Deep in the Shadows: Thief Design Analysis Part 1 – Light, Sound, and Stealth Systems

Our extensive design analysis of the Thief reboot begins with a look at its core stealth systems and artificial intelligence.

[one_third]Our extensive design analysis of the Thief reboot begins with a look at its core stealth systems and artificial intelligence.[/one_third]

Between Thief: The Dark Project, Thief II: The Metal Age, and a decade and a half of fan missions, the core design of Looking Glass Studios’ seminal stealth series has been thoroughly explored. In the ten years since Thief: Deadly Shadows there have been multiple entries in the Splinter Cell, Hitman and Metal Gear Solid series, each of which have taken the core concepts of stealth gameplay in different directions. Which path does Eidos Montreal’s take on Thief follow? Does it look to its progenitors for inspiration, or to the more recent successes of the games influenced by them like Dishonored and Mark of the Ninja? Or, does it attempt to stalk its own path through the stealth genre?

An understanding of what type of stealth game Thief is, and where it stands in relation to the genre as a whole, requires an examination of its design in detail. In would be misguided to ignore the influence The Dark Project and its successors had on the stealth genre, and so where it is relevant I will draw direct comparisons between Thief and the rest of the series whose name it bears. However, my goal is not to examine how well Thief functions as a continuation of the series, but rather whether it can stand on its own within the stealth genre.

A unified lighting scheme and an art direction focused on visual fidelity over obvious stylisation makes it difficult to differentiation between usable objects and scenery elements.
A unified lighting scheme and an art direction focused on visual fidelity over obvious stylisation makes it difficult to differentiation between usable objects and scenery elements.

Before we consider the design specifics of Thief, I want to talk about some of the general principles of the stealth genre, and the methodology behind my analysis.

Underlying stealth games is a power imbalance. You are underpowered and outnumbered. These are games of information management and exploitation. You overcome the challenges the game presents, and you redress the power imbalance – not through force of arms, but through subterfuge and guile. You succeed not because you are better armed than the forces arrayed against you, but because you have access to more information and the means to use that to circumvent or eliminate them.

One important way stealth games facilitate this exploitation of information is by granting you superior movement options. You have the means available to reach parts of a level that the enemy non-player characters can’t. You have a greater knowledge about the environment because you have access to more of it.

At their best, stealth games focus on the player and their own expressiveness. The way players use their superior information to devise and execute solutions says as much about their personality as it does about the underlying game systems. Creating systems that support this expression is difficult. It’s not uncommon for stealth games to instead fall back on offering predefined solutions to the challenges presented, a trap the original Splinter Cell is guilty of.

The way players use their superior information to devise and execute solutions says as much about their personality as it does about the underlying game systems

Two important concepts relevant to this principle of player expression are intentionality and improvisation. Intentionality is, broadly speaking, your ability to formulate plans and execute them. You want a specific NPC to go somewhere, to do something, and you use your understanding of your own abilities, the game environment, and AI behaviour to achieve that goal.

Sometimes things will go wrong. Even the best laid plans don’t always survive contact with NPCs. In those cases it’s time to improvise, to react in such a way that you can adjust the state of the world back to a point from which you can formulate a new plan and return to the intentional phase, closing the loop. For games to allow for improvisational play, they need to be fault tolerant. There needs to be enough options for failure recovery that you have the means of returning to the intentional phase without being thrown back to the last checkpoint. They should support partial failure.

Games that don’t allow for intentionality feature challenges with predefined solutions and little room for creative problem solving and experimentation. The aim is to discover and implement the correct solution in order to proceed. Games that don’t allow for improvisation have low fault tolerance, where deviation from what is expected or allowed leads to a hard failure state; a game over. The most common cause of bad stealth game design is failing to offer room for either intentionality or improvisation. There is one path through a given encounter and if you fail to use it you have to start over.

Both intentionality and improvisation benefit from providing clear information regarding your environment and the tools available, or at least from putting you in a position from which that information can be inferred. This readability is not about having perfect information, but about being able to quickly and accurately parse what information is available. Ambiguity and uncertainty still have an important role to play, as they encourage taking actions on partial information. You know what you want to do and have the means at your disposal of achieving your goals, but uncertainty about the state of the world or of elements within it creates tension as all the factors affecting your success are not clear.

Most of the boxes and ledges in this space can be climbed, despite the lack of any obvious visual signposting.
Most of the boxes and ledges in this space can be climbed, despite the lack of any obvious visual signposting.

How well a game supports these forms of play is governed by systems design and level design.

Intentionality is best supported by level design that allows you to obtain information about the world before choosing when to initiate action. For example, you enter a building through a first floor window and can analyse the patrol routes of guards below from a safe position before choosing when and where to engage. This is something stealth games share with a certain strain of the action genre, exemplified by the outposts in Far Cry 3. It is also supported by systems design that gives you multiple means of obtaining information about the state of the world, such as being able to determine the position of out of sight guards through audio cues or interface elements.

Improvisational play requires consistently implemented rules that allow you to react to changing circumstances quickly with a good idea of what the consequences of those actions will be. The outcomes need not always be exactly what you predicted but they should be discernable with clear reasons why something occurred the way it did.

Level design and systems design are dependant upon each other. Systems don’t operate in isolation devoid of context. An analysis of how Thief functions as a stealth game, and how it supports player expression through intentional and improvisational play, requires an examination of both forms of design and how they work in concert. This article serves as part one, a breakdown of some of the systems of Thief and how they function, the rest of the analysis will consider the level design more specifically and how these systems are engaged with in context.

Stealth Systems

The model used to represent the concept of stealth provides the framework within which success and failure is determined; it sets the boundaries of the simulation and what aspects of it players need to be mindful of.

In Thief, there are two primary means of determining how likely you are to be detected: your visibility and your audibility. The former is represented both by the observable ambient light level and by the UI. The latter can be determined only by directly listening to the sound of your footsteps within the environment.

Thief uses a Light Gem as an interface signifier of your visibility, it is entirely black, entirely white, or black with a white outline. This final state exists for feedback purposes only, in terms AI response and behaviour there is no concept of partial illumination; visibility in Thief is measured on a binary scale. The clearest indicator of which is the visual effects applied to the screen when you are hidden or become visible. Visibility is indicated by the entire screen flashing white as you move into the light, concealment by a swirling black shroud effect around the edges of the screen. That this shroud is present both when the Light Gem has an outline, and when it is entirely black, reinforces that the partial state is not actually a discrete state at all. The white outline is there to indicate a threshold: you are about to change state, until you do you are, for all intents and purposes, still concealed.

Like it's audio Thief's lighting model is not entirely free from bugs.
Like its audio, Thief’s lighting model is not entirely free from bugs.

Previous games in the series had a wider gradation of visibility. Though AI response and behaviour couldn’t cover that entire spectrum, there was a ramping up or down of lighting intensity as you moved from light to dark and back. By moving, you could determine the degree of risk you were exposing yourself to by the rate of change of the Light Gem.

With no grey area between being in the light and being in the shadows, your current state is immediately readable. This readability is achieved at the expense of ambiguity. There is no uncertainty about how concealed you are, there is no play in the system. It is strictly binary. Grey areas are fertile ground for player expression, allowing for a degree of risk; are you the type of player who always plays it safe, or one who is willing to push the boundaries, to become partially exposed to reach that one piece of loot?

The power of grey areas comes from their ability to support partial failure

The power of grey areas comes from their ability to support partial failure: you didn’t get through that section while remaining complete concealed, but you didn’t quite get detected; you knew there was a risk and chose to take it anyway.

It is not the case that all scope for partial failure has been removed from Thief. Rather, it has been shifted from the the domain of player visibility to that of NPC reactions. They don’t immediately respond to suspicious activity, and can be partially alerted. Functionally, this creates the same benefits as a visibility grey area, but the change from a player-influenced factor to an AI behaviour leads to a difference in how it is perceived. If you manage to get through an area without fully alerting any NPCs, it can feel like you succeeded not through skill or luck, but because the NPCs were slow to react.

Outwitting NPCs is stripped of tension if they appear unobservant or unresponsive. Success as a sensation derives from something you achieve not from something the AI was incapable of preventing anyway.

[message_box title=”The Sound of The City”]

As one of the two primary senses through which feedback can be provided, audio is an important part of a stealth game. Detection of threats and feedback as to your own movement speed and noise generation is vital to gaining an understanding of your environment. Unfortunately, the audio system in Thief is inconsistent and bug-ridden.

Environmental and NPC-related audio effects are victims of hard cut-offs, with sounds playing correctly at one location and then stopping completely when you take a step forward. Lines of dialogue will play over each other, making both unintelligible and obscuring any other sound effects that might be playing.

Positional audio is occasionally accurate enough for you to determine the relative positions and movements of NPCs, but often it simply lies to you. Dialogue that is contextually coming from the other side of a window will play from directly above your head. Footstep sounds will change in volume for no reason, as if the NPC making them has suddenly teleported into a distant room.

Together, these inconsistencies and bugs make relying on the audio system for feedback a challenge, as the information provided might arbitrarily change or disappear completely at any moment.[/message_box]

Sound too played a greater role in determining your ability to remain undetected in the previous games. Multiple types of surface and a range of movement speeds, each with associated noise levels, created a matrix of audibility that you needed to be mindful of when moving through a space. As the properties of a surface were defined by its materials, the noise you were making was a constant consideration rather than a special case one.

Audibility in Thief is determined by the speed of your movements, and only in a few cases by the surfaces you are moving across. Walking and creeping is quiet enough that in most instances it will not lead to your detection, whereas running will always generate enough sound to cause NPCs to become suspicious. The only surfaces that directly impact the sound of your footsteps are patches of glass and water more than ankle deep; only moving extremely slowly will ensure you are not heard when traversing them.

Carpets deaded your footfalls but this is only a concern if you are walking or running, given the ease of use of swoop there is little reason to ever walk anywhere.
Carpets deaded your footfalls, but this is only a concern if you are walking or running. Given the ease of use of the swoop ability, there is little reason to ever walk anywhere.

Confusingly, throughout the world there are small pools of water on the ground that will cause the sound of your footsteps to change when you move through them, but that don’t modify the amount of sound you’re making as perceived by NPCs. You can move back and forth across a patch of water directly behind an NPC and provoke no reaction.

There is little need to worry about the sound you are making, as you can swoop everywhere and only move slower when you encounter a specifically noisy surface. With such a limited impact on the underlying stealth model, your audibility becomes something you can largely forget about – except in specific special cases.

Artificial Intelligence

The behaviour of NPCs is the primary means by which feedback is provided to the player as to their success or failure. Stealthiness is a relative property, it requires an active antagonistic force.

NPCs in Thief have three distinct modes of behaviour differentiated by alterations in animation, movement speed, and dialogue barks. The first two of these stages also have discrete sub-stages, though the only feedback regarding which of these an NPC is in at any given moment is provided by the UI. These three modes of NPC awareness are represented in the UI by an eye symbol over their heads. As an NPC starts to become suspicious, the eye will slowly fill from an outline to a solid white symbol. This is accompanied by a whispering audio cue that remains even if the UI is disabled. As these are the only feedback mechanisms for this ramp up to alertness NPCs can appear slow to react.

Once the eye is full and they have become what the game statistics screen describes as “suspicious”, NPCs enter the first of the behaviourally differentiated modes which, for consistency, I’ll refer to as the Suspicious Mode. Signified in the UI by three white dots above the eye symbol, each additional event that provokes a reaction from an NPC turns one of those dots into a larger red pip. Upon passing through all stages of Suspicion Mode, the eye turns yellow and NPCs enter the second, and more aggressive, Search Mode.

Search Mode behaviour is differentiated by an increased movement speed, a more aggressive stance, and drawn weapons

When Suspicious NPCs move toward the area from which the suspicious activity originated, they will do so at their standard pace while maintaining an upright stance. Search Mode behaviour is differentiated by an increased movement speed, a more aggressive stance, and drawn weapons; guards lean forward with their heads projected as they actively scan the environment for you. The tone of their audio barks changes too, becoming less unsure, which is indicative of their increasing conviction that they did in fact detect something. Like the Suspicious Mode, Search Mode actually has three separate sub-states which are indicated only by changes in the UI and not in the animation or behaviour of the NPCs. The eye symbol changes from yellow, through to dark yellow, to a more pronounced orange staring eye.

Enough suspicious activity at once can cause NPCs to bypass Search Mode entirely and immediately enter Combat Mode. This usually happens if you are observed directly, or if another guard raises the alarm. Once enough suspicious activity has been detected to cause them to enter Combat Mode, represented as a crossed out red eye, NPCs will instantly become aware of your position regardless of what caused them to finally enter that mode – even if you have been causing distractions and are yourself hidden.

This is a problem with how the different modes ramp up from one to the other. There is no alternate mode after Search that models the behaviour of an NPC that is constantly observing something suspicious without an identifiable cause.

On Master difficulty NPCs instantly ramp up to two red pips upon entering suspicious mode.
On Master difficulty, NPCs instantly ramp up to two red pips upon entering Suspicious Mode.

NPCs that don’t become suspicious enough to enter Combat Mode will slowly cool down through the two preceding modes until they reset to their normal patrol behaviour. During this period, any suspicious activity will stop the cool down and initiate the ramp back toward Combat Mode. For a limited period after becoming suspicious, NPCs are quicker to react to stimuli. Eventually they forget there was something to be suspicious about, but it takes time for them to completely reset.

The animation and audio feedback of this cooldown is unclear. In most cases, an NPCs that has been made Suspicious and is now calming down behaves no differently to an NPC that has never become Suspicious at all, with the exception of the symbols above their head. If the UI is disabled, you may find two NPCs that appear to have differing levels of responsiveness because their underlying states are not communicated clearly. This poor readability of NPC behaviour is compounded by a variety of technical glitches that can see NPCs becoming stuck on geometry, or refusing to move even after becoming Suspicious.

NPCs in Thief are quick to react, and slow to forget

NPCs in Thief are quick to react, and slow to forget. Given that only a finite number of discrete NPC states can be represented, the ability of NPCs to eventually forget is important. Otherwise, you create a situation where every NPC is constantly Suspicious or actively in Search Mode. Such behaviour is useful in a limited capacity, such as highlighting the increased danger of a certain location, but a sustained period of high tension is both psychologically exhausting and limits your available options.

NPCs in Thief also react to proximity and motion. Though you can determine in absolute terms how close you are to an NPC and how fast you are moving, there is no feedback on how this is perceived by NPCs. There is no UI or other visual indicator that you can use to gauge the effect proximity and motion are having on your likelihood of being detected. Nothing tells you how close is too close, until it’s too late.

[message_box title=”Animal Instincts”]

Along with human guards and civilians populating the levels, there are also caged animals in certain locations. Either dogs or birds, these animals react to sound and movement, triggering an alert if they detect you. Their precise triggers, and the overlap in their abilities, is not immediately clear.

It is possible to swoop past dogs without alerting them completely, but this is dependant on the level design, as you need to able to move far enough away from them to prevent them becoming fully alerted and barking an alarm. A combination of level design and poor feedback leads to a situation where dogs are effectively static NPCs that create an area around them through which stealthy movement is difficult or outright impossible. Often, your best option is to either neutralise them with a Choke arrow, or find an alternate path. Oddly, though Choke arrows will incapacitate dogs, they will outright kill birds, so if you are attempting to avoid killing NPCs, you won’t be able to use Choke arrows on birds. Given the ease with which Choke arrows cancel out the threat posed by dogs, they can feel like just another equipment-gated obstacle, rather than an active AI-controlled threat.

The caged birds operate under much cleaner principles. They can detect you if you move too fast past them. This leads to interesting encounter design where it’s necessary to move past both a guard and a caged bird. You can’t simply swoop behind the guard as that will alert the bird. Instead, you have to slowly edge past both of them hoping that the guard won’t decide to turn around.[/message_box]

Leaps and Bounds

Player intent as expressed through the available control options serves as the primary means of systemic input. Movement allows you to gain a positional advantage over NPCs and is therefore a core aspect of stealth play.

Alongside standard directional movement, Thief features a freerunning system similar to that used by Assassin’s Creed. You can climb and vault over parts of the environment, moving rapidly across rooftops and through streets with only limited inputs. The exact means by which the environment is traversed, and where you move to next, is partially determined by the game interpreting your desires – which means that sometimes you will vault over a railing when you wanted to grab the rope hanging on the far side of it.

The clutter on top of this platform is enough to prevent you climbing up, despite there appearing to be room.
The clutter on top of this platform is enough to prevent you climbing up, despite there appearing to be room.

More of the environment can be climbed than is explicitly identified by the white paint and scratches used to highlight specific ledges; there is more scope for exploration than this obvious signposting suggests. Any object that has enough flat space on top to stand upon can be scaled, though what constitutes “enough space” is not always clear, especially from a distance.

Supplementing this freerunning system is your ability to swoop, to move rapidly and silently over short distances. Swoop is multi-directional, can be used to move around corners, and can even be cancelled out of if necessary. It’s a silent action, but still affected by the space you pass through. If you move across a noisy surface, or through an area of illumination, then you can still be detected. It has a stamina-related cooldown, but this doesn’t take effect for quite a while, making it overpowered. You can chain multiple swoops together to cross a large area in a very short time. One notable balancing factor is that you cannot chain swoops simply by holding the button down; each swoop has to be a distinct and discrete action.

Swoop has some similarities to Dishonored’s Blink, and the “Mark of Serenity” power from Mark of the Ninja

Swoop has some similarities to Dishonored’s Blink, and the “Mark of Serenity” power from Mark of the Ninja, both of which teleport you to your destination, compressing the space between into nothingness. Powerful and expressive, these two powers do, for the moment in which they are used, take you out of the simulated space; you overcome the obstacles ahead of you by essentially breaking the laws of physics. Swoop is a more grounded mechanic. It requires that you actually pass through the space itself, therefore whenever you use it you must consider that space in a way not necessary with either Blink or Mark of Serenity.

When using a mouse and keyboard, the swoop mechanic is (instead of being a dedicated button as it is on a controller) bound to the same key as jump and climb. This changes the functionality of swoop from one that can be used at any time, to one that is context-sensitive. When using a controller, climb, jump and run are on the left trigger, and swoop is on one of the face buttons. Not only is the functionality separated in this manner, but the button layout of controllers means that each action requires a different hand.

The systems of running and climbing have their own problems with readability and consistency, but the way in which the different movement functionality has been separated for mouse and keyboard input creates an additional set of problems. There seems to be little reasoning to this decision beyond the desire to allow for separate run and jump buttons, despite such a concept being at odds with how movement has been designed.

White paint on edges and white scratches on walls indicate places where you can climb up to a higher level, these indicated paths are not the only places you can climb though they are the clearest.
White paint on edges and white scratches on walls indicate places where you can climb to a higher level. These indicated paths are not the only places you can climb, though they are the clearest.

Swoop is possibly the best single mechanic in the game, but because of that it can easily become the dominant strategy. The means of countering this with stamina and the need for discrete key presses are not enough to prevent this becoming the default means of movement. You might be detected swooping through a brightly lit area, but it also grants you the ability to get far enough away that it won’t matter if you are.

Control over your movement is a fundamental aspect of stealth games. It’s vital to your ability to execute plans and response to changing circumstances. Though Thief’s freerunning mechanic is less prescriptive than it may initially appear, the moment you move in a manner you didn’t choose, or run impotently into a wall you thought you would climb, your trust in the system breaks down. When you can’t trust a system, your natural inclination is to avoid using it. You develop plans that mitigate it, rather than rely upon it. When that system is as integral as movement is to a stealth game that’s a problem.

Getting Handsy

Interactions are the other means by which player intent is translated into a systemic input. Movement and senses might grant you an advantage over the NPCs, but it’s through interactions that you exploit that advantage.

Your means of interacting with the world of Thief are through either direct manipulation, for which there are specific and dedicated animations, or the use of your arrows. There are tools that can be purchased and used on objects in the world but, once purchased, these operate in the same fashion as direct manipulation.

The animation of interactions makes the act of manipulating the environment a non-trivial and potentially fraught one. Interaction is not an abstraction; it requires the willingness to sacrifice control for a limited period and risk exposure. Unfortunately, whether you will be detected or not suffers from poor feedback, making it an often uninformed risk. Animations have been created to be performed from a set position relative to the object with which you are interacting, meaning you will frequently find yourself shifted into a position from which the animation can successfully play out.

You might become visible when interacting with objects, but even those that clearly make noise don’t register as such with nearby NPCs. An apparently audible action near a caged bird doesn’t provoke any reaction.

Every time Garrett’s hands reach out to grab an object or press a switch it’s a tacit reminder of all the other objects and switches that he can't interact with, all the other elements of the world for which a first person animation has not been created.
Every time Garrett’s hands reach out to grab an object or press a switch, it’s a tacit reminder of all the other objects and switches that he can’t interact with, all the other elements of the environment for which a first person animation has not been created.

Dedicated animations for each interaction not only define what can be interacted with, but also the manner in which those interactions can occur. You can’t lean around a corner and press a button, because there is no animation for that. You can’t snuff out some candles while hanging from a rope, because there is no animation for that.

In certain locations are vases and bottles that can be smashed to cause distractions, or broken unintentionally, causing NPCs to become suspicious. Though conceptually similar to the patches where glass is strewn on the floor, these two mechanics are unconnected; smashing a glass bottle does not create a patch of broken glass. Additionally, these objects only exists in the world as a means of player input into the system; they do not exist for the AI. If an NPC walks across a patch of glass, the sound of their footsteps doesn’t change. You cannot pinpoint the location of NPCs relative to a patch of glass or other surface; instead, you have to rely on the audio system, which is not always reliable. Oddly, NPC footsteps do change when they walk across water, however, in external environments, patches of water on the floor are so frequent, and irregularly positioned, that this proves of limited utility in terms of revealing their position.

Arrows allow you to interact with the environment from a distance, but much of the functionally provided is a duplication of that available through direct interaction. Water arrows can dowse light sources, but there are often more candles than torches in an environment and the former can be snuffed out by hand. Broadhead arrows can cause damage from a distance, but direct takedowns are faster and require no expenditure of resources.

Blunt arrows seem to only exist so that Broadhead arrows can be made more expensive, discouraging their use

There are also overlaps of functionality between the different arrow types themselves. Choke arrows can be used to temporarily stun NPCs and incapacitate animals, or to dowse torches; why they share an ability with the Water arrows, which can only be used to dowse torches, is unclear. Blunt arrows allow you to interact with objects and switches from a distance, yet that behaviour is provided by all the other types of arrow; there is nothing a Blunt arrow can do that any of the other arrow types can’t. Blunt arrows seem to only exist so that Broadhead arrows can be made more expensive, discouraging their use.

When not used against specific objects or NPCs all arrows share the same basic functionality. Each can be used to draw NPCs away from their current position; firing them at the floor or walls will cause NPCs to become suspicious and move to investigate. Aim too close and a NPC might enter Search Mode; aim too far away and you will waste an arrow without the NPC even noticing.

The network of interactions between objects in the world and the tools at your disposal is full of shared functionality and odd exceptions. There are eight types of arrow available, two of which (Sawtooth and Blast arrows) are direct upgrades of other arrow types, and one of which (Blunt arrows) provides functionality that is already offered by every other type of arrow, but at a reduced purchase price. Rope arrows open up access to different parts of a level, but these are special case items and can only be used on specific anchor points within the environment.

It’s not hard to see the possibilities for systemic depth that could have existed

It’s not hard to see the possibilities for systemic depth that could have existed. Arrows can be used to trigger switches from across the room, turning off lights or activating machinery, but they do nothing if fired at the light bulbs themselves. What if firing an arrow at a light bulb caused it to smash, creating noise and possibly a patch of glass on the floor below? You would have permanently disabled that light, but in a way that left obvious evidence, created an environmental hazard and alerted nearby NPCs. If patches of glass also had an impact on both the sound of your movements and that of NPCs, not only would you have created an environmental hazard for yourself – you would also have created a means by which you could determine the movement of NPCs through that space.

Combinatorial explosion is always a risk when modelling interactions between objects, but the range of options available in Thief feels incomplete. Too many logical interactions are not possible, and what tools you do have are prescriptive in their utility; they are instance-specific consumables, rather than means of modifying or disrupting the simulated environment in a systemic fashion.

There are few options for permanently changing the environment, if there is a patch of glass on the floor you can't do anything with it.
There are few options for permanently changing the environment, if there is a patch of glass on the floor you can’t do anything with it.

Your interactions with the environment can arouse the suspicions of NPCs. Opening doors and leaving them open can cause NPCs to investigate, though this behaviour is inconsistent; the degree to which a particular change in the state of the environment will provoke a reaction depends on a number of factors that are not easily readable. You can put a torch out, but an NPC will likely relit it again within moments; this shifts the actions you take into the short term. The interactions you have with the world are about provoking an immediate state change and then exploiting it. Increasing the pace of your engagement with the world is a design approach full of potential, but one that Thief never fully explores. Too many of the options available to you lead to the same outcomes.

You have a variety of tools at your disposal to directly impact the state of NPCs, but they all resolve to three basic interactions: distract them, partially alert them, or partially alert them while inflicting damage.

The scope for player expression is limited as so much of what you might want to do requires that somebody else has thought of it first

The scope for player expression is limited as so much of what you might want to do requires that somebody else has thought of it first and ensured that suitable objects exists within the world to allow it and that the necessary animations have been created.

The Dark Project provided means of permanently altering parts of the the environment to your benefit. Dishonored offered a greater range of objects with which to interact. Mark Of The Ninja focused on direct and highly readable manipulation of NPCs. Compared to these games, Thief offers few options for meaningfully interacting with the environment or the NPCs within it. Those it does provide are often temporary, leaving its highly detailed environments feeling like sets that reset after you have passed through, rather than worlds you can actively engage with.

Focus Testing

Focus is the catchall term for a variety of optional upgrades that have little in common, some change your interactions, some provide additional feedback, others  alter your ability to be detected. There’s little thematic or systemic consistency to Focus, as if the different components were once part of other systems that have now been grouped under a unifying concept.

In its basic form Focus will mark out in blue the parts of the environment that can be interacted with. Like alternate vision modes in other games, this serves as a means of compensating for poor affordances. The visual fidelity of the environment is high which makes it harder to determine what is aesthetic clutter and what are usable objects.

Focus can also provide additional feedback on the movement of NPCs and enhance your abilities with tools and weapons, but these additional powers require purchasing upgrades. The descriptions of what each of these upgrades does can be confusing, and the powers granted often of limited utility. For example the second “Stealth” upgrade includes the “Ability to stay concealed while using Focus” which suggests either that using Focus previously made you more visible, or that this upgrade would allow you to become temporarily invisible when using it. Neither of these are the case; using Focus doesn’t appear to have any measurable effect on your visibility. It’s possible what this means is that you will remain concealed if using Focus to pick a lock or perform some other direct interaction, however the ability to pick locks with Focus is related to a different upgrade and such a dependence, if it exists, is never explained.

The more visually complex version of lock picking is only accessible with a Focus upgrade.
The more visually complex version of lock picking is only accessible with a Focus upgrade.

[message_box title=”The Master of Unlocking”]When lockpicking, the “sweet spot” for each of the pins is either in one of the four cardinal directions, or for the more advanced locks, one of the four ordinal directions. The actual size of the “sweet spot” may be small, but can be located rapidly if you focus exclusively on the diagonals. Using the mouse to pick the locks actually serves to make them harder, as locating the cardinal points isn’t as straightforward as it is with a thumb stick and controller vibration.[/message_box]

Purchasing an additional Focus upgrade costs one hundred and fifty gold from the Queen of Beggars (on Master difficulty). Purchasing the second is three hundred gold; the third, four hundred and fifty. The cost of upgrading Focus escalates exponentially as a means of balancing its utility, but little of the functionality offered is powerful enough to warrant it.

Focus is given a position of relative prominence, with an upgrade system, specific consumables, and a dedicated space on the UI – but it lacks a coherent identity. It’s not clear what Focus is for, what it wants to be, or what purpose it serves. It’s a loosely-related collection of “helper upgrades” that you can spend money to unlock. None of the functionality it offers is vital. It’s the stealth equivalent of “driving aids” in a racing game, the abilities provided are primarily of benefit to players new to the stealth genre. Yet, rather than being selectable from an options menu, they are instead locked behind an upgrade system.

The Dropoff

With a binary visibility model and audibility determined by movement speed rather than surface properties, Thief feels like an attempt to replicate the clarity and readability of Mark of the Ninja in three dimensions. Even the presence of dogs, birds, and the flashes of lightning that occur on some levels evoke Klei’s 2D stealth side-scroller.

Some aspects of Thief’s design work to support this goal, but too often it is undermined by others that are inconsistent or arbitrarily restrictive. Without the grey area offered by an analogue visibility system, partial failure and player expression need to be supported in other ways. Mark of the Ninja achieved this by providing a variety of means of interacting with the world and the NPCs within it, restricting how many tools you could deploy at one time rather than the circumstances in which they could be used. With limited means to alter the simulated space, your abilities in Thief, for all their detailed animations, are left feeling artificial. The prescriptive nature of Thief’s design betrays a fear of “systemic messiness”; of allowing anything to happen that hasn’t been accounted for.

With limited means to alter the simulated space, your abilities in Thief, for all their detailed animations, are left feeling artificial

The wire cutters can be used to disable traps, but that's all they can do.
The wire cutters can be used to disable traps, but that’s all they can do.

Player expression is limited to being a Ghost, an Opportunist, or a Predator, these are the three ways in which the game judges your behaviour; another parallel to Mark of the Ninja. In the broadest sense, these break down into: avoiding interactions, interacting with the environment, or interacting directly with NPCs. The way this is presented within each encounter is not as obvious as having to choose between three different doors, though there are times when the level design can make it feel like that’s exactly what you are doing. If you’re a Ghost, you will use your Rope arrow on the provided anchor point to climb up to the ledge and bypass the NPC below. If you are an Opportunist, you will use a Water arrow to dowse the conveniently positioned torch and swoop past the NPC while he moves to relight it. If you are a Predator, you’ll swoop behind that NPCs and perform a takedown, allowing you to move on undisturbed.

If your preferred play style fits within the boundaries of either the Ghost or Predator paths, the restrictions on your actions can go unnoticed for several hours as, to an extent, they line up with the inherent restrictions of the Thief’s systems. If you try to experiment, to push against the boundaries, to be expressive, it soon becomes clear how few options there are and how much of the game is built on isolated instance-specific interactions.

If you try to be expressive, it soon becomes clear how few options there are and how much of the game is built on isolated instance-specific interactions

The three tools you can purchase – the wrench, the wire cutters, and the razor – highlight the prescriptive nature of the underlying design. There is no reason not to purchase all of these the moment you can afford them. Each has one specific use (two in the case of the wrench, but this secondary functionality is dedicated to collecting a type of unique loot), and once you possess them there is no reason not to use them whenever you encounter the object they interact with. You will want to disable every trap you find because there is no reason not to and rarely any risk to doing so. You will want to steal every painting you can cut out with the razor, because again there is no reason not to and doing so incurs no penalty. Your choice of whether to use these tools or not is irrelevant; there is no reason not too beyond self-imposed restrictions.

Thief is Mark of the Ninja without the scope for expressiveness and creativity, where the restrictions of the latter encouraged you to come up with your own ways around a problem, to improvise. Restrictions in the former just lead to compliance; the options available offer little scope for interactions that haven’t already been explicitly accounted for.

In Thief, your scope for intentional play is limited to those options the game has explicitly made available. You can’t climb up to gain a height advance and the increased situational awareness that provides unless some part of the scenery has been marked as climbable or a Rope arrow anchor point has been placed nearby; there’s no stacking of crates to create your own way out of an area. The solutions you devise to the challenges you encounter will rarely be ones you have come up with through creativity and experimentation, but pre-existing ones you have located within the environment.

There is little concept of materiality within the environments of Thief, a wooden surface functions the same as a stone or a dirt one.
There is little concept of materiality within the environments of Thief, a wooden surface functions the same as a stone or a dirt one.

There are elements, like swoop, that smartly engage with the underlying stealth systems and can be used in multiple non-prescriptive ways, but your core interactions with the environment, and the tools available, have been designed with a much less open “lock and key” mentality. Each environmental problem has a predefined range of solutions, sometimes, you simply can’t do something if you don’t have the necessary tool.

The design of Thief supports a limited range of playstyles, and is at its best when you focus on moving swiftly between patches of shadow and interacting with the world only sparingly. Within this narrow range, there is scope for engaging and memorable encounters, though as so many elements of the game feel like isolated special cases, the onus is on the level design to arrange the various elements into interesting and challenging configurations. Unfortunately, sometimes the level design not only fails to support this style of play, but outright demands that you abandon it.

This post is part of Deep in the Shadows: A Thief Design Analysis series. The link to the next instalment is below.

thief design analysis part 2

Or, you can click below to return to the Deep in the Shadows contents hub to find links to every part of our analysis.

thief design analysis contents

9 thoughts on “Deep in the Shadows: Thief Design Analysis Part 1 – Light, Sound, and Stealth Systems

  1. Justin, thank you for such a thorough investigation. Would you say that Splinter Cell: Chaos Theory also falls into the key-to-lock approach? I haven\’t played it yet, but did hear some very good mentions of it.

    Looking forward to the 2nd part of your analysis.

    1. Chaos Theory offers a much broader array of tools for manipulating the environment and the AI, providing greater scope for experimentation. It rarely feels like your opinions are being artificially restricted.

      A number of the concepts I discuss above (specifically Improvisation and Intentionality) are developed from ideas I was first exposed to in a 2006 GDC presentation (Designing to Promote Intentional Play) by Clint Hocking the Creative Director and Lead Level Designer of Splinter Cell: Chaos Theory.

  2. This is an interesting article that manages to articulate a lot of what bothers me about the new Thief game.

    One minor nitpick though: while the lighting system in the nuThief is definitely modal, it is definitely not BI-modal. As in there exists states between completely hidden and completely visible.

    I played this game on default settings so I spent a lot of time watching those little eye icons above the guard\’s head. If I stepped out directly into the light and directly into a guard\’s LOS, that eye goes from completely empty to completely red in a second.

    However, in partial light, even in a guards LOS, that eye will fill up before passing to yellow and then passing to red.

    I cannot say for certain, but I got the sense that the eye filled up more rapidly if I were in relatively brighter areas, in the guard\’s periphery, or closer to a guard.

    White, yellow and red seem to delineate the three possible states for the AI, and visibility seems to dictate the speed at which the AI transitions between those three states.

    Your poitn about the shroud, I think, is just poor UI design, which this game is plagued with.

    1. The lighting model may be more nuanced than the binary system I\’ve described, though I\’ve not seen definitive evidence of that.

      From my observations and testing, in terms of illumination alone there is no third state, the white outline on the Light Gem is only there to determine a threshold between the two states. What can make this confusing is that unlike previous Thief games you are not completely hidden when in the dark, as proximity now plays a more prominent role. If you are close enough to an NPC you can trigger their ramp up to suspicion even when in total darkness, moving into the partial state has no measurable impact on the speed of that ramp up, however moving into the light will cause it to increase dramatically.

      While researching this article I spent several hours specifically testing the stealth model and AI behaviour as they are at the heart of a stealth game. From that I determined that the factors impacting your detectability are: proximity to an NPC, position relative to an NPC\’s LOS; and current light level. The first two are analogue properties that are not communicated directly through the UI, the third is a discrete binary property that is communicated though the UI but somewhat inconsistently.

  3. A trend that is seen more and more often in L&S Stealth Games. Splinter Cell Blacklist for example does also not conceal you completely in darkness but makes it so the darker the spot your are in the slower the enemies detect you and the closer they have to be to start detecting you. Though in SCB there is definitely a system that uses multiple layers of shadows, although feedback is unfortunately only given in a binary form, meaning that there is some disconnect between knowing how visible you are and the real visibility.

    My main problem with Thief, and at that to a (much) lesser extend with SCB is definitely:

    1. The linearity of the maps/levels
    2. The lack of variable options on how to interact with your environment and manipulate it.
    3. The lack of alternate paths
    4. The lack of elements that add atmosphere and keep you invested in the world and the characters, make you feel like you are walking through a living and breathing world

    Chaos Theory had a really sandboxy design concept and managed to always give you multiple possibilities of handling a situation (example: retinal scanner: you could hack it or force an enemy to open the door for you, keypads: find the code, interrogate an enemy for it or hack it, AND it always even had a possibility to just go around such obstacles as well using alternate paths) and a lot of interactions such as hacking computers, picking locks, reading emails, interrogating enemies, shutting down the power, etc. A design that allows for a multiple-approach of situations WITHOUT necessarily having to interact with enemies. You can basically use the environment and manipulate it into giving you an advantage. The last point is the amazing way how all those little elements I mentioned turned the world into a living and breathing organism. You got to know your enemies, not only the villains but also the simple goons/sentries you were fighting, through emails, conversations, interrogations. WNM, the news channel gave you and impression on how the world saw the situation, how much was publicly known and how the publicity reacted to certain events.

    Because the core Stealth gameplay and the elements they have, some of the systems and the gadgets are NOT BAD. The problem is the inconsistency in various systems, the game-flow-breaking errors done, (sometimes) the lack of attention to detail, and last but not least the things I mentioned above, mainly going into the direction of a lack of freedom and opportunities to express yourself as a player, interact and invest in the \”game-world\” you are moving in, playing in.

  4. Are you planning on continuing the analysis?
    I do not know how many parts are planned, but it\’s been a while since you posted anything here Justin.

    1. Originally there were to be two articles this and one focused on the level design. That second article ended up at over 10,000 words so it\’s now going to be turned into a series (akin to my series for The Metal Age) with one article for each level plus introduction and conclusion.

      Unfortunately this type of analysis takes time. Those articles are almost ready and should all be completed by the end of E3. They\’ll go up then, likely over the course of several days.

Leave a Reply

Your email address will not be published. Required fields are marked *