 The universe, in many senses, is a dance between order and chaos. It has certainty embodied by the seemingly fixed laws of the cosmos, but uncertainty expressed by the strangeness of quantum mechanics. If the games we create are a microcosm of the universe, it makes sense that they might reflect this duality. Knuckle bones, better known as dice, have been used for centuries to simulate the whims of the fates, and games like snakes and ladders are argued to have emerged in ancient India because they reinforced the arbitrary nature of reality reflected in their philosophy. An ethic was encoded in games to reflect the culture they arose from, but games are also played because they give us control over this randomness. In ancient China, Goh was used alongside three other art forms as an admissions test into high society. In his seminal work, The Ambiguity of Play, Brian Sutton Smith outlines different metrics of play. He cites progress, fate, identity, power, imaginary, and frivolity as archetypes that have existed through the ages. Play is never one thing because play exists in a specific cultural context that molds it to its particular whims. Play is uncertain. However, this seems to suggest uncertainty only exists in certain contexts, but this is not true. Games that require strategic intentionality fall on a spectrum of complexity. We might start with Tic Tac Toe, then graduate to chess in our adolescence, and finally, the estimate is that there are more games of Goh than there are stars in the universe. However, Goh is more fun in the technical sense, precisely because its possibility space creates more uncertainty for the player. It has a higher branching factor, more computational complexity, more strategic versatility, more room for uncertainty. The takeaway point here is first that we have games that invoke different amounts of uncertainty because of the cultural context they arise from, as well as the fact that uncertainty and the depth that arises from games are not mutually exclusive. To extend this point, the ability for AI to conquer humans at games of various kinds has always been a marker for how far AI has come. Checkers was conquered a while ago, and then chess, and now even Goh can't resist the tides of artificial intelligence. When AlphaGo beat Goh champion Lee Sedol at Goh, it used a method called Monte Carlo Tree Search, which actually plays games randomly to prune a decision tree with more certainty by calculating probability. Tetris, perhaps video games equivalent of Goh, shows us another form of uncertainty to enhance complexity. To generate its tiles, it uses a random number generator to author its play space. However, it's not actually fully random, as then absurd configurations of the same piece might crop up. But it is sufficiently random to thwart pattern recognition. The uncertainty in Tetris makes the game one that tests a different faculty to Goh, showing how uncertainty can be invoked in the design of games to craft emergent forms of play. This is Greg Kostakian, and he wrote the book Uncertainty in Games. In it, he says, if you're not uncertain of what's going to happen in a game, why even bother playing? He runs through a list of more than a dozen forms of uncertainty we use in games, and how this affects our capacity for design, but also makes the deeper philosophical point about how uncertainty is built into the structure of games. However, games are not just enclosed systems. They are a microcosm of the properties of the universe, and so the parity between reality and the magic circle become important to take note of. Uncertainty is everywhere in our quest to discover the secrets of the universe. Correlational studies don't imply causation, in addition to being beholden to significance tests. We proceed by induction with most scientific questions, but one result to the contrary can delegitimize an entire body of work. In Heisenberg's Uncertainty principle suggests that at a fundamental level, knowledge of one property of subatomic particles precludes knowledge of another. In his seminal work, Risk, Uncertainty, and Profit, Frank Knight identified what's known as Knightian Uncertainty, which he distinguishes from risk. Risk is an instance where you can reliably refer to a probability distribution, whereas uncertainty characterizes a situation where one doesn't even exist. For example, if we know for a fact the probability of the various hands in a poker game, and have some inkling of the different variety of cards on the field of play, we can make calculated risks about what to do, making decisions based on expected value. Conversely, if I were to throw you into a game like Go right now, with you having no experience of it before, chances are you will just randomly play moves. There will be no intentionality. John Maynard Keynes once stated, about the matter of uncertainty, there is not scientific basis on which to form any calculable probability whatsoever. In games, when we have no basis for accurate strategic foresight based on information, human beings tend to defer to heuristics and strategic conventions. Heuristics drive a lot of strategy in games, including games that are conventionally about skill. Studies show that chess masters are really referring to a database of stored patterns, as opposed to brute force look ahead. In his now famous book, The Black Swan, Nassim Talib coins the term, the ludic fallacy, which he cites as the misimpression you can use models to predict reality. Models don't have all the information, and black swans, extremely unlikely events, can come from outside the model and are wholly unpredictable from within it. We can take his concept and invert it though, calling it the reality fallacy. This is the misuse of our understanding of uncertainty outside games and applying it within the magic circle of play. With games, we can limit the effect of outside influences, control the transfer of information, and reliably construct the states of mind players have down various avenues of a decision tree, so it requires a different conceptualization of uncertainty. The Keynesian and Nitean type of uncertainty is what we will call fundamental uncertainty, but it is only one of many types of uncertainty we will explore. Let's use a commonly invoked framework in game design to ease us into this. In his GDC talk, Luck and Skill in Games, and elaborated on further in his book, Characteristics of Games, George Scafalius points out how luck and skill are not mutually exclusive. A game like Tic Tac Toe has low skill and low luck, Bingo has high luck and low skill, Poker has high skill and luck, and Go has low luck and high skill. What does he mean by this though? And how does it relate to uncertainty? His conception of it is that skill is a pure calculation, like multiplying 43 by 34, and luck is experienced as random variables you cannot control, whether it be by using a dice, shuffling, or a random number generator. Luck is when the dynamics of player are such that one can't act with intentionality or strategic foresight, and hence there is less of a payoff for more skill. There are many design purposes for luck, which can also be called randomness. First, it allows for weaker players to compete, evening the playing field. Next, it forces adaptive play, making it so that improvisational play is more important. Finally, it can generate novel and exciting moments within the field of play. What's interesting is that Go has a low degree of luck because it has a high amount of uncertainty, and there can actually be skill in luck, like how poker at a high level is extremely skillful. In the art of game design, Jesse Schell explains how the calculation of expected value is a skillful form of probability estimation, and can be incorporated into skill-based design. Even when luck does intervene over enough games played, more skillful players of poker will win out. Luck's skill, risk, and uncertainty are actually independent variables that can be invoked and combined in different measure to create a particular type of experience. Skill is the ability to act with foresight, planning, or execution. Luck is the extent to which you can't control factors in a game. Risk is the extent to which you can invoke probability to act with some degree of control, and uncertainty is the many ways these elements express themselves during the dynamics of play. Elias introduces some important things in our conceptualization of this, though. He claims that the number of tiers a game has can tell us about its skill. He claims that luck not just makes a game more accessible, but also allows people to blame their failures on the fates than interesting rhetoric of play. Skill can refer to strategy, memorization, deception, manipulation, cognitive dexterous, or any number of other attributes. Luck or randomness can be input or output randomness, and can come in a variety of colors. Input randomness is like using the dice to determine your moves and monopoly, and it requires you to adapt to this in the play space. Output randomness exists in games where you can make a move, and then a randomness output device determines what happens, and it exists in a game like Risk. Input randomness supports more skillful foresight, whereas output randomness is more uncertain. The color of randomness, as explained in this GDC talk by Jeff Engelstein, is the amount of fuzziness you are dealing with, and we seem to have a predilection for a certain kind of randomness. White noise is when there is zero correlation between the last and next result. Brown noise is where the correlation is very high, and pink noise is where there is a big chance of a small change, and decreasing chance of larger changes. Boardgame conventions have come up with different ways of introducing unpredictable events, including dice deltas extending the extreme or using multiple dice. In Advanced Squad Leader, crazy things can happen if you roll a 2 or 12, limiting the appearance of randomness, but having it still be a factor. Exploding dice in the Uncharted SEAS games has it that if you roll a 6, you can roll again and it grants you a multiplier, giving a slight chance of major damage. When we turn to video games, we use sporadic critical hits, variably scheduled loot drops, and even feedback loops to create this. The Call of Duty feedback loop system makes it so that even amateur players can get on a lucky streak and completely dominate the battlefield using killstreaks. However, it is rare enough to still be uncertain. The levers of randomness can be used to make games more accessible, introduce adaptive and creative play, create the right type of noise, be used to modulate games using feedback loops, and can also be used to tell a narrative. Uncertainty is not a unified concept, though. It can express itself in many ways and come in many forms. Greg Kostakhan's book has a fairly robust conceptualization of the different forms of uncertainty, so we shall use it as a reference. The first kind of uncertainty he defines is performative uncertainty, and this is the uncertainty that you, as the player, can actually perform the tasks needed to finish the game. Extremely difficult games like Dark Souls and Super Meat Boy leverage this to entice players to meet their challenges, which is mostly predicated on players improving their skills to match the challenge. The second type of uncertainty, Solver Uncertainty, is a similar kind of uncertainty, except it adds cognitively challenging tasks. The puzzles of the Witness, for example, very rarely require you have rapid reflexes, but it stresses the need to remember and apply a web of conceptual constraints. These two forms of uncertainty fall under the broader label, difficulty, and are often seen as the core of what makes a game a game. Designers leverage an array of tools and players in this state of uncertainty, whether it be flow theory, implicit interaction design, or in some instances, dynamic difficulty adjustment. These games also invoke various subtleties of design to craft different experiences. A puzzle game like Steven's Sausage Role crafts player uncertainty by making it seem impossible to figure out what to do, but as explained by various puzzle designers, this is needed to elicit the sensation of Eureka when the solution dawns on players. The third kind of uncertainty is called player uncertainty, which is the phenomena where you have no idea what your opponent is going to do next. This exists in any game that has another player or a sufficiently capable AI system and requires we deal with it in specific ways. Most games in this realm leverage a few non-nash equilibrium games to create gameplay. The reason for this is that they prevent any Nash equilibrium or situations where a dominant solution for all parties arises. Rock Paper Scissors Dynamics can be seen in everything from the throw, block, and hit cycle in fighting games to the infantry, tanks, and aerial units of battlefield. Matching Penny's Dynamics are in any game that involves trying to match your strategy to an opponent's choice and plays out pretty much like a penalty shootout. This means that the uncertainty at the level of the player creates strategic uncertainty at the level of games broadening the possibility space. The reality though is that information deficits, mixed unit configurations, and the fuzziness of dynamic play unlocks another strategic dynamic called Yomi. Yomi is the act of trying to read something new, and this involves that we bluff, lie, scout, research, or attempt any action that reveals their intentions. Games of this ilk are testing our ability to deceive, manipulate, and read others, skills of vital importance in an uncertain world. It also makes these games have virtually infinite depth, as it effectively creates a new game every time you play a different player. Preventing a dominant solution in a game means that all the elements need to be balanced so that there is always interesting decisions to be made, and this takes us to uncertainty. Sid Meier famously called games a series of interesting decisions, but what he really should have said is that they are interesting precisely because there is uncertainty. He outlines risk versus reward decisions and long-term versus short-term decisions, and this maps on to a form of uncertainty Greg Kostakian calls analytic uncertainty. Analytic uncertainty is basically the uncertainty that a company is not being able to fully anticipate the state space of a game, but this is precisely what makes it engaging. In this case, you can make small risk reward decisions, like choosing between stacking your blocks higher to get a Tetris, at the risk of ending the game prematurely, but you can't predict the chain far enough to solve this game beforehand. One domain this version of uncertainty applies comes when you try to define the depth of a game. Elias tries to elaborate on some methods of gauging depth, including the number of tiers of players that exist, the extent to which better players win out over worse players, the cognitive strain of a game, or just a raw mathematical calculation whether depth should be measured independent of player perception is another area of dispute. In this paper, depth in strategic games, the authors propose a measurable property of a game's formal system, D, that corresponds to the capacity of a game to absorb dedicated problem-solving attention and allow for sustained long-term learning. They go through state spaces, computational complexity, branching factors, skill chains, and strategic ladders to uncover D, which they classify as the capacity for a game's system to allow for a ranked population of strategies. It's still an area of ongoing inquiry, but it seems clear that uncertainty has a role to play in understanding of depth. The next type of uncertainty is called procedural uncertainty, and is the use of random number generators, seed generators, and procedural systems to craft gameplay. A game like Spelunky uses procedural level design to keep players engaged, but it made sure to combine this with more curated design to manage it well. All levels are on a 16 grid that starts at the top and ends at the bottom. A game like Elite used a procedural seed generator to generate its vast cosmos, and we see this now in games as varied as Minecraft to No Man's Sky. Procedural systems are also going to become increasingly important as we push towards more dynamic story generators as well. This brings us to another form of uncertainty, called information uncertainty, which is about deliberately obscuring information from the player to craft specific forms of gameplay. Early strategy games had a system called Pog of War, which prevented you from seeing what your opponent was doing early on. In order to explore the map, you had to send scouts out, making gathering information a part of the dynamics of play. A recent game, Into the Breach, put a twist on this formula, allowing you to see ahead, but as the developers explained, this fundamentally changed the game and forced them to make it about defensive play instead. This shows how altering information constraints can change the game you are making. People often compare poker to chess, suggesting the former is a game of imperfect information, and chess won with perfect information. But is this necessarily true? In poker, you are unaware of certain cards on the table and are not privy to whether they are bluffing. As you play, you are engaging in a Bayesian process of updating prior probabilities, as new information is presented to you, and truly skillful players need to adapt as information reveals itself. In chess, all the pieces are laid bare on the board. However, you are not aware of what is happening in their mind, the skill level they possess, and the strategy that is being employed. The information uncertainty is not about the physical board, but the uncertainty you have about the game's decision tree. This brings us back to Yomi, which is fundamentally about being able to read the other player. In a game of Virtua Fighter, throw beats block, which beats hit, which beats throw, but the real depth of the game comes from it's Yomi. You have to anticipate what your opponent's biases are, what they think your preferences are, and how this tangled web can go add-in for an item. The final form of uncertainty we will discuss here is a mix of narrative, semiotic, and perceptual uncertainty. In his GDC talk, The Play of Stillness, Brian Opton talks about the importance of forms of play that just take place in a player's head. One form of this is called anticipatory play, which is what might happen between two turns of a chess game. In games with intentional strategic foresight, the time between turns is sometimes just as important as the moments of interaction themselves. However, he further elaborates on the importance of interpretive play, which is the act of interpreting the meaning of the events that transpire in a game. For example, a game like PT is just about walking through a corridor, but the real play comes from piecing together what is going on. Uncertainty in a game's explicit story can actually be a powerful form of engagement, although most other art forms also have this, they don't require as much interaction on the part of the player to unearth their mysteries. This form of uncertainty is extremely important when crafting environmental storytelling, distributed forms of meaning, as well as systemic story generation. What drives the obsessive community of souls born games if not the ambiguity uncertainty and mystery of its worlds? Sometimes you need to make sure things are clear to create different forms of uncertainty. Mark of the Ninja has very clear affordances for vision, sound and enemy awareness, but this is blended with information obfuscation to create dynamic challenges for the player. And the Legend of Zelda Breath of the Wild uses an art style that draws analogies from the real world to enable its systemic play in boldening player choice. In the underappreciated shooter Space Giraffe, the game inundates you with perceptual uncertainty, making it very hard to see, but this actually adds more depth in that it forces you to pay attention to alternate cues, turning obfuscation into a game of strategic depth. Like we examined though, uncertainty is not only important for crafting engaging gameplay, but also in the communication of stories and ideas. In Rimworld, the designer made sure that there were very simple and clear cues for what was happening, arranged in a hierarchy to ensure players could actually read the stories embedded in the game, to generate interpretive and emergent uncertainty. In The Witness, solver uncertainty is married with interpretive uncertainty, to craft a game whose theme was about uncertainty, the uncertainty we all have when it comes to the pursuit of truth. At a deeply conceptual level, games are compelling precisely because they are uncertain, but this should come as no surprise because the universe we inhabit itself exhibits these properties. Games are disordered systems that are amenable to feedback loops and chaotic emergence that yield much strategic and narrative depth for players. However, because games are interactive ergodic texts, we bring order from the chaos. In his book, Trickster Makes the World, Lewis Hyde argues that Tricksters were the first mythological artists, as they inhabit the realm of uncertainty with frivolity and mischief. Play and art were two sides of a coin that existed on the periphery of what is known in the realms of the uncertain. In games, we act out this metaphor of artful creation and playful destruction by being forced into states of uncertainty. When Einstein claimed God does not play dice with the universe, he was speaking about his reservations with quantum mechanics, but perhaps we can reframe his anxieties. The universe may be chaotic, but we can forge order within it by bringing these dynamics into the games we create and playing with them.