Game Design

Way back when I started this blog, one of the first essays I did was about conceiving of a game as the combination of three related spaces – physical, mechanical, and narrative – and gameplay as the act of allowing the player to explore these spaces. I think this perspective is still useful, though sorely in need of revision now, five years later (I’ll likely return to it at some point in the future). However, whenever I think about how different games emphasize one or the other of these attributes, whenever I try to draw a hard line between where one space begins or ends, I run into a bit of difficulty making that division. The mechanical and narrative spaces are fairly easy to delineate – one is the actions you can take in an environment and how that environment reacts to those actions, the other is the story that is told about those actions and the context in which they take place. However, the physical space, which one would expect to be the most intuitive of the three, is a bit more difficult to delineate – and I think it has to do with how we create physical space in games.

The problem I keep running into is that the physical space of the games is actually created by means of mechanical and narrative elements. The mechanical aspect of the space is your ability to move around on some parts of it and to have your movement blocked by others, and the narrative aspect is the colors and textures and what they suggest about the world you’re in. Together they create something that feels like a chunk of physical world to explore, but there’s nothing actually physical there. A sense of physical space is created, but it is not separable from the mechanical and narrative elements that contribute to it – not in the same way that the mechanical and narrative elements are separable from each other.

It may be extremely obvious that the physical reality of the game world doesn’t exist, but it’s suggestive that we create the illusion of a physical reality through recreating the parts of reality which interest us most as humans. That is, when we encounter an object, our concerns are a) what can I do with this? And b) what does it mean that this is here? This, of course, has very little to do with the actual material world, where objects are made of many different bits and pieces, covered with bits and pieces of everything else, subjected to forces we have an incomplete understanding of. What’s noteworthy is not that we are simulating a reality, but that we are simulating outwards in, out from the superficial aspects we find ourselves most interested in, down into the more fundamental aspects such as mass and warmth only as we find those necessary to power the superficial simulation.

Tangentially, I am now quite certain that if we had any way to simulate texture and taste in games we would have done so long ago, as these are also superficial aspects of great interest to human perception.

It’s fascinating that we so many of us consider what we’re doing to be realistic. What we do with games is render exclusively that which can be seen: every 3d object is an empty shell, every character who is modeled is simply their exterior with nothing inside, and interior parts only created as they become necessary to render when they are ripped apart from the exterior (a common scenario in games). We see what a human, or a house, or a rock, looks like, and reproduce what we see, when that is inherently only the most superficial possible version of that thing.

Something from physical reality is translated into signals for our brain, is stored as a symbol representing that object; then our brain conducts our body to create an object that can reproduce those signals in another brain. That is what we call art – or, at least, representational art.

So, with games, we started from the simplest version of the most superficial reality, and from there we’ve managed to make more detailed and convincing forms of that representation. Perhaps we could simulate a reality based more on what we know to be there than what we see to be there? Even a primitive simulation of a more complete reality could lead to new and interesting artistic pursuits. Or, perhaps, since we are unmoored from the physical basis of reality, we could create a simulated world far wilder and stranger than we can while paying lip service to material reality.

Mostly, though, I just find it amusing how much we like to act like we’ve come anywhere near a reality simulation when our approach is in essence purely superficial. How very human of us.


Among other ways to think about games, one that I rarely hear spoken of is their capacity as attention engines. Among the many social and emotional needs we have as human beings is our desire to be heard – even more than to say anything specific, we just yearn to be able to jam a flag into the dirt for people to see. Having someone listen to you, even passively, can be hugely rewarding, emotionally and even intellectually, as you feel connected with the world.

Before most games, we had ELIZA. ELIZA is a simple chatbot made to emulate a psychotherapist, one who answers every question with a question. “What are you thinking of?” “How did that make you feel?” “Why do you say that?” Despite being created with an almost parodic intent, to show the superficiality of human-machine communication, people felt a genuine connection with ELIZA, and sometimes even a degree of therapeutic benefit. Now, in 2017, there are lots of ways to be listened to by machines. Most of us have a machine in our pocket who will answer our questions as best as it is able, and will listen and respond to anything we say no matter how inane – though, perhaps, not in a very satisfactory way. Still, unsatisfactory answers are not necessarily too different from what we’ve come to expect from genuine human social contact either.

A lot of what we want from games is for them to just respond to what we say to them. a game lives or dies on its ability to react to us, to listen to what we are trying to say. Because real artificial intelligence is a very long way away still, our methods of communication are usually greatly restrained in games to enable them to react in a satisfying way. A game’s controls are the language we use to speak to it: Frustration ensues when a game misunderstands what we are trying to communicate, or when it doesn’t allow us to communicate the thing that we desperately want to. We describe these sorts of problems as control issues, which I suppose says as much about us as it does about them.

This is what people really want a lot of the time when they ask for non-linearity. They don’t care about replay value and they don’t care about getting the sex scene for the character they like, they just want the sensation that they are being listened to. They want the video game equivalent of someone nodding along and saying “Mm, wow. Interesting. Huh. I see.” That those multiple branching paths and endings and romantic partners are available is chiefly valuable because it reifies that sensation, makes it feel solid and responsive – that there is, indeed, someone listening. And, perhaps, there was, 18 months earlier, a designer who listened to them by way of the player proxy voice who lives in his head, who he designs for.

A little while back the game Passpartout: The Starving Artist became a small-scale hit. Passpartout is a game where, playing as the titular starving artist, you paint using an extremely simple art program, and passersby choose to either buy your paintings or not, occasionally giving explanations as to what they like or don’t like about the picture. Now, the engine for evaluating these paintings is seemingly pretty simple, apparently rewarding the artist more frequently for time spent than for technical skill, but it serves its purpose, effectively convincing the player that the game is paying attention to what they are creating, that they’re not just creating into a void – which it all too often feels like artists are.

Unfortunately, Passpartout is quite similar to a prototype made by Jon Blow called Painter. He released this prototype for free on his site, and in talking about it he described it as a failed prototype because it failed to realize his vision of a strategy game where you create paintings to appeal to different gallery owners and curators and achieve success. Some statements made on Twitter suggest that he was dismayed that a game so similar to his failure could be a success – but there was no actual reason for the Painter prototype to be a failure except that it failed to achieve the vision he’d had in mind. The element of trying to appeal to tastes was never actually very interesting. What’s interesting was making a painting and having it be seen and acknowledged, of being told that something you’d made had worth. All the judging algorithm had to do was make it so the game could reasonably successfully determine which you’d worked hard on and which you’d hastily crapped out and evaluate them appropriately, just to ensure that you knew it was paying attention.

This may seem fake or trivial, but this loop of the player communicating something and the game responding is the core of what a game is. It doesn’t need to be a real or detailed response, it just has to be real enough to show the player that someone, or something, is listening.

A game is a collection of symbols and rules for how those symbols can interact with one another. I described these as being obstacles and tools, one of which defines the parameters of success and the other of which are used to navigate through those parameters – but the division between the two isn’t necessarily as harsh as I implied. Sometimes obstacles and tools are the same – that is, sometimes you can use one obstacle to navigate another.

There are a number of interesting examples of this – you could jump off of an opponent’s head to reach a higher platform or, inversely, you could jump to a higher platform to cut off an opponent’s pursuit. You could lead a group of enemies from one faction to encounter a group of enemies from another faction so they start fighting each other, or you might reposition a trap to activate another trap safely. You might even intentionally jump on an exploding trap to boost yourself to a normally unreachable rooftop. In addition to obstacles sometimes behaving like tools, on occasion a tool will take on the characteristics of an obstacle. The most common example that we encounter in games is probably the hand grenade, which is both an extremely potent weapon and also a convenient way to quickly separate the component parts of your bloody carcass.

There’s no reason for tools and obstacles to be different at all, in fact, when they all just boil down to being a set of physical properties and behaviors. There’s no reason for an object to know what its purpose is, only for it to behave in accordance with a set of instructions. In Spelunky, for example, you can pick up and throw almost everything in the game – rocks, pots, laser turrets, unconscious yeti – and, once an object has been thrown, it affects the environment in pretty much the same way no matter what it happens to be – well, unless it happens to be explosive, in which case its effects will be more dramatic. It’s actually fairly commonplace to throw a rock to set off an explosive only to have the rock blasted back in your face by the explosion and hit you… into a yeti who throws you onto a collapsing platform which falls on top of a mine which blasts you into a pit.

When items are agnostic of their origins and purpose, surprisingly intricate interactions become possible. In many other games, rocks would only be useful for hurting opponents or for setting off traps – in many other games mines would only be activated by the player, enemy bodies would be inert, explosions would only affect things which could be damaged. It’s fascinating what can be achieved when we allow items to be exactly what they are, instead of design them specifically to fit a particular role in the game.

Another interesting example is Phantom Brave, a tactical RPG from Nippon Ichi Software: In this game, the main character Marona is the only living character you control, with all of your teammates being ghosts she has befriended and which she can summon to help. The catch is, to summon a ghost she has to have something to summon them into, which can be something as ordinary as an everyday bush or rock or as ornate as a cursed sword. Whatever they get summoned into, they gain properties of that object so, for instance, if you summon someone into a tree they’ll probably be more vulnerable to fire damage. The other catch is that literally any object on the battlefield, including other characters, can be picked up and used as a weapon. These systems, interacting with others such as a system where every item comes with a stat-modifying adjective before it, enable some really strange and intriguing strategies. Sometimes it’s necessary to pick up one of your characters and toss them across a pit, pick up an enemy and beat up his teammate with his body, or summon a weak character holding a really nice rock you’ve found and have them drop it there for you to summon a more important fighter into. Later in the game some enemies are even scripted to start picking up and throwing powerful items off of the stage to keep you from summoning allies into them.

In some ways, this functional agnosticism of game objects is the default – when I say a game is a collection of symbols and rules, I’m not just speaking conceptually, but also in terms of how games are made, what the internal programming logic that goes into their operation is. So, one might ask, if this is the natural state of the game, and if this creates so many interesting and emergent situations, why do so few designers allow their games this kind of leeway? Unfortunately, when you increase the possibility space in a game this way, you also increase the odds of something going haywire, of an uncontrolled feedback loop or absurd dominant strategy that completely undermines the intended game design. Part of why this openness was possible in Spelunky and Phantom Brave is that these are very tightly controlled designs – Spelunky through having a small set of game elements with only a couple of methods of interaction and a straightforward and minimal progress path, and Phantom Brave through presenting a restrained and traditional tactical RPG interface. For a look at what this ends up looking like without this sort of restraint, take a look at Dwarf Fortress – which is an important and fascinating game, to be sure, but is not easily accessible to most audiences and results in scenarios which, though they are amazing stories, frequently represent bizarre and illogical breakdowns of the symbolic logic as the system recursively interacts with itself.

Still, it’s worth considering: How am I restraining this object, making it behave in a more constrained way than it has to – and a less interesting way than it could? How would it affect the overall design if these constraints were to just, maybe… disappear?

Fictional lies pose an interesting challenge. With the (many) lies we encounter in our daily lives, we understand that there is a reality that these lies are purposefully misrepresenting – that, even if we don’t know the truth, there is a truth to be known. However, when the entire reality of the thing is made-up, when even the truth of the world we are participating in is a lie, lies that happen within the fiction of that world become strangely insidious. Sherlock Holmes says that once you have eliminated the impossible then whatever remains, no matter how improbable, must be true – Sherlock Holmes also solved a mystery where a man drank a de-evolution serum and turned into an ape-like creature to murder the victim, so I feel his claims should be taken with a grain of salt. Within the world of Sherlock Holmes the impossible is quite possible, because he is a character in a book and the only guarantee we have against impossibility is the discretion of the author.

It’s particularly difficult to convey lies in games. In novels and movies the audience doesn’t need to worry about what to do next, so they don’t actually need to worry about whether something a character says is true or not – it might be fun and interesting to think about, but determining falsehood won’t factor into what happens next at all. A lie in a game, however, can have consequences for how the player experiences that game, possibly leading them on a wild-goose chase or towards a decision which will turn out to be disastrous. Also, since people don’t expect to encounter lies in games, they tend to regard any information which doesn’t pan out and isn’t explicitly revealed as a lie to be a mistake or, at best, a vestige of cut content. For instance, the villagers in Castlevania 2 may have been lying or the game may have been poorly translated – as best as I can tell from a cursory look online it appears to have been both – but when players were misled they read that not as thematic but as accidental, and blamed the developers for slipshod work.

Because of this, certain conventions have arisen in games when it comes to falsehood, and these are rarely betrayed. Lies are almost always constrained to the narrative of the game, while rarely affecting the actual gameplay. That is, being lied to by your commander has become such a common game trope that it’s surprising on the rare occasions that you’re not betrayed, but since being lied to or not makes absolutely zero difference in how you approach the problems presented by the game it doesn’t matter. It’s just fluff.

This is a solution, of sorts, but it also removes most of the narrative power and interest from falsehood. Another solution might lie in informing players that they will be lied to and that it’s on them to believe or disbelieve what they’re being told, but this results in its own set of problems – asking players to determine whether someone is lying is a core mechanic of LA Noire, but in reality every actor whose performance was captured for the game was, in fact, lying. They’re actors, that’s the job. So it’s a matter of determining which lie was the more convincing looking lie – which is really a shit way of determining when someone’s lying, since accomplished liars are much better at being convincing than people who are unaccustomed to lying, even when the latter are actually being completely truthful.

I suspect that the only way to integrate falsehood into gameplay in a way that’s satisfying is to leave traditional failure states behind completely. If we refocus the game’s design around exploring a story, rather than ‘winning’ at it, then the player is free to believe or disbelieve what characters say based on what they think is the most interesting to the story or on where they’d like to see it go next. This also opens the door for playing a character role, where the character can be trusting or cynical, rather than analyze the scenario for optimal play.

Even then, something is lost, because this makes the deception stakes low for the player. Is there a way to satisfyingly integrate falsehood into a single-player game, when the systems of the game and its narrative are being conveyed through the same channels? How can the player know what to trust, without the game being scrupulously honest at all times?

Not too long ago, and for a lot of the history of video games, the visual quality of a game has been decided entirely on how ‘realistic’ the graphics are. Using photo textures, true-to-life lighting models, and increasingly sophisticated shading systems, we tried to – and, indeed, continue to try to – create rendered images that are completely indistinguishable from a photograph. On the one hand this makes a lot of sense – I mean, photorealism is often regarded, rightly or wrongly, as the height of technical mastery for a painter, so shouldn’t game graphics aspire to the same thing? On the other hand, what a tedious aspiration this is, for a medium that could do literally anything, portray any kind of weird and wild reality.

Fortunately this is no longer the aspiration for most games. This may have as much to do with the problems inherent in trying to produce to this quality of fidelity on a budget as with any shift in aesthetic priority, but the end effect is that realism is no longer the universal standard of quality – in most games, that is.

It’s interesting and a bit dismaying to look at the games where ‘realism’ is still prized. War games, mostly, and particular first-person shooters. These games are mechanically some of the most distant from their source material – wars full of permanent death, permanent destruction, permanent loss, portrayed in a manner where everything can be redone, remade, regained, with a quick checkpoint reload in single-player or starting the next round in multi-player. Sure, the same can be said of most games, which usually have dramatic stakes and some sort of loading/reloading system, but rarely does real and tragic loss sit quite so closely to quick and easy consequence-free gameplay. There’s something exceptional and grotesque about using real wars, some quite recent, as set-dressing for your shooty game, and selling that illusion with state-of-the-art graphics.

The reason why realistic graphics have become less popular, aside from budgetary reasons, is that we’ve realized that graphical style can communicate something about the nature of the game and the world it takes place in. The reason why it’s odd that realism is still the art style of choice for military-themed shoot-em-ups is that what this art style conveys is: “this is reality, this is what war is like, it’s gritty and bloody – and also painless and fun and inconsequential!”

Perhaps they’re pressured to adopt this realistic style by market forces – it is, after all, easy to appreciate realism because we know what reality looks like. It also makes them appear faithful and respectful to the realities of war in a certain way, since they study real war to make sure they can replicate its aesthetic, and perhaps the desire to use a realistic style is in some way a response to the massive narrative and mechanical disconnect noted earlier. They keep pushing this aesthetic harder, and though they still shy short of presenting the screams of agony, the begging for mercy, the child casualties, how long before they wear this, too, as aesthetic? How long before the fans defend these choices, as well, as being ‘realistic’ to the war portrayed, when realism is the furthest thing from the mechanics of the game experience?

Maybe this all seems very alarmist, but the reason why this bothers me is how often people who advocate real actual war position themselves as being realists, as just being pragmatic, when they talk about the necessity of armed conflict. The way we frame discussions of war as being willing to do what’s necessary, willing to see a hard thing through, it seems similar to the way we smear dirt and blood over things to make them seem real and true, wearing the aesthetic of sacrifice instead of trying to understand what is lost. And, to be clear – this isn’t just games. We wear blood and suffering as a costume, while quietly shuffling past all the actual blood and suffering, in all sorts of media.

So perhaps it’s just market forces that make it so every game that’s about being a person, about real and painful loss, looks like a cartoon – while every game about getting to be a cartoon, about being Itchy and Scratchy killing each other over and over again, looks like footage from a war zone. Perhaps I’m just worried about where the market is forcing us, and what will happen when we get there.

I’ve been playing a lot of Dishonored over the past couple of weeks, and I’ve been, for the most part, loving my time with it. Of course, the more I enjoy a game, the more frustrated I get with the few things that stand out to me as issues. I went over this a bit a few days ago, about how difficulty changes which are usually used to push a player to explore a possibility space can, when used without care, constrict that possibility space. Modifying the difficulty of a stealth game can be tricky: After all, there are a lot of easy handles to grab to modify the challenge of a traditional action game, numbers such as damage, health, speed, powerup effectiveness… It gets trickier in a game that’s not based on overpowering force, but on sight-lines and motion, suspicion and awareness. In stealth games, what is difficult or not is often based on intricacies of positioning and movement, rather than whether one number is bigger than another.

Because of this, it’s entirely understandable that Dishonored’s attempts to increase in difficulty over the course of the game, and across difficulty settings, largely boil down to the game becoming more punishing of mistakes. That is, rather than asking you to succeed against a greater challenge, it mostly asks you to succeed against the same challenge but with a smaller margin for error. This is not completely ineffective, since it does add some tension – but since the actual consequences, in a game with quickload, are so negligible, it doesn’t really change the experience of playing the game except to make it more finicky. Heightening the consequences of mistakes just removes any chance to recover from them, any chance to retreat, to improvise, and replaces these with just reloading your last save. Does this incentivize more careful planning? Not especially effectively, when the worst that happens is a quick load screen and then another attempt at navigating the same challenge.

I suppose some might say that this is an issue with the player being able to freely save and load, and I think this is not an invalid perspective, but I prefer to look at it differently. I think the issue is more that the designers approached the creation of more difficult enemies as a way to push people away from the using the exact same tactics that were discouraged by every other enemy, but more punishingly and aggressively. Ideally, each new enemy would add some new factor the player had to contend with, a new and different challenge that forced the player to approach it in a new and different way. The tallboy enemy type, essentially a set of armored stilts, does this quite effectively: This enemy type cannot be choked or easily stealth killed, and also can see into areas other enemies cannot because of its height. However, other enemy types, such as the Music Box Overseers, and even the late-game basic enemy types with improved attack abilities, mostly just serve to make it less feasible to recover from a mistake while still being vulnerable to the exact same tactics.

I thought it would be an interesting design exercise to consider how I would try to improve the game – at least what I would consider to be an improvement, I know tastes vary. What follows are my notes for a fantasy patch for Dishonored, one which would push the player to vary their approach to the game’s obstacles while still allowing for different generalized styles of play. Having not played Dishonored 2, it’s entirely possible that I’ll say something that appears foolish in light of the changes made in that game. Oh well.

Fantasy patch notes:

Cannot knock out opponents using the choke-hold on very hard difficulty. On all other difficulties choking works on any opponent from behind, regardless of alert status

More objects are throwable, including all vases and dishes as well as swords and armor dropped by enemies.

Throwing enemy bodies, dead or alive, at an enemy causes a long stagger.

Damaging alarm stations in any way now sets off the alarm.

Enemies are now staggered by attacks that hit them during their attack, but only after their attack is complete.

Enemies sometimes do desperate attacks while staggered, which increase their stagger time but can be dangerous when careless.

Tallboy models revised with heavier armor, drop attack no longer possible, have a blind spot directly under them beneath their sight-line.

Music Box Overseers are visibly armored front and back, though there are enough gaps to make choking and stealth kills possible. They’ve selected elite troops to carry the music boxes, so they’re all visibly taller and the armor has red highlights. Music boxes now, rather than disabling all magic use, rapidly drain mana – once mana is drained, they continue to quickly drain health. This effect is weaker the further away the player is from the music box, and up close the drain is enough to kill the player in about 3 seconds. This drain rate is percentage-based, so the same regardless of current mana/health, and the lost health and mana will regenerate if the music box is removed. Some Music Box Overseers are set to constantly play, and will only stop if knocked out or killed. A new effect has been added to make the range of the music box more clearly visible. Being behind a wall will offer some protection from the box, but it continues to affect the player. Music box no longer slows down player movement.

Armored Butchers no longer have a ranged attack and deal damage that results in near-instant death at melee range. They now always explode on death or knockout, alerting everyone nearby and dealing slight damage. This makes knockouts impossible on non-lethal and ghost playthroughs. However, the player can also pickpocket the oil tanks powering the armor, leaving them immobile, though they can still cry for help. While immobile they can be picked up and moved like any other body.

The intent with these changes to create a game that’s a bit more dynamic. Meticulous planning is still the strongest route to success, especially with the new types of obstacles and complications you have to plan for, but you also have more room to improvise a recovery, both in lethal and non-lethal play. I tried to make the aspects I dislike less obnoxious without actually nerfing them – that is, I feel that these versions of the Music Box Overseer and Armored Butchers are actually much more challenging and dangerous than the extant versions, but also more interesting to play against.

Playing a game is learning a set of habits, a set of reactions, and trying to tune and optimize those habits and reactions towards those that most frequently successfully achieve the game’s goal. Calling it a strategy might, sometimes, seem a bit grandiose, since we make a lot of these decisions through habit and muscle memory, but the decisions we wire into our brains are still strategic constructs. However, the approaches we take must shift with the situation the game puts us into: Depending on the circumstances, a good habit may yield bad results, or a foolish strategy might win big.

Weak players tend to rely too much on the general case best strategy, where more experienced players recognize the situational nature of tactics. A case I recently saw was a game where you have three dice, a number you’re trying to beat, and have a single reroll of as many dice as you choose. Now, the obvious best strategy for trying to optimize your roll for high numbers is to reroll every die that lands on 1, 2, or 3 and to keep every die that lands on 4, 5, or 6 – this will yield the average highest result. However, when the goal is not just to roll high but to specifically reach a threshold, the strategy must change to suit the circumstances. As an extreme example, let’s say the target is 18: You now need to reroll any die that comes to something lower than 6 to have even a slight chance of winning. Or, if the target is something else relatively high, like 16, even if it’s possible to keep a 4 and still win, that would mean you’d need to roll 6’s on the other two dice, which is significantly less likely than rolling 16 on three dice – though both are a bit of a long shot.

Of course, that’s quite an edge case, and I can’t see any scenario short of trying to roll an 18 where you’d want to reroll a five, but there are many examples of when the necessity of trying to reach a particular success threshold affects the strategy. An extremely common example, and one I’ve talked about before, is balancing attacks that do damage quickly in small chunks vs attacks that do damage more slowly but in bigger chunks. The faster attacks usually need to be significantly more effective in terms of damage rate to balance out, since strong single-hit attacks have the advantage of allowing hit-and-run tactics, of defeating a weakened opponent more quickly (potentially preventing a devastating counterattack), and of being more difficult for the opponent to effectively react to. Not only does how close you are to the success threshold of defeating your opponent drastically change what attack is most effective, so does the opponent’s mobility, preferred range, and their own most effective attacks.

A particularly interesting example of being forced to adapt your strategy is in heads-up collectible card games like Magic and Hearthstone, where how many cards your opponent is currently holding is a tremendously important situational modifier to your strategy. The more cards your opponent has, the more likely they are to have something that immediately counters whatever your general-case strongest available move is: This means it’s often worthwhile to lead with a generally sub-optimal move just to draw out your opponent’s countermeasures.

On a more meta-gaming level, there’s a curious kind of balancing that takes place in many competitive games. The stronger the move, the more players have learned how and prepared to counter it – and, commensurately, the weaker the move, the more unexpected it will be. This sort of self-balancing can only go so far, and it of course only works under circumstances where the weaker move is distinct enough from the stronger move that the same countermeasures aren’t effective against both. Unfortunately, these strange automatic balancing mechanisms don’t really work in single-player games – even if you could make an AI that chose its strategies in a manner indistinguishable from a human player, if the players knew it was an AI they would probably feel it was behaving arbitrarily and erratically – after all, why would it ever pick the obvious ‘worse’ move?

Despite that limitation, this principle isn’t restricted to competitive games. Often what separates a good game from a bad game, nearly irrespective of genre, is how much you have to pay attention and adapt your strategies to the situation the game puts forth. As I noted a couple of days ago, part of the role that difficulty plays in game design is in raising the stakes and responsiveness high enough that you have to adapt your strategy from your default in order to succeed.

How can we build to enable situational decision-making? There are two pieces: First, a system that generates diverse situations; second, a set of tools that gives a number of discrete ways to approach these situations. Many games present essentially the same challenge over and over – such as a room full of guys that need shooting. Under these circumstances, the player is likely to figure out one tactic they like and use it over and over again, which gets dull sooner rather than later. Many games present the player with a set of tools that are essentially identical – weapons with different attack animations and damage numbers but the same underlying mechanics. Now, no matter how diverse and interesting the situational challenges presented the player, they’re all just nails to be hit with a hammer.

These requirements may seem basic, but a lot of games honestly don’t do an amazing job of meeting them. Of course, a game can have many other fine traits, can even be fun to play – but if you want the systems of the game to be enjoyable on their own merit, there has to be something there for the player to react to. Without that reaction, that decision-making process, they’re just working an assembly line.