Glitches, Players, And Metagames

This piece was originally published as portions of a proceedings paper for ISAGA 2014, an academic conference on simulations and games. I (James Murff) provided the perspective of entertainment game design, while Elizabeth Tipton provided the perspective of educational game design. It has been edited and improved for clarity and a casual audience.

In 2009, my decade-long conversation with my mother - a professor of statistics - on the comparisons between educational and entertainment games was made public during panel discussions at ISAGA and ABSEL.

At that time, the focus was on improving educational game design through an understanding of problems commonly seen during the play of entertainment games. More recently, the conversation has turned to player agency, emergent gameplay, player archetypes, and metagaming.


There are many tales of games not turning out as expected.

Sometimes games have so many glitches that they are unplayable; for example, Big Rigs: Over The Road Racing, where the physics are abnormal, the AI is non-existent, and there are no win conditions. Others are vaguely playable, but suffer from glitches which inconsistently destroy the experience; for example, Ride To Hell: Retribution, which morphed from an open-world revenge fantasy into an exceedingly offensive linear game so bad that the only positive remark in Electronic Gaming Monthly was about the pause menu.

However, glitches do not necessarily lead to unplayability. The Elder Scrolls V: Skyrim has hundreds of bugs where non-player characters behave unexpectedly; they repeat the same routine over and over again, or their animations glitch out comically. Players turned these glitches into a form of egg hunt, posting videos of these odd behaviors on YouTube. Bethesda, the game’s developer, even kept certain glitches in the game just for their entertainment value.

Red Dead Redemption's glitches pertaining to NPCs associating with the wrong animation and sound set - the gunslinger-dog, the bird-people, the cougar-man, and the donkey-lady - were immortalized online due to their comedy. Minecraft took advantage of one glitch by repurposing it, creating the character known as the creeper from a failed model for a pig. The creeper was subsequently merchandised into stuffed toys, action figures, LEGO and apparel. It has even appeared in TV shows, music concerts, and games produced by other companies.

Glitches are not restricted to the entertainment gaming industry; they also happen in educational games. Two papers (Wolfe & Jackson, 1989; Wolfe, 1993) discuss situations where the advertising demand function within business games violated the economic concept of advertising elasticity of demand, and players were apparently unaware that the games’ weren't accurate to real-world principles. In terms of learning outcomes, this is of serious concern, as business games are used for training/evaluating players’ business skills (Greco et al, 2013).

However, sometimes glitches can occur in educational games without derailing outcome. During one play of the supply chain simulator The MIT Beer Game (Sterman, 1992), small coins were used to represent delivered product. A player stumbled and dropped about 400 of them on the floor. As coins flew everywhere, the players demanded that the game stop, as those downstream in the supply chain did not receive their product. A local businessman participating in the game exclaimed “Have you never heard of a beer truck wreck?” The game then continued as normal.

Glitches can also be caused by deliberate errant player behavior. In an international business simulation contest, a team performed well despite the clear displeasure of the members. When asked, one member reported that the decisions were not what the team had submitted to the contest site, but were rather those of a single member who had been overruled by the rest of the team. When confronted, the suspicious team member admitted to changing the decisions after their turn-in. The team proceeded to “fire” that member and went on to win the contest.

Emergent behavior

Emergent behavior is when specific components of a game behave as they are supposed to, but either go too far or interact with other mechanics in ways the developer didn’t originally intend.

The Radiant A.I. system developed by Bethesda for The Elder Scrolls: Oblivion - and then expanded for use in Skyrim - allowed non-player characters to interact with the game world by establishing goals and needs which then had to be met by the AI. While this gave the games a more organic feel, it also caused non-player characters to satisfy their programmed needs in strange ways. In perhaps the funniest example, players committing crimes were caught despite using stealth and avoiding the gaze of human NPCs. Bethesda finally tracked it down to chickens reporting the crimes to guards.

Emergent behavior is not restricted non-player interactions. In Crackdown, the Agency SUV gains traction and suspension as the driving stat is improved. When the maximum level is reached, the car can grip the side of a vertical surface, allowing the player to drive up the sides of buildings.

Emergent behavior in game engines also aided in the construction of now-ubiquitous game mechanics. A glitch in the programming of Street Fighter 2 allowed for players to skip the recovery time of a move by performing another within a specific time window. Though originally unintentional, this mechanism allowed players to string together moves to deal more damage to opponents, creating the now common feature of "combos". This former glitch has even seen deliberate iteration through mechanics such as Guilty Gear's Roman Cancels and Street Fighter Alpha's V-ism mode.

Straferunning - the act of strafing diagonally to increase player speed - was an engine glitch in DOOM that was later embraced by the community. Bunny-hopping - jumping the moment you touch the ground after a previous jump in order to build momentum - was an engine glitch in Quake that was later codified in Quake 3 Arena's level design.

id Software is also responsible for the creation of rocket jumping. In both Doom and Quake, rockets push back characters when they explode, which allowed players to jump significantly higher and farther by shooting rockets at their feet. It was embraced by the competitive community, and was widely copied in other shooters from Half-Life to Team Fortress 2. Team Fortress 2 in particular incorporated rocket jumping to allow heavier characters a greater sense of mobility at a nominal health cost.

Emergent behavior is not restricted to the entertainment gaming industry either. During the play of a marketing simulation many years ago, students provided their decisions by punched cards. The program, at that point in time, did not validate inputs. One team input the actual number of units to be manufactured, forgetting that the number was to be coded as a thousand unit block. When the computer run was completed, they ended up with an overwhelming amount of inventory. Due to the feature that automatic loans were provided if a firm underestimated its cash needs, the team also had an insurmountable debt.

During play of the business simulation Capstone, one poorly performing team suddenly changed their price from $24 to $24,000 near the end of the game. This resulted in zero sales, as expected. However, industry demand was determined by a combination of all prices in the market, so total sales also went to zero, completely collapsing the marketplace. When asked why, the team stated that they assumed the lagged effect of price on demand would let them sell a few to make a lot of money before the market collapsed.

Hofstede & Murff (2011) reported on an experience where an old, well-tested game, So Long Sucker, demonstrated emergent behavior when used in a multicultural classroom. This game was designed by a group of American mathematicians to demonstrate unstable coalitions. When played by American students, cooperation existed only in the short term, as expected. When played by Taiwanese students, however, a new behavior emerged as cooperation existed well past the point where it was obvious that not everyone could win. When both groups were mixed, a new behavior emerged among the American students; they adopted some, but not all, of the behaviors of the Taiwanese students.

Rewarding Emergent Behavior

Many developers attempt to simulate the process of emergent behavior with varying success. However, the only true way to harness emergent design is to watch how players approach a particular game. For example, Dishonored and Deus Ex determine the player's path via previous choices. Although both games allow and even encourage violence, it is possible to complete them without doing anything worse than knocking somebody out.

Intentional emergent behavior is not restricted to these genres. Completion of the puzzle, not adhering to a set machine design, determines success in physics puzzlers such as The Incredible Machine. The very first modern roleplaying game, Dungeons & Dragons, encouraged players to develop their own house rules, styles of gameplay, and narratives. This is not a new idea; H.G. Well’s encouraged intentional emergent behavior in Little Wars in the chapter entitled “Ending with a sort of challenge.”

This intentional emergent behavior can also be seen in educational game settings. In The Bunny Festival, developed at ISAGA Summerschool 2012, the specifications for a product were laid out in an instruction sheet. These steps appeared to be rigid, but upon careful reading, it was apparent that there were no real restrictions, and players could fill the specifications any way they wanted to. When the designers realized that their project encouraged emergent gameplay, the game was altered to run for two production rounds - separated by a mid-game debriefing - to encourage players to “think outside the box.” Players continue to find unanticipated ways to satisfy each of the specifications in every session since the rules change.

In 2005, Peter Molyneux, the designer of Populous and Theme Park, commented that players of the next-generation of games would “want to customize the experience, setting their own goals in a world that they can play around in." Players of games today grew up in an environment that encourages emergent behavior. This is particularly evident in a recent study of undergraduate students of information technology (Dahlstrom, 2012); 55% of the students “wished instructors used more simulations and educational games.”

Emergent behavior, no matter the source, gives the player the feeling of having discovered something. Even if a game is designed specifically to encourage players to try a particular behavior, allowing the choice reinforces the belief that the game allows for true interaction and thus enhances player agency.

Player Agency

Where designer intent is how the designer wants the game systems to impact the player's choices, player agency is how the player makes those choices. Player agency is what separates a game from a non-interactive medium like a book or movie, as it has the potential to affect the course of the game, even in miniscule ways. Call of Duty has very little overall player agency; the player’s only interactive role is to kill enemies to advance in the game. Conversely, Dwarf Fortress is often lauded as a prime example of player agency, as every small decision has a natural and evolving impact on the game world.

Player agency doesn't necessarily have to affect the game, either. In games like Gone Home or Spec Ops - generally linear, rigid narrative experiences - the illusion of player agency impacts the player's emotional state. The White Phosphorous scene from Spec Ops is notorious as one of the most emotionally impacting and memorable scenes in shooters, while the continued revelation about the protagonist's sister throughout Gone Home gave LGBT players representation where there is usually none.

This matters because games are not merely lectures from a designer to a player. They are more akin to a dialogue; the designer dictates, and the player refutes or agrees. This is obviously abstracted, of course. For example, the “dictation” of enemy placement is “refuted” by how the player eliminates the enemies. It’s a conversation where one party — the designer — can’t hear the other party’s side of the conversation until the conversation itself is finished. In this respect, it's very similar to the notion of postmodernism as laid out by Death of the Author.

This is especially important to remember because games are not static. The way players approach games and interact with them changes over time. This is not just a social or cultural trend, but an exploration of the old-made-new. Players are still finding quirks in old games like Legend of Zelda and discussing them and posting about them in places like the Speed Demos Archive. Any designer's goal should be to build a game that offers the player enough agency to encourage dissection and exploration years after release.

Problems have been long noted with the completely predictable environment of the typical classroom; this setting is particularly dissonant for students who have been raised with games that encourage player agency and emergent behavior. This means that educational games must be designed with a greater focus on interlocking and observable systems rather than direct teaching. Learning should at least be partially auto-didactic; players should interact with the game and at least recognize new concepts, even if they don’t understand them. Students may even break the game in an unexpected manner, as in the So Long Sucker experience.

In a predictable environment, the teacher would force the students back on track to the planned learning outcome. Through skilled debriefing, a teacher using a game designed for greater player agency can expand the lesson and address the concepts discovered by the students. The teacher can even use emergent behaviors to teach a different lesson than the one initially planned. Thus, players learn in a more natural way through the use of their internal agency, rather than being told what they have to know. However, the potential for deviation from the planned learning outcome due to player agency appears to be problematic in the current educational environment of common standards and rigorous assessment.

Player agency can often be tied to the principle of “players will do anything they can figure out how to win.” A level in Portal requires solving several smaller puzzles in order to complete a large puzzle that allows movement to the next level. Valve found that experienced players skipped the main puzzle entirely by utilizing the game physics in an interesting way. Rather than taking a few minutes, as expected, these players solved the bigger puzzle in seconds. Yet not all players figured this out, as players differ greatly in personal goals and attentiveness when palying a game.

Player Archetypes

Players behave according to a principle that is best paraphrased from physics: “Players seek their lowest natural state.” In other words, players seek the easiest, most straightforward avenue for their goals as possible. These goals, however, are not necessarily those defined by the game designer. Rather, these goals vary from person to person, and can be roughly codified into three archetypes: Power, Challenge, or Exploration.

Power Gamers

Power gamers focus on victory above all else. They will bend — or even break — the rules in order to gain an advantage. This often manifests itself as using weapons that are stronger than others (“unbalanced”), utilizing strategies that have a high chance of winning against weaker players (“cheesing”), and memorizing bugs that grant some advantage (“glitching”). These players will run roughshod over other players, and even the game itself, in their bid for victory.

Power gamers are usually not motivated to play games which have no designer-defined goals. They need some sort of structure in their games to motivate their play, or they move on to a new game. Make a power gamer play a sandbox game with no goals whatsoever — such as Minecraft or Dwarf Fortress — and they will likely quit out of boredom.

This is not necessarily bad, though. Power gamers are the easiest players to predict, as they always follow the win condition, often to extremes. To put yourself in the mind of a power gamer, you work backward from victory to figure out how the win condition can be circumvented, exploited, or otherwise cheated.

Consider a power gamer playing “capture the flag” in a game that allows for the use of vehicles (such as Battlefield). The victory condition is to capture a specified object, usually a flag, at the other team’s home base and return it to the player's home base. Working backward from the victory condition, it’s clear that the power gamer will obtain the flag in the quickest possible manner. They will then most likely seek out a vehicle soon after grabbing the flag; this will, after all, confer a distinct survival advantage and thus make victory that much easier. If there is no vehicle available, they will use weaving, unexpected paths, and difficult-to-reach areas to throw off pursuers and make it back to home base.

The best designers, rather than punishing these players for outside-the-box thinking, make their unusual behaviors a part of the game. For example, when the shortcut in Portal was discovered during play-testing, the developers left it in and formulated a challenge around it. The developers of Dishonored noticed that players were using their given supernatural powers in unusual ways and combinations. Rather than whacking the players' hands with a proverbial ruler and telling them no, these alternative strategies were embraced by the developers, leading to a wide variety of paths that converge on a single goal. Player agency was enhanced through the designers learning of alternatives from the power players.

In an educational setting, these players are the ones that are most likely to rip your game apart within the first session or two. They will always seek out the victory — whatever that may be — and will attempt to achieve it as quickly as possible. If they can’t, they will continually try new tacks until one works. These are the students that ask the wellworn question: “Will this be on the next exam?”

Challenge Gamers

Challenge gamers are those who play games not because they want to win, but because they want to be challenged. This means that they often devise unusual tests in order to heighten the difficulty. Refusing to use certain gear, approaching the game from a far more difficult angle, and playing against players who are obviously and significantly better are all a part of the challenge gamers’ repertoire of tricks. Challenge gamers seek to win, which is similar to a power gamer, but only if it conforms to their own internal ruleset.

Much like the power gamer, challenge gamers thrive in environments that have defined goals of some kind. However, they also do well in a sandbox environment, as this allows them more freedom to express their need to challenge themselves. This makes player agency a powerful, and important, component of that expression. For example, Dwarf Fortress is very appealing to challenge gamers, as it offers a mix of sandbox (build your fortress and watch its inhabitants) and explicit goals (grow in wealth, fight titans, and bring in more dwarves.

Where the power gamer is exceedingly easy to predict, the challenge gamer is far tougher. Where you can simply look at the optimal path to a goal for the power gamer, you must also consider modifying that path for the challenge gamers. For example, a challenge gamer might refuse to move past a game obstacle until it is defeated, where a power gamer will often simply circumvent the obstacle entirely to conserve resources and save time.

In an educational setting, challenge gamers are those players who often force themselves into more difficult situations. This is rarely conducive for their learning, but is very helpful to the educator, as the knowledge gleaned from challenge gaming can be used to further refine the game in question. Think of these as pruners; you can use them to locate and remove extraneous lessons from your game until the one left is the one you intend to teach. These are the students that will latch onto a particular topic, such as dinosaurs, and then proceed to learn everything that they possibly can about that single topic before moving on to something new.

Exploration Gamers

Finally, there are exploration gamers. These players indulge in games not because they want to win or to challenge themselves, but because they want to explore the systems that make the game function. They are the most difficult player archetype to predict, as they often indulge in behavior that is directly contradictory to the explicit goal of completing the game. For example, exploration players may just spend time looking at and admiring the textures in a room rather than progressing to the next area of the game.

Rather than deriving pleasure from completing the designed challenges, exploration gamers derive pleasure from whatever personal agenda suits them. Where one exploration player might enjoy listening to every sound in the game and judging how they interact with each other, another might focus on seeing how NPCs react to unusual stimuli. They may set off on a quest of their own devising to locate unreported glitches, or they may invent completely new ways to play old games. You can’t purposefully design a linear game around an exploration gamer, as they will rapidly lose interest in a game that forces them into a particular win condition outside their own goal.

Exploration gamers thrive when placed into environments with plenty of experimentation; while sandbox games are obvious choices, any game with unusual systems will generally hold their attention. For example, an exploration gamer may play Worms 2 in order to better understand the way the wind physics work, or they may play Euro Truck Simulator 2 to explore every single road in the game.

In an educational setting, exploration gamers are incredibly valuable. Not only do they expand the parameters of the game by approaching it in their own unique way, they open dialogues with the designer that the designer can’t predict. This means that a designer can learn more about their own game through an exploration gamer than they originally thought possible. In the classroom, these are the students are difficult to motivate in a traditional lecture and exam structure, and getting them to turn in their homework can be exasperating. Yet, these are also the students that will create new ideas and find unexpected connections between old ones.

All of these archetypes, however, have one thing in common: they want to be invested in the game. Whether it be through victory, challenge, or exploration, each type of player has their own ideals and goals. The easiest way players accomplish these goals is through metagaming.


Metagaming is the act of creating a “game outside the game” by altering the game parameters through non-game actions. If a game is made up of structures — mechanics dictated to the player by the designer — then metagaming is made up of metastructures — mechanics dictated by the player to the game.

Metastructures usually act as a recontextualization of existing mechanics or motivations. For example, the most common form of metagaming is metacompetition. When players can’t directly compete with one another, they compete in other, more creative ways.


Tournaments - those gatherings where players compete to see who the is the best - are an integral part of gaming, and are the most commonly-encountered metastructure. These tournaments can be as explicit as an actual convention - such as EVO, the world's largest and most prestigious fighting game tournament - or as small as jockeying for the top spot on leaderboards - lists of players organized by their ranking in particular statistics.

Emotional metagaming — fakeouts and mindgames — is common within the fighting game community, but is ineffective outside local multiplayer games. This doesn't prevent players from trying in games like League of Legends or Call of Duty, though, which gives those games their notoriously bad reputation for toxic communities.

A few games, EVE Online being the most notable, incorporate political maneuvering as a metastructure. More wars in EVE are won and lost in the smoke-filled back rooms than on the battlefield. This form of competition may be unusual, but it is still a metastructure, as those politics are not directly managed by the games mechanics; they are merely encouraged through the way the game's systems intersect one another.

Competition doesn't necessarily have to be a tournament, though. Speedrunning - completing a game as quickly as possible, including using tools to make the impossible easier - is a natural offshoot of the common arcade goal of completing levels as quickly as possible, and has seen a resurgence in recent times thanks to the popularity of events like Awesome Games Done Quick.


Not all metastructures are competitive, though. Some bring players together in the spirit of mutual cooperation.

Theorycrafting - examining a game's mechanics and concepts to find optimal solutions - is a form of metacooperation that almost every internet-savvy player engages in. This is common across all genres, and acts as a socialized version of powergaming. This usually manifests as collaborative data collection, where players band together collate information and theorycraft. RPGs such as World of Warcraft and Dark Souls, and survival games such as DayZ, are excellent examples of players working together to dissect and understand a game.

Then there's using violent games in constructive, cooperative ways. The most popular indie game of all time, Minecraft, originated as players using the lesser-known Infiniminer to build structures rather than complete objectives. Within Minecraft, giant cola cans float in the sky and immense yet intricately-detailed castles dot the landscape, all built by one or more players. Players have even gone so far as to create Turing-complete calculators powered by dwarves in Dwarf Fortress.

Successor games - those cooperative, singleplayer-turned-multiplayer structures where players hand off singleplayer saves to other players through the internet - have resulted in tales such as the infamous Boatmurdered, where belligerent dwarf-eating elephants died in a biblical-level flood of lava.

The comp-stomp - playing a competitive game against the game’s AI, rather than against other players — is common amongst those who play strategy games such as Civilization 4, but rare in other genres. This is because training bots to play a reflex-based game realistically is far more difficult than training them to play strategy games, as strategy games have more discrete, solvable situations and a cheating AI in a strategy game is not as difficult to overcome as a cheating AI in a reflex shooter.

Rare Metastructures

There are also some metastructures that don't fit either category and seem obvious in hindsight, yet are so rare that it’s surprising they were ever discovered. The clearest and most recent example of this is the Twitch Plays Pokemon phenomenon. This particular metastructure used a videogame video streaming service and internet chat to create a place where thousands upon thousands of players could attempt to play the same instance of a game at the same time; imagine a hundred thousand players all mashing buttons on the same hand-held Gameboy. This social experiment of using internet viewers as a collective mind to input game commands to a turn-based, low-penalty-for-failure game is something wholly original and entertaining, if somewhat frustrating at times. During the first run of the game, it took over 16 days of continuous play to complete a game that averages around 50 hours.

Educational Metastructures

Metastructures can also occur in the classroom. When So Long Sucker was used in a Swiss-system tournament, four students played to lose, but only some of the time. During the debriefing, these students explained that were curious to see if what they had learned about negotiations and strategy in another course could be implemented to attain a goal they had self-determined. They were attempting to manipulate the tournament to ensure that only they would be in the finale together.

In a nationwide simulation contest, a team did some very sophisticated data analysis of the other teams’ results, and made very good estimates of the other teams’ decision making processes. One of the judges noticed and asked the team how they were able to accomplish this task. A student explained a few analytical tools that he was using, and then turned to the judge and said, “Are you able to comprehend what I am explaining to you?” The student was not playing to outperform the other teams, but rather to outperform those administering the game.

Final Thoughts

Gamers do not all approach games with the same goals and motivations, just as students do not all learn in the same manner. Players may create new metastructures and avenues of play that the designer is unable to predict. This then is yet another reason why debriefing is absolutely necessary for establishing what player got out of a game. Goals laid out by a designer may not be the goals adhered to by a player. After all, “most learning is not the result of instruction. It is rather the result of unhampered participation in a meaningful setting.”  (Illich,1972)

If you wish to support my writing on games, please consider donating on Patreon.