category of Player Psychology capturing psychosocial dynamics. I proposed roleplaying as an underexplored solution to the quarterbacking problem which is also uniquely designed to additionally address the need for learning in educational cooperative games. To demonstrate the Roleplaying paradigm, I propose two games (and link to their rules and print-and-play materials), one educational and one entertainment game, which use the Roleplaying design pattern. Finally, I briefly compare Pandemic and Spirit Island to give examples of how this taxonomy can be used to evaluate how well games address the quarterbacking problem.
The author wishes to thank Nathan Partlan, whose playtest comment spurred this research.
Featured Image by ElectricFootball @Wikipedia CC BY-SA 4.0
Josh Aaron Miller is a PhD candidate at Northeastern University where he studies game-user interaction and transformational games. His work combines game design principles with the psychology of learning and motivation to create seamless and engaging user experiences in non-game contexts. He designed Descent Into Madness, a Lovecraftian horror one-shot RPG system published in the Indiepocalypse anthology, and he is currently the lead game designer of Foldit, the protein-folding game.