Deep in the Shadows: Thief Design Analysis Conclusion
[one_third]Our deep dive analysis of the Thief reboot concludes with a holistic look at its stealth systems in the context of the game’s level design.[/one_third]
Levels in Eidos Montreal’s Thief reboot lack consistency. From the fundamental spatial design, to NPC and object placement, to contextualising aesthetics, there is no encompassing identity. Some levels feature forced progression through a strictly unidirectional space, others are open and self-directed. Sometimes you will be free to use any of the tools at your disposal, at others you are not even permitted to draw your bow.
Lacking a consistent implementation, the underlying systems of Thief are rendered unreadable and unpredictable. Intentionality and improvisation are undermined when the circumstances under which plans are made, and events reacted to, are liable to change without warning.
The world of Thief is not a coherent whole, but a series of smaller encounter arenas with strict limits on how they can connect to each other
Connections between areas are not organic; their construction as a sequential series of stealth arenas linked by soft loading transitions is poorly concealed. This obvious, and seemingly arbitrary partitioning of space destroys any sense of coherency the spatial design might otherwise have created.
The frequency with which you are required to hammer at a button to force open a window, or hold up a fallen beam so you can duck under, soon renders such interactions rote. Sometimes these transitions will fade to black momentarily if they are not taking enough time to cover the streaming of the next area, or they will cut abruptly to a hard level transition signified by a load screen. Even when handled smoothly, these transition points serve to constrict the environment into a series of choke points between larger sections. The world of Thief is not a coherent whole, but a series of smaller encounter arenas with strict limits on how they can connect to each other.
The nature of these transitions leads to a disjointed progress through the world, where a hard transition through a load screen can either mask a distance of a few feet or the entire width of the City. The exterior of one space can lead to an interior inconsistent with the structure as seen from the outside. A vast cathedral that looms above you as you approach is much smaller and more restrictive within.
There are locations that are built as single explorable spaces. These make up the core of a number of the chapters, but they are surrounded by introductions and finales that restrict your available abilities to ensure a specific playstyle. Follow these NPCs along their predefined path, escape across this burning bridge – but only along the path provided.
While exploring the main section of these chapters, avoiding or actively confronting NPCs, the rhythm of play is dictated by your own actions. Things occur at your own pace because of decisions you make. But every time you reach the conclusion to a chapter, the pace is altered artificially. There’s nothing organic to the tension created; your actions within each discrete section of a level are irrelevant to the sections that follow. There’s little entanglement, and the only consequences are those predefined. If you knocked out all the guards in the gardens of Northcrest Manor, the only impact that has is on the statistics screen. You never have to go through that space again, so it doesn’t matter what you did as nobody inside the Manor will react to, or even acknowledge, your behaviour. Consequences are artificial and inelegant, defined purely as numerical scores rather than changes to the environment or the dynamic of a level.
Consequences are artificial and inelegant, defined purely as numerical scores rather than changes to the environment or the dynamic of a level
Compare this to the infiltration of Dunwall Tower in Dishonored. Alerting the guards as you move through the grounds means that the Lord Regent will remain within his secure room on the roof, rather than returning to his bedroom, creating an additional challenge should you wish to reach him. Even this is a more “special case” example of entanglement than that offered by Thief: The Dark Project, where the design of spaces and your ability to explore then in a non-directed manner meant you could never divorce yourself from your previous actions. If you knocked out a guard in order to gain entry to a building, their body would remain where you hide it until you made your escape, to potentially be found by another guard, or simply to remain undiscovered, allowing you to leave the way you came in. The level design of The Dark Project paved the way for entanglement to flourish as an emergent property.
Within each chapter of the reboot, there is one objective which, over the course of the level, is broken down to a sequence of sub-objectives. With a rare few exceptions, these sub-objectives involve reaching a specific point in the level. Those that don’t will require you to grab an object or operate a switch. Objectives never rely on or exploit any of the underlying systems beyond basic movement and generic interaction.
You can steal the loot from the belts of NPCs, you can even take their necklaces without them noticing, you can disable traps and unlock doors, but only once within the story chapters are these abilities built into the objectives – and even then, it is optional. Beyond a handful of plot-specific items, you don’t need to steal anything to complete an objective. Of the few plot-relevant objects you do need to steal, several of the thefts occur inside the altered reality created by the Primal, a deus-ex-machina that is never sufficiently contextualised.

One recurring pattern in Thief, likely caused by technical constraints, is that the size and complexity of a level is inversely proportional to the number of NPCs within it. This can be seen most noticeably in Chapters 5 and 6. Both of these chapters feature spatially large levels, with multiple routes and scope for exploration, but they also feature either no, or a limited number of, NPCs. This is a problem for a stealth game, as you can’t use your tools and creativity to avoid detection when there is nothing to detect you, no antagonistic force to provide resistance.
The size and complexity of a level is inversely proportional to the number of NPCs within it
The ratio of space to number of NPCs is such that they are either densely clustered in a small area, or there are a limited number of them on long patrols or standing idly in isolated locations. In the latter instance, the speed of swoop makes it a straightforward matter to put some distance between you and them, allowing you to move around the levels with impunity. There are areas throughout Thief that feature smart design, with use of layout and lighting that has the potential to create interesting stealth encounters. Unfortunately, these are frequently undermined by this lack of NPCs. Some of the most memorable locations in terms of spatial design are those that are empty.
In the entire game, there are only three situations where actual physical keys are necessary to unlock doors despite there being hundreds of locks that can be picked. If you approach a door with a lock that can be picked, you have to participate in the lock picking minigame or find another way into that room. You will never be able to locate the key for that lock, as it doesn’t exist. Picking a lock in a room devoid of NPCs is a time sink; you can fail as much as you want, and eventually brute force your way through. Compare this to the sense of tension created when picking a complex lock next to a sleeping guard, or in a patch of light along another’s patrol route. The function of individual objects is most strongly defined when they are layered upon each other, which is done smartly in some levels, but not others.
Too often, variety is offered by restricting your tools and changing the rules governing their use, rather than through modification of the systemic elements of the game. Parts of Thief that are less restrictive encourage you to cultivate your own style and toolset, which you may then have to abandon for enforced escape sequences. It’s a similar problem to that created by the boss fights in Deus Ex: Human Revolution.
There is no systemic identity to the levels
There is no systemic identity to the levels. Consider Thief: The Metal Age; First City Bank and Trust was systemically about traps and random encounters. Sabotage at Soulforge was the ultimate hostile space, with metal surfaces everywhere and rooms teeming with Karras’ mechanical creations. This latter example, the conclusion of The Metal Age, generates its hostile atmosphere not simply from its size and the challenge presented by the patrolling robot inhabitants, but through the abundance of metal surfaces that are both noisy to walk across and entirely divorced from the design of the rest of the City, where metal and other hard noisy surfaces are a luxury. With hostility from both object interactions and thematic and narrative context, Sabotage at Soulforge is The Metal Age manifest: the death of the organic and the rise of the machine. Compare this to the un-boat of The Dawn’s Light, which is just a mess of thematic inconsistency and hostility created by bespoke rules.
At the high level, there is no arc to the level design. The plot presents dramatic twists and moments of tension that are not supported by the structure of the levels. Each chapter is isolated from those before or after, to such an extent that you could swap the order of the five chapters in the middle of the game and, beyond the cutscenes that occur outside your control, they would not feel out of place. You will still deal with the same limited set of NPC and object types in areas that have similar spatial layouts.
There is little sense of escalation in the design. One level will focus on a design conceit that another underplays or completely abandons. The House of Blossoms is the densest level with the most NPCs. The Baron’s Keep is the most obviously scripted and cinematic. Moria Asylum is the largest level in the game spatially, and Northcrest Manor is the strongest stealth space and one of the few levels that feels like a coherent location. Any one of those could serve as the climax of the game and it would not feel out of place. The systems, and the encounters in which you engage with them, don’t develop beyond the form in which they are first introduced.
The final three chapters feature the introduction of no new design elements. The tension of these final levels comes instead from forcing multiple objects into small spaces and limiting your movement options. These chapters are challenging because it’s harder to move around with so many NPCs in such a limited space. The utility of using thrown objects to distract guards, which until then has been a primary means of avoiding confrontation, is diminished when doing so alerts every NPC in the section.

In isolation, the core systems of Thief, the tools and movement options available and their relationship to visibility, are a mixed bag, but their interactions offer scope for engaging and challenging stealth encounters if layered together. Sometimes this occurs, but at others the disparate systems exist in isolation. A problem underlying the entire game is that there are rarely enough NPCs; those that are present are not in good locations to push against your actions. Instead, you have rooms that exist just for you to take everything from with no sense of risk or challenge. In some instances these rooms are ones that would logically exists in such a location, but too often they are simply used as dumping grounds for collectibles.
There are torches to dowse and candles to snuff out in areas where no NPCs patrol
There are torches to dowse and candles to snuff out in areas where no NPCs patrol. There are locks to pick in entirely empty rooms where nobody could ever hear you fail. The former is a redundant action, the spatial design of most levels means you will never return to a location so the time you spend dowsing that torch is never going to aid you at some potentially later time. The lack of complete concealment in dark areas means any dowsed torches only provide a limited safe space anyway. Locks on doors and chests in unguarded rooms are simply time sinks, preventing you from stealing whatever loot might be on the other side of the lock. If the keys existed within the level, you could spend time exploring the space and building up a mental model to aid you later, and only then finding the key. Discovery, and the rewards that come with it, fit conceptually the notion of being a thief in a way that a lockpicking minigame can’t.
Action games can create entanglement through ammunition limitations and weapon selection. But in a stealth game it’s often possible to get through a space using only your movement abilities, so entanglement can’t be created through resources. Dishonored’s way to solve this, along with also being an action game, was to make mana a finite resource. There are only a set number of vials that exist, and each power uses a specific amount of mana, so using a power at one point might mean you are unable to use it later. Thief’s Focus mode operates differently; it needs to be activated and then continuously drains until either deactivated or it runs out. It is also not as directly useful, and can even be disabled completely.
Resource management isn’t a concern, as in a lot of areas you don’t even need to use any resources to get through. Bottles are potentially the most important resource for stealth play, given how useful distractions are, but the game removes bottles from your inventory without warning, only providing access to them to at specific points. There are certain levels you have to get through without a bottle, because that’s how the game has been designed.

When Thief steps back and lets you explore, steal, and manipulate the NPCs, it feels, if not just like The Dark Project, at least like it understands what made that game work.
There are moments where the elements of Thief coalesce; you’re edging carefully through a guarded room, relying on partial observations of patrol routes to slip past unnoticed. You swoop around a corner as a torch bearing guard turns back toward you. A swipe, and you grab the loot from the table beside you just as the guard rounds the corner. Before your position is illuminated by the flames from his torch you climb up, over the bookcase and drop down on the other side into darkness. A swoop to the door, you pick the lock quickly, knowing another guard is due to patrol past any moment, just in time you open the door and swoop through, closing it behind you so as not to raise any suspicions.
Within the heart of the House of Blossoms, or Northcrest Manor, there are moments like this where Thief provides a freeform space for exploration, tense encounters, and creative stealth play. If more of the game had been made up of level design like this, it would be an easier game to analyse. Unfortunately, such sections make up only a small part of the game’s eight main chapters; their scale, complexity, and general quality vary with little sense of escalation or pacing, and they are connected by sections where your available abilities are either partially or entirely restricted, to the point where sometimes you are literally just pressing forward.
With more moments like the House of Blossoms or Northcrest Manor, and less like the Keep or the Hidden City, Thief would be a much stronger game. Given that intentionality can and will lead to improvisation, it is when the game puts you in situations like those found in the House of Blossoms that it is able to offer the strongest stealth play.
Thief breaks its own rules in order to enforce the type of experience it thinks you should be having
With predictable frequency, Thief breaks its own rules in order to enforce the type of experience it thinks you should be having. The result is a design that lacks conviction, one desperate to provide memorable moments but so terrified that you might somehow mess them up that it rarely lets you do anything within them. Instead of varying pacing through changes in the challenges you face, and the means by which you can use your abilities to find creative solutions to problems, Thief strips those abilities away and for most of the “important” moments relegates you to moving forward while things happen to you.
There’s no intentionality when you’re not offered enough information to formulate a plan, and the rules keep changing on you. Thief has strong moments, but too often it wants to direct you toward what it wants you to do, what it thinks will be an engaging experience, rather than allowing you to explore and let such moment emerge organically from systemic interactions and smart level design. Many of Thief’s problems can be seen in the space of a single chapter. As well-designed as the opening exploration of Northcrest Manor is, your confrontation with the Baron is bookended with the same restrictions and enforced tension that makes the closing chapters feel so flat.
There’s a third of a great stealth game lurking within Thief, a third of an average cinematic action adventure, and a third that could be good but doesn’t commit to anything. Unfortunately, you can’t just play one part. Every chapter includes some portion of all three, with sections arranged in a structure that lacks consistency, or is sometimes just incoherent, and can’t wait to change the rules on you.
Excellent piece to conclude a great series. I really enjoyed reading these!
I agree with you that there aren\’t enough NPCs, but I also have to ask: why the lack of enemy variety? The initial trailers advertised a shield guard, one that might use brute force and frontal defense to discourage open combat. That would make a big difference to stealth! Instead, all the enemies are just re-skinned swords- and bow-men. Also captains, who apparently cause disarray among their troops if you take them out.
An example of a stealth game with good enemy design was Mark of the Ninja. Guards with shields, guards with swords, elite guards that required tools, even nightvision troopers! Finding new guards was part of the fun, and you had to go up against them in ways that helped stealth.
Hell, just look to Thief 2 for an example, since there were robots and guards with knock-out proof helmets!
Over the course of writing this analysis I built upon the work of a range of incredibly smart people. As they have been a significant influential on my own design sensibilities, and this article series in particular, I wanted to include links to some specific reference material relevant to game and level design, particularly as it relates to stealth games. What follows is a non-exhaustive list.
Clint Hocking.
\”Designing To Promote Intentional Play.\”
\”The Interesting Thing About Bishops.\”
(http://clicknothing.typepad.com/Design/hockingc_GDC06_Intentionality.zip)
(http://clicknothing.typepad.com/Design/hockingc_GDC04_Bishops.zip)
Harvey Smith:
\”Systemic Level Design.\”
(http://www.witchboy.net/wp-content/uploads/2009/08/systemic_level_design_harvey_smith_.ppt)
Harvey Smith and Randy Smith:
\”Practical Techniques for Implementing Emergent Gameplay.\”
(http://www.roningamedeveloper.com/Materials/RandySmithAndHarveySmith_GDC_2004.ppt)
Randy Smith:
\”Level Building For Stealth Gameplay.\”
\”Design Fundamentals of Stealth Gameplay in the Thief Series.\”
(http://www.roningamedeveloper.com/Materials/RandySmith_GDC_2006.ppt)
(http://www.roningamedeveloper.com/Materials/RandySmith_GDC_2002.ppt)
Thanks for the extra links!
Love the work of Randy Smith and Harvey Smith.
Also, if you happen to have missed these, a really enjoyable listen! Podcast series of interviews with various staff from Looking Glass Studios, talking about their games (System Shock, Thief series, and beyond), and related games such as Trespasser, etc. Fascinating!
So many great people and games… They are listed here from the latest one to the oldest one…
http://gambit.mit.edu/updates/audio/looking_glass_studios_podcast/
Interviews include… Randy Smith, Daniel Thron, Terri Brosius, Eric Brosius, Ken Levine, Paul Neurath, Austin Grossman, Greg LoPiccolo, Marc LeBlanc, Dan Schmidt, Tim Stellmach, and Laura Baldwin.
AWESOME.
Would love to hear one with Doug Church, but alas, the warlock doth hide in his burrow.