Level Theory: Tom Clancy's Splinter Cell

Level Theory: Tom Clancy's Splinter Cell


While gameplay mechanics encompass the rules and tools of play, level design is the element that gives the game shape beyond its rules. Level design is often misunderstood as level geometry, but it is used as another term for challenge design. It encompasses everything from geometry to enemy and hazard placements. It is the design of a challenge-environment.

Level design provides for the challenges the player has to overcome while utilizing the mechanics of the game; it makes the rules and tools applicable to its environment. Systems in themselves can provide challenges as well; Tetris, for example, does not feature level design as its system is sufficient to provide procedural challenges without ever increasing difficulty.

Yet if we are talking about more traditional non-procedural design, level design is one of the most important parts of any game and even more so in the stealth(-action) genre. Stealth games have to walk an especially fine line in designing their challenges; their systems must be lethal enough to discourage non-stealth approaches, yet provide enough tools to overcome scaling challenges. This is why many stealth games provide the player with multiple, sometimes emergent approaches created by both systems and level design.

Tom Clancy’s Splinter Cell is one of the earliest stealth-action games that learned lessons from both Thief and Metal Gear Solid (and possibly Sons of Liberty, released just a year before), unifying aspects of both games into one.

Systems Introduction

Splinter Cell plays much like the Metal Gear Solid games: third person vantage with the ability to stick to walls and lean out of them, but it also features a more sophisticated movement element with sound and lighting playing a role similar to the Thief games. While the Thief games have the light-gem (now being replaced by a light-gauge, and later games having the noise-gauge filling a similar purpose for sound), Splinter Cell utilized the analog stick, or mouse-wheel to decide the speed at which the character was moving and provided a far more precise control over the character. While the game was certainly more precise in this department, it wasn’t entirely analog. There are only 8 speeds the character can move at and hence 8 (unmodified) levels of noise, which were later expanded to 12 levels in Chaos Theory.


Similarly to Thief, the game also utilized different ground materials to generate different noise, but expanded these into more physics-based and interactive hazards like suspended chains, fabrics, or broken glass. Shooting out a light could create an unfavorable situation later where shattered glass from the bulb betrays your position.

Splinter Cell also features a sophisticated lighting system, this time utilizing projected shadows from animated objects, and multiple colored light sources as graded illumination of the character. For a game from 2002 it features one of the best lighting engines and has aged quite well despite its low resolution textures and low polygon models.

SplinterCell 2014-01-20 02-12-56-70

Given that a lot of the game takes place in complete darkness, the designers provided the player with Sam Fisher’s signature three-lensed Night Vision Goggles. Thermal Goggles were also introduced to allow the player to see through smoke or other thin opaque materials like paper walls or drapes as well as heat signatures on keypads or mines (in later installments EEV and EMF vision was added). Like in the Thief games, these lights could be manipulated. The player could shoot out most lights or overload them with a specific gadget. Switches to turn off lights and other light-emitting objects like computers or monitors also existed.

The AI could recognize the change in lighting and would investigate; if an NPC sitting at the desk had the office light suddenly switched off they would investigate and try to turn them back on.

The enemy AI had a sophisticated sensory system similar to Thief, based on light and sound but also included movement as a parameter. The AI distinguished movement of objects that are not “common” and even understood things like trajectory. Throwing an object to attract attention could lead to multiple outcomes. Either the object went through the visual range of the enemy and triggers it to investigate towards its point of origin (the players last position), or alternatively it did not pass through the visual range and the AI would investigate the sound it made on impact (destination point). If, however, the object hit an enemy directly that enemy became immediately hostile.

Furthermore, the Splinter Cell AI also distinguished multiple alert and detection states that would inform its behavior. Simple noises or a visual glimpse of the player would trigger an investigation period in which enemies would coordinate to find their origin, however missed shots or other aggressive direct attempts the enemies like smoke grenades or gas from diversion cameras would trigger a hostile state that made the enemy aggressively shoot at the last predicted location of the player, or even possibly even trigger an alarm if it thought itself at a disadvantage. Similarly, if a body was to be discovered by the AI, it would trigger a state of active hostility and a triggered alarm. In later games the AI would even equip or use their own vision devices like night-vision or thermal goggles to counteract the utilities provided to the player.

Similar to Metal Gear Solid 2, guards can also be held up or even taken as a body-shield, but in Splinter Cell instead of dropping a dog-tag the enemies can sometimes be interrogated to reveal information about the level, objectives, resources, or access-codes in absence of a detailed map like Metal Gear Solid’s soliton radar, an invaluable resource.

An iconic element of Splinter Cell is Sam Fisher’s split-jump which he can perform in tight spaces. It is a similar mechanic (and similarly iconic) to Solid Snake’s box as it will usually conceal the player from sight, above hanging light sources, and can be triggered quickly as a way to disappear temporarily. From the split-jump position the player can fire his weapon or drop onto enemies to knock them out, although the same can be achieved by dropping from ledges or other elevated positions and is not context sensitive.


With these sophisticated tools there was a lot of approaches a player could engage in and the systems were sound enough to provide for varied emergent scenarios. However, for every emergent scenario there must first be a base-state from which to engage, and this is where the level design comes in to provide both challenge and a way to teach the player these nuances.

Level Theory: Mission 1 – Police Station

I have singled out this mission as it is the first “real” mission the player will engage in after the tutorial in which the system basics are introduced. The tutorial level is structured rather simply: first as an obstacle course to learn movement and navigation of the environment and second a simple introduction to the stealth systems in a controlled environment without risk.

The Police Station mission is kept in line with the tutorial’s structure, it features similar scenarios to reinforce the lessons from the tutorial. First is a simple environment navigation-test that starts the player out in a backyard where the gate is locked and can not be scaled. Detection in this segment is of no importance as the NPCs are civilians are not hostile, but it will introduce the player to the sound-effect that is played once he is spotted. The player is free to experiment running around the backyard trying to continue on his way in a safe environment. When the player finally reaches what I consider the second segment, he is dropped in an empty warehouse on fire.

Again, there are no enemies here, providing for a “safe” environment for the second part of the light platforming, but the level introduces something that was not taught in the tutorial: fire hurts.

This part shows that the player is vulnerable to environmental hazards like fire, where the game steers the player consciously into a room with a collapsing ceiling to make the player take damage. Upstairs, the game forces the player to navigate a smoke-filled room that makes the player take damage and obscures vision: a valuable lesson for smoke or gas grenades later on. However, it is shown that if the room has an open ceiling by shooting the skylight, the smoke will clear out. If the player crouches, he can avoid taking damage altogether from the smoke, showing again a way to circumvent later hazards without drawing attention. Once the player makes it through the door on the other side, the loading screen transitions him to the second part of the level.

The second part is an isolated sandbox introducing hostile enemies for the first time. The player starts elevated above ground level so detection from civilians is not an issue, and after the player clears a jump from balcony to balcony the first AI test begins. The threat is relatively low here as the two enemies present in this encounter are separated and the first one is talking into a phone, signaling distraction, with his back turned towards the player as they come out onto the terrace. Both enemies are lightly armed with simple pistols that will not inflict too much damage even when spotted and can be dealt with brute-force if necessary. This environment is still “safe” with a medkit waiting for the player near the exit.

The position of the first guard mirrors the first enemy encounter in the tutorial where a guard needs to be interrogated and the player is expected to emulate this behavior by sneaking up to the guard and grabbing him from behind.

Of course, other approaches are available for experienced players, like ghosting past the guard and entering the house leaving the enemy undisturbed. However, a first-time player will learn to associate a single enemy with a its back turned as an invitation for a close-quarters takedown and/or interrogation.

Once the first guard is dispatched, its body would lie in direct illumination and possibly trigger the other guard passing by the window to detect it and come out. This serves as a reminder to hide bodies away from sight or to only kill them in advantageous spots, a very important part of Splinter Cell as the game checks for unconcealed bodies at the end of every level-segment (which can trigger game-overs, a feature removed, and even joked on, later in the series).

Moving on to the apartment there are multiple entry points, though all of them lead into the living-room, with a running TV providing a light-source and then into a brightly lit hallway and kitchen. This part reminds the player that he can manipulate light and is forced to create openings for himself in order to proceed, either by elimination of enemies or by eliminating elements that could lead to detection. Most first-time players will use force to play through this encounter as they still did not pick up on the nuances of the mechanics.

Once the player reaches the last room they are presented with a locked door with a keypad. If the player was paying attention to the hints from their handler, they will realize to look for hidden compartments. This is the place where most players struggle and often get stuck trying to find the secret alcove. This teaches the player to scour the environment for information, like the computer in the living room, or pay attention to details like an undisturbed painting (while others are tilted) and to utilize the OPSAT for additional information and guidance.

Once the player has found the alcove and received the code for the keypad door, the way is open to rappel towards the adjacent roof and proceed through the door and into the next loading screen.

After passing through the door the player is informed that the restriction to not engage on street-level is lifted. The player is then presented with a locked door to lockpick and once through enters a larger space overseeing a yard from a safe position atop a set of stairs.

Here is where the game poses the player the first real test that has to be solved by utilizing all elements that have been taught previously. Up until now the game has reinforced its philosophy by presenting safe environments in order to teach, but here the gloves come off. Splinter Cell is a game primarily about navigating the environment as well as manipulating the AI to your benefit. The whole game up until now has reminded us over and over how important negotiating its geometry is, now it expects the player to apply this knowledge.

The initial part of the test presents the player with a forced scenario. Two police officers round the corner towards a drunk on the bench and engage in conversation with him blocking the path forward. Across the yard a lone civilian stands on his balcony looking down on the scene and in the side-alley beyond an armed civilian patrols up and down towards the yard exit.

This is in total 5 pairs of eyes monitoring the yard, far too many enemies to take on in a straight gun-fight. The player might even remember how much damage a single handgun can do from the apartment segment. The player is forced here to navigate the environment to avoid detection as any single mistake will be punished by at least two gunmen. The player must pause and think, and if he does he is rewarded with the drunk on the bench passing out, removing him from the equation of eyes staring directly across his path.

Once the player takes the time to survey his surroundings, he might realize that there are two sets of ledges around the yard he can traverse, or the flight of stairs leading down into a basement that can be used to bypass the police officers and hide in the brush on the other side. There is also junk lying around that can manipulate the guards into moving from their brightly lit spot. So, unsurprisingly there are multiple ways to tackle this scenario as the video below demonstrates.

[the audio cuts out in the last minute, apologies]

The placement of the enemies is especially insidious here to create threat by both vision and sound proximity. Shooting one of the police officers will most certainly trigger the other to kill you and alert the patrol in the alley while the cone of fire for your pistol resets. Similarly, shooting the civilian on the balcony will trigger the police officers to go hostile immediately (the killshot that I present in the above video is pixel-perfect and impossible from that range without practice).

This level of threat is achieved in the level design by having overlapping cones of vision and the AI being able to trigger detected states across NPCs. The civilian on the balcony has no weapons, however he will shout if he spots the player.

This is mechanically the first version of a camera and alarm system introduced in the mission. The civilian has moving vision up and down the alley as well as down the yard and will spot the player and alert other NPCs, just as cameras will later into the game, and his position adds difficulty to being able to eliminate his threat of vision.

Forcing an encounter like this can be sometimes seen as a bad design choice as it severely limits player options. However, it is a necessity to limit the player to force them to come up with solutions for the problem that circumvent the limitations; that is the definition of a challenge.


As the video above demonstrates, the level design leaves a lot of freedom and creativity to the player in contrast to the limits it presents in this encounter; it is walking the fine line of being challenging but having enough sophistication in its systems to create complexity even in seemingly restricting circumstances. Placing another two guards in this yard would increase its challenge twofold, but considering the systems present, the encounter can always be completed through skilled manipulation and navigation as the tools the player is given are direct counters towards challenges the game throws at the player. At the very base the game is about visual range, visibility, light, shadow and sound; this does not change further down the line.

The rest of the Police Station mission is far less interesting to analyze as it only further reinforces the basics taught in this specific encounter. I have consciously picked these first scenarios to illustrate the importance of creating a test through level design and how the simple flow of this mission attributes towards teaching the basic mechanics that will be universally useful in more complex situations further into the game.

It is safe to say that later games up until Chaos Theory only improved on these concepts and philosophies creating an even more nuanced and sophisticated game with even more interactions and better graphical fidelity in service of the gameplay (for example, shadows thrown on lightly opaque materials as a way to identify threats through thin walls). Unfortunately, after Chaos Theory the reverse trend started in the franchise and hit a complete low point with Splinter Cell: Conviction which had little regard towards systems or level design.


  1. Aeiou says:

    Level design is hard, corridors are easy to make.

    The new Thief game should be yet another excellent example of how godawful modern level design is.

  2. As bonus material, heres a pic of an SP level that I couldn’t find a use for in the article.
    Also its a picture for ants.


    Its the Vertigo Cosmetics map in Pandora Tomorrow, on the website of the designer.

  3. Gbelly says:

    Is Double Agent that bad? I find it on par with earlier titles and I really liked how role of undercover agent was integrated in gameplay.

    • Double Agent is “ok”, it would be a lot better if they didn’t recycle the same map like 6 times and the game would stop crashing on PC which made it almost unplayable for me. In fact my steam version doesnt even start and just crashes to desktop.
      Otherwise its largely intact mechanically although it was a step down from Chaos Theory regardless considering depth and options.

  4. -- says:

    good writing; but what’s the point of the tutorial breakdown? it doesn’t really add to presentation and lengthens the article.
    also if your point was just to demonstrate level design a simple overview of the level would suffice. its quite long for a vidya article.

You can use basic HTML in your post. Gather Your Party will never share your email address with anyone, ever.