1

Designing with the Player Experience in Mind: Dynamic Difficulty Adjustment

Christian E. Erali and Professor Michael McCoy Jr.

potential for such a system and how a simple DDA system Executive Summary: One of the hardest aspects of game design can be implemented and used in most first-person shooters. is balancing the flow, challenge, and skill of a player throughout The final, polished system adjusts the difficulty across the game’s experience. As such, it is relatively easy for games multiple levels of challenge by assessing time per encounter and even encounters within these games to shift between either and damage to the player and adjusts difficulty through being too easy for the player, making them bored; or becoming altering enemy placement and enemy type. too challenging, frustrating the player. Both boredom and frustration can be the death of a game offering the players little reason to continue, especially if all the game can offer is death or boredom. However, using dynamic difficulty adjustment, games can regulate certain variables to fit the player’s current skill level and make the experience more enjoyable. Most dynamic difficulty systems work behind the scenes, adjusting unseen aspects of the game so the player never feels like the game is cheating them. This study investigates the positives and negatives of using dynamic difficulty and documents the creation of a simple system in CryEngine 3.

Index Terms: Dynamic Difficulty Adjustment, Design Optimization, Game Flow, Software Design

I. INTRODUCTION NE of the more interesting problems that O developers face during the game creation process is the issue of balance. Developers want to ensure that their Figure 1: Image showing the intended flow zone result of Dynamic product is consumable by as many people as possible. Difficulty Adjustment [2] Trying to create a game that challenges a long-time gaming veteran can push less experienced gamers away. Vice II. RESEARCH REVIEW versa, developing only for the less skilled repels the veteran Overview: and hardcore gamers. While there are several systems that allow players to choose a level of difficulty or augment the The potential for DDA systems has often been explored game’s challenge, these options can sometimes be lacking and researched by both academics and professional game and force the player to set a level of challenge with little if designers. While the academic papers focus on the any context. potential of DDA systems, their ideal applications, or outline some of the basic needs and calculations for such a However, one idea that has been around for some time is system to function. Meanwhile the blogs and articles of the a collection of systems referred to as Dynamic Difficulty game designers focus on certain examples and what caused Adjustment (DDA). DDA systems adjust various factors the system to succeed or fail. The next section discusses the and variables in video games to augment a player’s , CryEngine 3, and discusses some of the experience based on the player’s actions. These systems reasons why the engine was chosen as well as some of its allow developers to balance and build games that can limitations. Lastly, several games already use DDA change to fit a player’s playstyle and level of skill as the systems. For example, the Left 4 Dead Series with its well- gamer plays. known A.I. Director is a great example of a popular series using DDA. It the basic DDA systems turning an otherwise This thesis investigates many aspects of DDA systems simple game into a harrowing, new experience each time and explores their value through the development of such a the player played. system using CryEngine 3, the game engine used to create 2 [1]. Through dissecting and understanding the various iterations of the system, this thesis shows the 2

player’s action data and augmented a series of adjustable data points known as particle filters. These filters take the player’s data, in this case reaction time, and build a basic Texts: understanding of the player which it compares to the already existing player data. These systems then augments Due to the potential of such a system, there is a the experience and gives the player encounters based on the substantial amount of research about DDA systems. In her expected reaction, playstyle, and data of similar players. paper The Case for Dynamic Difficulty Adjustment in The system even changed the reward the player received, Games, Robin Hunicke of Northwestern University giving risky players more rewards such as more health and describes a DDA System in their game Hamlet that ammunition. measured various aspects of the player’s performance and then intervened appropriately [3]. According to Hunicke, One article that speaks about how exactly to apply and when it comes to First-Person-Shooters, players generally use DDA is Asher Einhorn’s article “Reducing Difficulty moved through four different states as part of the gameplay Dynamically and Invisibly” [6]. In the first part of his loop: three-part articles series, Einhorn discussed the need for • Search DDA to focus on more subtle ways to adjust difficulty and • Retrieve offered examples for both good and bad adjustments. To • Solving illustrate, a player in the heat of combat may not know • Fight exactly how much damage they take per-bullet from an Throughout a player’s experience, the systems behind enemy. This increase in damage, however slight, can help Hamlet endeavored to keep the players engaged by keeping to curtail a player’s more aggressive or reckless tactics and players in certain states and avoiding others. After the lead to a more challenging and rewarding fight. However, a system developed a database of information about the player does notice when the enemies’ health bar increases player, the Hamlet system checked if the player was or decreases due to the change in weapon effectiveness. “flailing” or having repeated difficulty with an encounter. Likewise, the player may not notice how quickly the If the system detected it, and to some extent predicted it, the enemies turn to fire upon them, as long as a hail of bullets system could then run a series of assessments and choose to eventually comes, but are more likely to notice when the either place extra health somewhere in the encounter, A.I. suddenly stops performing a certain power move or disrupt the enemy’s accuracy, or switch enemies for those action. Essentially, different aspects of a combat the player had already mastered. After some time, the encounters have their own level of readability and players system would check again to see if the change was are more likely to notice these aspects when they change. successful enough and assess its own alteration. Through At the same time, Einhorn also admits that in some cases this continual cycle of assessing, altering, and assessing, certain things can change, such as enemy number, as it Hamlet can mold itself to the current player. ultimately comes down to a designer’s judgement call.

While most designers tend to be positive about DDA systems, one designer named Ernest Adams and his article, “The Designer’s Notebook: Difficulty Modes and Dynamic Difficulty,” speaks out against DDA [7]. Adams advocated static difficulty in levels in response to a previous article written by Andrew Glassner. While Adams does not discount DDA systems as an idea, he did list off several potential problems they face such as, players feeling patronized, the potential exploitability, and how DDA systems do not work for all types of challenges. According to Adams, if a player notices that a DDA system is in play, the player may become frustrated as they feel the game is pulling punches and not giving them the challenge they wanted. A player may also decide to exploit a system if they realize one is in play. This is seen multiple times in massive-multiplayer-games where players who find a Figure 2: Example of a flow curve and how well a player plays based loophole tend to milk it for resources and loot until the on their stress level [4] developers finally correct the problem. Lastly, DDA Another interesting paper offers insight into designing systems cannot be applied to all games, such as puzzle DDA systems is Dynamic Difficulty Balancing for Cautious games, and may in fact lead to massive upsets in game flow, Players and Risk Takers by Guy Hawkins, Keith Nesbitt, leading to awkward encounters or broken levels where and Scott Brown [5]. This paper brings up the idea of players just flounder not knowing what to do. finding ways to reward “risky” playstyles. Through this study, the authors used a unique DDA system that took the Adams is not alone in his thoughts as there are plenty of games who do not have DDA systems and many are very 3

successful. Some of these games draw in players based on One last article, “Game Changers: Dynamic Difficulty” a strong gameplay experience, using an in-built difficulty by Adam Saltsman, briefly described three major types of curve to create interest and flow. Other games provide DDA that are used in three different games [11]. The first players with compelling stories and roleplaying Saltsman discussed was rubberbanding, which is most opportunities to draw in customers and keep them playing. prevalently seen in Mario Kart and other racing games. Ultimately it comes down to the game project and whether a Rubberbanding is a system or series of mechanics that game even needs or could benefit from a DDA system. restricts the maximum distance between the racers in first and last place. Mario Kart achieved this by adjusting the Similar to what Adams spoke about in terms of combat A.I. speed depending on which place the player was in; pacing, another important idea in game design is flow. faster if player is in the lead, slower if the player was farther Flow, originally defined by Mihaly Csikszentmihalyi, is the back; and by switching which items a player could pick up. idea of building levels and mechanics that work together to For example, a player in first never obtained a blue shell, give players a directed and focused experience. Sean the most powerful item in the game, while a player in last Baron’s article, “Cognitive Flow: The Psychology of Great had a much higher chance of receiving it. Game Design,” discusses in detail many of the major concerns and goals of flow including the four major aspects Another interesting system Saltsman discussed was the of flow. [8] The four aspects of flow are concrete goals, Resident Evil 4 difficulty system. A player begins the game goals fitting the player’s abilities, timely feedback, and by selecting a difficulty, then based on the player’s minimal distractions. Concrete goals is the idea of giving performance, mostly counting player death; the game the player a concise, directed target to achieve, all the better adjusted the difficulty within a range determined by the if this goal is within the player’s current skill level, initial difficulty choice. For example, in this difficulty something DDA systems can adjust for. Timely feedback system on a scale of 1-10, 6 was medium difficulty. If the provides players the necessary information to learn the player continually died and struggled through a section, the skills needed to overcome the current challenge. Lastly, the system might drop the difficulty to four, but no lower. If idea of zero or limited distractions means that the player the player exceled, the system raised the challenge to a should be able to focus on these challenges without risk of seven or eight, but stopped there, not wanting to reach a being confused of disoriented by extraneous details. This is point where the player could not meet the challenge. important for building a believable and rewarding Lastly, Saltsman discussed the A.I. Director in Left 4 Dead, experience for the player before integrating the DDA. which operated on a similar, but more complex version of the Resident Evil system. Instead of adjusting difficulty, Due to focusing the DDA system more towards the Left 4 Dead changed the number and quality of enemies as altering of enemy types and position, the foundational level well as adjusting health pickups and supplies based on the needed to provide a fun and interesting experience by itself. player’s performance. One article by Steve Gaynor, creator of Gone Home, entitled, “Basics of effective FPS encounter design (via Games: F.E.A.R and F.E.A.R. 2),” discusses one of most important aspects of strong encounter design, giving the player and the A.I. choice [9]. In the article, Gaynor pointed out the necessity to provide players options so they can effectively adapt to the A.I. they encounter. Players should also be able to plan, see the encounter or space before engaging enemies. That said, A.I. without some form of cover or ability to counter the player’s actions might as well be a stationary target. As such, it is important for designers to not just provide players multiple options during combat, such as flanking, vertical ledges, and cover, but also provide similar choices to the A.I.

Figure 4: Left 4 Dead Cover [12]

The Left 4 Dead [13, 14] series is a multiplayer zombie- Figure 3: Example Combat Encounter, Dungeons and Dragons [10] survival series where players battle their way through hordes of both normal and specialized zombies to get to the end of the 4 level. The game used an A.I Director, a specialized dynamic sure the game continued to entertain the player, Bethesda difficulty system that altered the positioning of ammunition implemented a system where the enemies leveled up and grew drops, health, and other important items. In addition, the A.I. along a similar growth rate to the player. This meant a director also controlled enemy spawn rates and encounter dungeon would still provide players the same or similar level areas, making sure each play through of a level was unique of difficulty to a level five player as it would a level fifty-five without disrupting the player’s experience or patronizing player. One unique aspect of the system was when designers them. placed certain types of enemies into the world; the enemy had a five-stage difficulty rating from very easy to very hard. This system had several advantages, allowing designers to focus on building interesting dungeons and experiences, allowing the system to provide player appropriate and interesting challenges [18, 19].

Game Engines:

During the initial development and research phase of the project, many game engines were researched and explored as potential tools to be used for the final project. There were three predominate choices that each had their own benefits and shortcomings when it came to the project: Radiant, Unreal 4, CryEngine 3.

The Radiant Engine, used by Infinity Ward and other developers to craft the series, was one of the first engines researched as a potential tool for the thesis. With an extensive list of available assets, a familiar layout in terms of tools and development structure, and competent A.I. Radiant had plenty to offer the project. However, due to a lack of immediate support in terms of problem solving and unfamiliarity with Radiant’s coding method, Radiant was the first engine marked off the list.

Figure 5: Resident Evil, Selectable Ranged Difficulty [15] Unreal 4, the game engine behind Paragon and the Unreal Tournament series, was also a promising tool for use in creating a DDA system. Being familiar with the Engine’s Resident Evil 4 and 5 [16, 17] have a user selectable level coding system, Blueprint, meant that more time could be spent difficulty system combined with a dynamic difficulty system. developing the system, not learning the engine itself. How it works is, when the player starts the game, the player Additionally, the engine had plenty of blocking out tools for selects their difficulty. As players play through the game, they the early stages of development and offered the most amount encountered challenges based on the difficulty they selected. of freedom as every system was a blank slate ready to be However as the player proceeds through the level, depending augmented and changed to become some new facet for the on how many times they die through a section, the game DDA system. However, this level of freedom also proved its adjusts certain aspects to be more or less difficult within a greatest weakness. With little background code to help make range. This way, a player who wanted a medium level enemies interesting, the DDA system could adjust everything challenge, will always be within a medium level difficulty, but it wanted, but the experience itself would be boring without never break into hard or sink into easy there by gaining access competent A.I. Additionally, there were little if any free or to a level of difficulty that eventually leads to frustration or available assets meaning new assets would need to be created outright boredom. This way the player still has some agency or bought from the online store. As such, the best engine for in which level of difficulty they want, but they are still this project quickly became CryEngine 3. challenged based on their skill. CryEngine 3 was a great editor for this project for several Both the Fallout and Elder Scrolls series use DDA in a way, reasons. The editor provided a fully functioning game, though it tends to be more pronounced and slower. Both the complete with a robust A.I system, environmental geometry, Fallout and Elder Scrolls series created by Bethesda are and other assets that just did not exist in many of the other Roleplaying games where players build a character and possible editors. The editor also had a visual scripting editor adventure through an expansive and creative world. As tool that added plenty of functionality but had plenty of quirks players slayed monsters, completed quests, and found treasure, and limitations that impacted and limited some of the system’s the player gained experience and leveled up, gaining new final design. Additionally, due to the game’s built in skills and abilities to aid them as they adventured. To make regenerating health system, relatively simple ammunition set up, and other factors certain aspects of the most common 5

DDA systems were not available to be explored and help aspects and factors the previous systems observed, but expand the effect and impact of the final DDA system [20]. essentially scoring the player at the end of each level.

For several decades now, game designers have worked to The original premise behind the DDA system for this thesis try to build for the player. While not seen by everyone as a was to assess the player based on their speed through a series perfect solution, many have researched and even built full of four, individual encounters. The structure of the average games with various types of DDA systems in place. While first-person-shooter, third-person-shooters and even adventure some are more extensive than others and some are more games, plays to this idea very well. As players progress successful than others, as game creators continue to explore, through a level, the player has opportunities to fight, shoot, build, and iterate, the capabilities of DDA systems can and kill, followed by low points with little if any action, continue to grow into the product the industry needs them to possibly even a puzzle or narrative section, depending on the become. game. This is part of creating an interest curve, giving players high points of intense action and combat interspersed with opportunities to breathe, catch up with the narrative, or just reload and rest a bit before the next big event. With combat split up into these preset encounters, it would be possible for a system to trigger a timer when the player entered a combat encounter and end once the encounter was completed, either through player death or reaching a predetermined point. Based on personal observation, more skilled gamers tend to have faster reaction and more efficient gameplay than less skilled players. Professional or even just skilled gamers move into spaces, acquire targets quickly and finish encounters with measured experience. However, less skilled gamers tend to charge into areas then back into cover, are fairly inaccurate, and take more time than their more skilled counterparts. It would be possible to time these different player types, then provide more challenging encounters based on this data. This focus on time, lead to the system’s tentative title “Clockwork.”

Working from this idea, the development on the system began. The scripting system for CryEngine 3 is a system known as Flowgraph, a node-based visual scripting language that later influenced the development of Unreal 4’s blueprint Figure 6: Crysis 2 Cover [21] system. The benefits of such a system allow for rapid prototyping, visual debugging, and data gathering. However, III. METHODOLOGY these preset nodes do limit what can be done with the system., A. Project Overview limiting potential data points to adjust and aspects to measure. It also impacts the structure of the system and how data is processed through it. For example, in CryEngine 3’s The artifact for this thesis was a four-encounter level built Flowgraph there is a section called math where all math in CryEngine 3 using assets from Crysis 2. While the level related functions are stored. Surprisingly, while there are both was no doubt important, and needed to be of high quality, its less than and equal to nodes, there is no greater than node. primary focus was to display the effectiveness of the While ultimately a small thing, it did impact the design of the underlying DDA system while still providing players an system, forcing the creation of long chains of checks to spawn enjoyable experience. As players played through the level, the appropriate encounter for the player. the system would gather data about the player then change future encounters so players could have an experience that better fit their skill set. As the DDA system was the central focus of the project, the designing of the physical level became secondary as the needs of the DDA system would no doubt influence the level’s final design.

Figure 7: Initial Version of Core Equation During DDA research, it was interesting to note exactly how many aspects and factors some games tracked to make The initial design for the Clockwork system worked off a decisions on what exactly to adjust. Some tracked player scoring system of a 1-10 scale, with 5 being the average. health, accuracy, and where the player was in the game, Though a somewhat lopsided scale, players began the test creating a more generic, overarching system. Others worked level with a score of 5. As players stepped into an encounter, on a more level to level system, still monitoring many of the a timer would begin and enemies would spawn. Players engaged the enemies, killing all in their path. Once the last 6 enemy of the encounter was dead, that particular encounter was considered over and the timer stopped. The player’s total elapsed time would then be compared to three different, preset times, as determined by internal playtesting, to assess the player’s skill for that encounter. Based on the player’s score, the player would receive either a -2, 0, or +2 to their total encounter score, meaning at the end of encounter one, the player’s score would either be 3, 5, or 7. Continuing, once the player crossed the trigger for the second encounter, the program ran a simple check to obtain the player’s current encounter score. The system then spawned an appropriate encounter for that score, 3 being an easier encounter, 5 being average, and 7 being the most difficult encounter. With the enemies spawned, a new timer began and once the second encounter was completed, the player was scored again with either a -2, 0 or +2. This became their newest encounter score and so on the system went. This allowed the system to adjust the difficulty of the player’s experience on an encounter to encounter basis. Due to the beginning trial and testing nature of this first equation, the guides for spawning the various encounter options were kept simple. They consisted of: was a Figure 8: Initial Overview Map and Key player’s score less than 7, equal to 5, or was their score 3 or less. This effectively broke the system down into a very While some simple hard, medium, or easy respectively and allowed for might argue that quick, efficient data gathering on whether the system was all the project working. needed was a whitebox to test During this time, there was some concern about what would and prove the happen if a player just ignored the spawned enemies. If the system, without player did just ignore the enemies, then the system would a polished level never trigger properly and the whole system would break. of some kind, Due to the early nature of the project, that concern was the project thought best answered with simple objectives to guide the would fail on player such as, “Eliminate all Hostile Forces.” The thought many points. was that using these objectives, the player would not proceed First, the project without first completing each objective there by keeping the would not player in the area until all the enemies were dead and the provide players encounter completed. While later playtests did show the an enjoyable experience. Granted, the focus of the project was objectives working most of the time, a contingency trigger set to provide a solid gameplay experience, but visuals and somewhat after the encounter was implemented to make sure atmosphere are just as much a part of gameplay as any other players did not just break the system either on purpose or by aspect. It is hard to feel like a super soldier saving New York accident. City from an alien invasion when all the “buildings” are checkered board. Another major necessity for the level to be a Another core aspect of the system that needed to be defined polished, functional level instead of just a whitebox was to was what exactly would the system change to create the make sure the system could operate in an actual level, not just different levels of difficulty. While many options were a test space. If the system was seen and applied only in a discussed, altering enemy loadouts through scripts, changing whitebox, certain aspects of the whitebox could easily cater to the environment, ultimately it seemed best to just get the the system’s ideal functionality, something that might not be system working on its own, spawning different options per- possible if built into an actual level. By eventually building a encounter but ensuring they were drastically different. Yet, polished level for Clockwork, the system would inform the once the system was proven viable, the idea of using the development of the level, but also the game’s assets and already built in capabilities of CryEngine’s A.I. Wave system limitations would inform the system’s final design as well. to create groups of various soldiers seemed a fine solution. Figure 8 is a top down map of the first whitebox space created These soldiers, being physical assets in the level, could then to test Clockwork. Primarily focused on just testing the be placed wherever to add another variable for adjusting, functionality, this whitebox mainly proved as a testing ground meaning the system allowed for the altering of the types of for various ideas, spawning techniques and flowgraph enemies in each encounter and their initial spawn location. systems.

Though it was not the central focus of the project, the final level needed to still provide players an engaging experience. 7

IV. DATA, ANALYSIS, AND DEVELOPMENT 20. This number was then added to the final value of the previous encounter or 50 if it this was the first encounter. The A. 2nd Iteration equation then divides the previous encounter score by 2 meaning it use the previous encounter’s score twice. This After several attempts at making the system work, and ensured the player was assessed to see if they were or were not building around the quirks of the engine, testers successful improving. played through the test area in different ways to trigger different encounters. While this system was very basic and After the first and second encounter, the equation used the served only to see if the engine would be able to handle the two most recent encounter scores divided by 2 and 3 mass spawning and other background mechanics, it was great respectively. This allowed previous encounters to influence for seeing how it would work in engine and pointed out some the final score, but have less and less impact as the player important pieces of the final structure. proceeded through the level. Players who get lucky or faltered in one encounter, would not be punished or hindered by the The next iteration of the system expanded on much of the system for too long of a time. Additionally, it served to keep original and worked to provide more options for the player to the player experience more even, reducing rapid changes. potentially encounter. Seeing as the first version of the system mainly existed as a testing foundation, now came the Lastly, depending the last encounter, the calculated value development and design to make Clockwork and actual was normalized to keep the encounter scores within a 1-100 system that assessed players over the course of a level, not just scale. The first encounter of the level was divided by 1.5 due a test platform. For starters, the number of averaging times to only having two encounters, effectively dividing the first was increased to 5 and the whole system was broadened to encounter by one and the previous levels value by two. Future double digits to allow for more varied encounter scores. This encounters were divided by 1.833 with the current encounter meant that our new average was set to 50. Additionally, now not changing, the previous encounter divided by two, and the each of the changeable encounters had at least six different 2nd previous encounter divided by three. options, one spawning depending on the player’s current encounter score. The reason there were six possible options This final number became the encounters final score for that for each encounter was due to the lack of a greater than encounter, which determined the next encounters difficulty function, the system forcing long lines of less than checks to level. It did this by, once the player stepped into the trigger to see which wave to spawn, and to offer just that little bit more spawn the next encounter, the system analyzed the player’s diversity in potential encounters. most recent score and compared it to a series of less than checks in ranges of 10. If the player’s score was greater than 90, they received the hardest encounter. If the player’s score was within the 80’s the system spawned a hard encounter. If the player’s score was within the 70’s they received a medium hard encounter and so on and so forth. There was some discussion about taking all the encounters of a level into account, but tests showed that including a fourth encounter did not impact the final encounter score enough to be included.

During this time, the showcase level was also under Figure 9: 2nd Iteration of the Core Equation development. The original premise for the level was the player was tasked to eliminate all hostiles in the area so the

US army can set up a staging ground against the Ceph Building off the tests of the initial prototype system, the invasion. The player proceeds through the streets of New second iteration equation maintained a more even transition York, fighting various C.E.L.L. groups before finally ending between one difficulty and another. The previous equation, at the level exit. While still in whitebox, much of the final while simple, shoved players from one difficulty level to playspace was well on its way to completion. Building off the another, with no appreciation for the player’s previous structure of the test area, the final level was to have four actions. To address this, the new equation takes a player’s encounters spanning over a large city block. previous scores and averaged them with the current encounter score to create a final encounter score. 1. Set Up Encounter = Get an initial reading on player

skill level to set the following encounter to suit; set in a Players began the level with a preset score of 50, which corner store of a larger building represents their score from the previous level. The first 2. Street Encounter = Longer ranged fight with cover for encounter was a static encounter in that it was not altered by enemies and player; first opportunity for DDA to alter the system at all and was used to obtain a baseline. Once the an encounter; set in a street modified to accommodate player completed the first encounter, the player’s elapsed time the infected by the government was compared against the five preset times. Depending on 3. Bank Encounter = Multi-floored fight with players how quickly the player completed the encounter as compared fighting enemies on the ground floor and second floor, to the preset times, they would receive a +20, +10, 0, -10, or - altered by DDA system 8

4. Park Encounter = Large enemy base with a significant E. Once the player completed the second encounter, they number of enemies to provide sufficient conclusion, could then proceed into the third encounter, held in a altered by DDA system large bank. F. The third encounter at point F focused on multi-level combat, with players fighting enemies on the ground floor and second floor of a large, open air bank. G. After all the enemies were eliminated, the players would leave the bank and exit out onto the street below. H. The player engages the enemy at the final encounter I. In the initial design for this encounter, the higher tiered players would encounter a stalled but still dangerous vehicle mounted turret which would spawn at point I and limit the player’s route of attack and deal out heavy damage. J. Once the player completed the encounter, they were free to end the level by walking to point J. While some of the original level did stay the same throughout development, as is often the case, much of the final design needed iteration and change to bring about the best, polished result.

While this equation was a good start towards giving the player a smoother experience, tests soon revealed a major flaw. The starting idea for the timer section of the equation was that a player who was of superior skill to the average gamer, often moved through spaces and encounters faster and more efficiently than their less skilled counterparts, something that is often seen in high level play of tournaments and such. However, the internal playtests showed that the primary’s particular style of play went against this idea. The primary, a 3-year veteran of law enforcement, focused on staying in cover, taking whatever shot the A.I. presented him, and firing accurately rather than charging in like the super soldier that he was in game. This left the primary with usually horrid time scores, but with little if any health damage or loss in accuracy. This showed that games like Crysis 2 specialize in giving players choice on how they interacted with and proceeded through the level. Some players might cloak and stealth their way through, others might charge in like a freight train, and others might stay in cover and play it safe. Ultimately, while the now working system successfully spawned different enemy collections per-encounter, the system was still failing to consider a player’s actions to assess and work appropriately. For the system to be successful, the Clockwork Figure 10: 2nd Iteration of Showcase Level system needed to be able to, or at least attempt to, accommodate whatever playstyle the player employed.

As for the critical path: B. 3rd Iteration A. Players begin at the player start and can observe the first encounter area from point A. B. Engaging the encounter, at B, players would receive Seeing the system could not work off time alone, their first assessment of their skills, triggering the timer development began to look at other assets and pieces of at the beginning of the encounter and ending once all CryEngine 3 that could be used to gather more information the enemies were eliminated. about the player and their relative skills. At first, the idea of C. Continuing through the store and back out into the checking if the player was using their various powers, mainly street, the player can rest for a time around point C, the armor and cloaking abilities, seemed like a solid idea. taking in some of the devastation of the city thanks to However, there were various problems with this subsystem the Ceph invasion such as not being able to reset the subsystem after the initial D. The player engages the enemies of the second trigger and what exactly would this provide for the DDA encounter. system. Is a player who uses their powers an unskilled player 9 trying to find an edge, a reactionary player who finds themselves in a sudden bad situation, or a player who is using their abilities appropriately, and should the player be punished for using their powers or applauded? There was also an exploration into checking the player’s accuracy. However, with the variety of weapons and their individual power levels, and a lack of available nodes to check for accuracy, this idea could not be explored. After much discovery and iterating, the first attempt at a health checking system was successful. Using a specialized node called a sequentializer, the subsystem checked the player’s health every few seconds to see if the player’s health had dropped below certain gauges. Whenever a player entered an encounter, the health tracker would begin checking the player’s health. If the player took damage, a negative relative amount was added to a what became known as the health buffer value. The idea behind the buffer value was to provide player’s a bonus to their encounter score if they avoided taking damage and as a simple way to include the health tracking system. If a player entered an encounter and avoided taking damage, the buffer value would be positive, adding to the player’s encounter score and pushing them into a higher tier of difficulty. As the encounter went on and the player took more and more damage, the buffer value would slowly decrease and eventually turn negative, subtracting from their score and potentially moving Figure 12: 3rd Iteration of Showcase Level them into lower tiers of difficulty. The buffer value was added to the equation for this iteration as can be seen in Figure 6. Everything else about the equation remained unchanged.

Figure 11: 3rd Iteration on Core Equation

During this time, the level also saw some significant development. While the level itself remained unfinished, many of the final pieces were locked into place and much of the gameplay was ironed out. In the above map, the green square captures the space that saw the most significant changes during this time. This included finalizing the gameplay for the first and second encounters and at least bringing the third encounter to a playable state. As for the level, itself, the pieces for encounter one were set rather quickly as there was only two corner store pieces and only one with the necessary structure. More cover and the second corner store were added as well, providing player safe places to advance towards the second encounter. The second encounter notably changed from a car filled street to more of a checkpoint area. This was predominately because the two different hazmat tents made overall better use of the space than a whole bunch of cars. By turning the second encounter into more of a security checkpoint, not only was the second 10 encounter improved immensely due to extra cover and better did not have enough impact on the equation to feel like it was spawn locations for enemies, it also just provided a better use being counted as equal to the timer. This meant while the of space. buffer value was useful in switching people through different difficulty tracks, its value was artificially implemented into the equation. The idea of the health buffer number also needlessly added to the complexity of the equation. How much should the buffer value be, why is it that damage is tracked going from zero and up yet scored with negative numbers? This made gathering data on the equation somewhat confusing. Add in the fact that the health counter system, mainly the sequentializer, was inconsistent sometimes firing randomly or skipping over certain health gauges altogether, a new system and new equation was required.

C. Final Iteration

The fourth and final iteration of the Clockwork system ended up being much simpler than was initially envisioned. Using a normalization equation, the new system took both

Figure 13: 2nd Encounter, Showing use of hazmat tents to create health and time equally into account. It also meant that the better gameplay scale of success shifted back down too smaller, yet more precise numbers, with zero being the average. The new health subsystem also relied on time, but simply compared the value The third encounter was the space that saw the most intense of the player’s health every two seconds. If the player’s health change. Unfortunately, as potentially cool and interesting as changed, then the difference was recorded and added to a total the two-floored bank encounter was, the CryEngine lacked the amount of damage taken during the entirety of the encounter. necessary pieces and assets to make the Bank function. The Once the encounter was over, the total damage was subtracted only pieces that existed were exterior pieces, meaning much of from the average total health damage for that encounter then the desired interior encounter was impossible. As such, a divided by that same amount. This lead to a float number that, carpark replaced the bank, providing for a slightly longer level if it was negative, meant the player did well and, if it was and more interesting use of space. The players traveled up the positive, meant the player did poorly. For example, let us say, carpark level by level, encounter a small side fight along the the player begins an encounter and starts at 100 health. At way. Moving to the top floor, players faced the third some point in the encounter, the player takes 10 points of encounter which evolved in many ways to focus more on close health damage, taking them to 90. The health subsystem quarters fighting. During this time, attempts were made to records those 10 points of health damage. The player ducks finalize the final encounter, but it was clear the encounter into cover for a second to allow for Crysis 2’s regenerating would need more attention and revision before finalization. health system to kick in, gaining them back up to 95. The player suddenly takes another 10 damage, taking them to 85. The health system records this damage meaning the player’s current total would be 20. Since the health system was only concerned about the damage taken, Crysis 2’s regenerating health never reduced the amount of damage recorded by the system. The player eventually completes the encounter and the health system stops recording their health for that encounter.

The same idea was also applied to the timing part of the equation, with the player’s elapsed time compared to a preset average time then divided by that same preset time. The averages for both health and time were compiled from a series of informal playtests. During these playtests, players ran Figure 14: Carpark that replaced the Bank encounter through the level as best as they could, encountering and fighting whatever encounter spawned as determined by their With the level moving forward and the system working skill level. Once the playtest was completed, the data was well, it seemed like the project would continue almost without analyzed and the two most extreme examples were removed issue. However, as more and more tests occurred, it became from the pool of results. This cropping of testers would catch clear that the health tracker subsystem and its implementation the more extreme cases or blunders that can occasionally arise had many problems. For starters, the addition of the health during first-person-shooters who might not be that counter was seemingly tacked onto the end of the equation and experienced. For example, one player got lost in one of the 11 encounters and added to their time significantly which warped the rest of their experience and potential data. As such, this data was cut. Now to keep it fair, a player on the other side of the spectrum, a very highly skilled player who mowed through each encounter, was also cut from the data pool. This left a more balanced data pool of 15 playtests that created a reliable sketch of the average player’s skill per-encounter. Playtests also showed that when given the objective to kill all the enemies in an area before proceeding, a majority of the testers completed these objectives without incident. There were a few testers who did proceed without kill all the enemies in an area, but these were a very small portion of the playtest population. Even still, the additional trigger to ensure the encounter’s completion was implemented later as a precautionary measure. In future iterations, it would not be Figure 16: 4th Encounter, Final Overview Shot impossible to use in game barriers and doors to further ensure players complete each encounter. The area that saw the most amount of change was the final Both the value for the health and time were then added encounter. After weeks of trying to make the park idea work, together alongside any previous encounters, which were the park was replaced with an office building and its exterior divided by 2 or 3 respectively, then divided by either 1.5 for foyer. The park as an idea was interesting in many ways, but two encounters or 1.833 for three. This last number became without the right messes and assets to really sell it, the space the player’s current encounter score which was used for the often looked strange and played oddly. This compounded spawning of different waves for the next available encounter. with oddities in CryEngine’s landscaping tool lead to strange, Essentially as the player completed an encounter, Clockwork mono-chrome landscapes where the grass of the park look gathered more and more information not just about how they stilted and dead, it was not something that fit in a polished did in that encounter, but took their previous actions into experience. As such, by changing to an exterior space of an account as well. This makes the system more efficient and office building, player still ended up with a large space to fight effective the longer the level and even allows the system to and a space that felt like it the desired anchor of enemy forces track the player’s abilities even between levels. This system in the area. Large, well defended with walls and cell was very simple to understand, easy to gather data from, and barricades, the foyer of the business building, once a lovely soon provided consistent information and results. place to sip coffee with a colleague and discuss the Johnson account, changed into a fortress of guns and steel, providing a polished final encounter for all players.

During this time, there was also the addition of what are known as “cool down fights.” In other first-person-shooters, there are plenty of these small moments that exist to keep the player on task or act as minor activities to keep the player from growing bored as they move towards the next major Figure 15: Final Iteration on Core Equation encounter. As these combat moments were incredibly small and easy, at maximum reaching three opponents, it did not During this iteration, many of the encounter layouts and seem necessary to include them into the DDA system as multiple options were revised and finalized. However, to players even on the lowest difficulty would be able to handle better understand the encounter system and the idea behind the encounters with ease . They mainly existed as a shot of altering the encounter make up and the enemy position, it is excitement and to make the showcase level resemble other best to first gain and understanding of the final version of the first-person-shooter levels. level.

This last iteration saw plenty of revision and finalization that helped to make it a polished experience. Figure 17 shows the final layout of the level and highlights the major areas of change with a green outline. First, the opening of the level changed slightly to provide players a sense of safety before they entered combat. Instead of spawning in the middle of the street where, on occasion, enemies would spot the player and engage, the player now spawned inside a small garage. This provided players a space to gather their wits, find resources, and prepare for the rest of the level. 12

and soon receive their objective for the entire level, “Destroy all Hostile Forces.”

Figure 18: Exterior of Player Start and First Encounter

B. Heading out the garage, the player’s gaze takes them right to their first encounter in a small, corner grocer. At first there seems to be only one lone guard, but upon entering the store, four more soldiers appear and a brief fight insues. While panicked survivors have left most of the shelves completely bare, the shelves and tight quarters provide plenty of cover for both enemies and player.

Figure 17: Map Overview, Final Figure 19: Interior of First Encounter The final version of the top down map, as seen above, shows not only the basics of the space, but also shows the critical path and the four major encounters. The critical path C. With all the enemies dead, the player proceeds through the of the level goes as follows: remains of the store and oout onto the street again. The signs of the invasion are fairly plane here with rubble and ruined A. The player begins the level at the Player Start, in a large cars scattered throughout the street. The player turns the garage for some sort of big-city store. The space is filled with corner and sees that part of the C.E.L.L. pallisade has boxes and crates, some of them torn into as survivors tried to collapsed, allowing for a direct assault on the next area. find anything that might help them make it to the next day. However, if the player explores a little they can find a more Two such survivors lay near the northern portion of the covered route that may give them an easy shot on one or two garage, gurgling out their last as they wait for the infection to guards before entering the combat zone proper. take them. The player can find a pistol and ammunition here 13

fenced off area providing the player and opportunity to restock if necessary.

Figure 20: Encounter 2, Checkpoint

D. The player engages the enemies in the second encounter, the first encounter that is effected by the DDA system. The Figure 22: Encounter 3, Carpark area itself not as destroyed as the previous section with cracked asphault, and some scattered trash. It is the C.E.L.L. F. Continuing up through the carpark, the player comes to the soldiers were setting up some sort of base or quarantine zone thrid encounter, a large contingent of C.E.L.L. soldiers taking to take in survivors before eliminating them. position on the third floor of the carpark. With plenty of cover

and routes for attack, the player has to be mindful of not focusing on one enemy, unless they want another soldier flanking them. This encounter can be rather rough as most of the enemies have grenades and can easily land one near the player. And in the confined spaces, if a player gets a version of the encounter with a shotgun enemy, it can be rather devastating. After completing the encounter, the player can continue on their previous route up the carpark, but is soon blocked by a door. However, this little side area provides players another opportunity to gather more ammunition and spawns a guard on the opposite side of a partially closed garage door, directing players towards the correct path.

G. Exiting the third floor of the carpark, the player can either drop down onto three soldiers on the street below, carefully Figure 21: First floor of Car Park; Tiny in-between fight pick them off from their high vantige point, or sneak down using the stairs. At the end of the small cooldown encounter, the player sees a small barricade and must jump over it to E. With the second encounter defeated, the player moves into proceed. Crossing over the barricade spawns the final the next area, a carpark. Largely entacked, the garage shows encounter. little sign of the invasion. Just up the ramp from the second encounter, the player encoutners two C.E.L.L. soldiers as a brief cool down fight, providing some action without starting a whole new encounter. There is some ammunition in a small,

14

Figure 24: Overview of Encounter 2

Above, there is a top down screenshot of the second encounter. In the image, each blue circle represents an enemy

Figure 23: Encounter 4: Final Encounter unit that is tied to the encounter. During a given playthrough, only a fraction of the soldiers are spawned in relation to the player’s current skill level. The enemy soldiers are split up H. The final area of the level is a business building’s exterior into A.I. Waves, the name of an entity in CryEngine 3 that foyer. A once calm and relaxing place to take a cup of coffee essentially holds several AI and acts on them as a group. in the morning, talk about the johnson account, or just sit and Below is an example of one wave for Encounter 2. rest from the busy day has transformed into a fortified anchor for C.E.L.L. operations in the area. With more guards and fortifications than any other encounter, the player must navigate through the encounter taking fire from many directions as they pick away at the defenders one at a time. Players have multiple routes into the area, either going straight through the front, jumping over the wall, or moving around and taking the encounter from the northern side. This encounter pushed players to their best and provided a suitable conclusion to the level.

I. Once the player completed the final encounter, they Figure 25: Overview, Encounter 2 Easiest Difficulty received a score for the entire level and are able to see what encounter score the player earned for the next level. With all the enemies dead, they player recieves a notification to Each green and yellow “W” represents a member of the proceed to the flares and a helicopter arrives to pick the player second encounter’s easiest difficulty, option 1. As seen in the up, ending the level. image, the enemies of the wave are spread out, and arguably not in the best situations for cover and safety. Additionally, the members of the wave are set in such a way as to act as D. Results three separate, smaller encounters within the larger, overarching encounter. This structure of smaller engagements Before development even began on the system itself, there within the larger one, allows less skilled players the was the question of what exactly the system changed between opportunity to defeat one or two enemies before the rest can difficulties to make one option easier and another harder. As bring their firepower to bear. What is not perceivable from seen throughout the research section, there were multiple the image is the type of enemy each “W” represents. things that could be adjusted such as placement of health CryEngine has a long list of A.I. that are available for use for packs and weapons, altering certain stats and abilities of building complex, coordinated encounters. However, wanting enemies, even changing the physical layout of some of the to ensure that an encounter was not suddenly broken by the encounters. And while the idea of altering an enemy’s health inclusion of a rocket soldier or something else, the level and stats was discussed, ultimately the system gained enough encounters predominately pull from the more generic selection divergence in difficulty by which enemies spawned and of soldiers. This leaves us with the following: where. • shotgun wielding soldier • a submachine gun wielder • a longer-range rifle • a captain • and (in a few special circumstances) a grenade launcher. There was some discussion given to adding Ceph enemies, a race of alien invaders, as part of the finale, but Ceph are not 15 only rather limited in weapon options, at least when compared they do not offer a lot of damage output. However, the big to the C.E.L.L. soldiers, but it seemed rather strange for problem for this encounter is the two shotgunners just outside players to be fighting C.E.L.L. soldiers throughout most the the central arch of cover in the center. If a player fails to level before suddenly switching to a whole new enemy type recognize that threat, the player can easily find themselves who deal more damage and have completely different tactics. flanked by one or two highly damaging enemies. While the Not to mention that any combination of the two factions rifleman is towards the back of the encounter, he is still not in would lead to far too many variables to calculate or control the most ideal spot, meaning while the soldiers can fall back adequately to gather adequate data. and take up a strong position, the intervening time allows the player to either eliminate the rifleman or counter act the Normally, encounters like this would more than likely have choice by moving out of line of sight. Another nasty trick that two shotgun soldiers towards the front, perhaps two can come into play is the shotgun soldier in the far back of the submachine gun soldiers in the middle, and the longer-range encounter. Due to his positioning, he will more than likely rifle towards the back. However, this wave is primarily made make his way into the hazmat tent, providing the player a up of submachine gun soldiers, with one shotgun wielder nasty surprise if they charge in unprepared. The encounter is towards the front of the encounter. This is because the still an easy encounter, but still provides more of challenge submachine gun soldiers are arguably one of the more than option one, allowing this encounter to be a strong, ineffective units against the player. Sure, the A.I. may put a average encounter for the average player. lot of fire and lead down towards the player and it may sound dangerous, but the relative accuracy and lower damage serve primarily to scare the player, force them into hiding as they formulate a plan of attack. The one shotgun soldier can be nasty, but it is only one and can end up providing the player a powerful weapon they can use to complete the encounter. While some of this is more theory than reality as ultimately it falls to the robust A.I., who may choose to flank the player or not, the encounter is arguably stacked in the player’s favor to provide them a potentially challenging, yet fun and frantic experience. Figure 27: Overview, Encounter 2 Hardest Difficulty

The above image is the same encounter as the previous two, except this is the hardest option for the encounter. In the easiest version, the enemies were spread out, and easy to pick off one by one, or battle in smaller groups. This version of the encounter is far more dangerous and not just because of the two additional enemies. First, the enemies are relatively more clumped together and are generally in better opportunities for cover and retaliation once they discover the player. The layout of the enemies is also weighted more towards the A.I. The two soldiers nearest the jeep can immediately duck behind there or advance into the incomplete hazmat tent to gain cover or charge the player. The one soldier in the completed hazmat Figure 26: Overview, Encounter 2 Average Difficulty tent has a shotgun, potentially severely hurting the player when they enter the tent. The predominate arch of cover in the The above image is the exact same encounter, only middle of the encounter is guarded by a captain, a soldier with outlining the enemies and layout for option 3 of six, a higher a larger than normal health pool and a powerful rifle. If the tier, but still middle of the road level encounter. As this is an player and captain begin to swap blows, it might allow the encounter that most of the average players receive, it is an other A.I. enough time to better position themselves and flank encounter that endeavors to push the player and provide them the player, again if the A.I. decides to do this. This layout of a strong experience, without forcing perfect gameplay. While enemies and cover provides both the player and the enemies at first, the only noticeable difference between the two options with plenty of tactical choices and with the right gun in the is the additional solider, there are multiple alterations that help right place, a player can be pushed to their limit, giving them a to make this an ever so slightly more difficult challenge. First strong and worthy challenge. off, the position of the soldiers is more clumped towards the center of the encounter meaning players can perhaps get an With this system of encounters and their various options, easy elimination off the front most soldier, but soon trigger the this leads to a total of 19 potential encounters and several other soldiers to take cover and move to engage. The soldier hundred combinations of encounters, depending on the make up is also different with three shotgunners, two machine player’s skill, in order to provide players an enjoyable and gunners, and one rifleman. The machine gunners do still compelling experience. However, not every game needs a provide a rain of lead, but not as much as the previous level as DDA system to create these compelling experiences. Many 16 games rely on a simple difficulty curve, increasing the Figure 21 shows all the potential options for each encounter numbers of enemies, introducing more challenging types of throughout the entire level. The chart shows that each enemies, all so the player is appropriately challenged encounter is designed to be slightly more difficult than the last throughout the course of the game. Wanting to provide in its track, often through changing the types of enemies that players a similar experience, the various layouts of each spawn. This system all works off the foundation created by encounter aim to provide this same curve of difficulty over the first encounter, which tests each player to assess their time. This way, even if a player stays in one track for the skills and obtain a baseline. After the system obtains a entirety of their experience, they are still challenged and are baseline, a player who sticks directly to the fourth option in pushed to improve their skills as opposed to dropping them every encounter, for example, first deals with six soldiers in down farther and farther until the game becomes boring. the second encounter, seven in the third, and ten total in the fourth. While this simply points out just the simple increase in the number of soldiers, the table also notes the enemy make up of each encounter and gives a brief description of the area and layout. So while encounter 3 might look harder than encounter 4 in some ways, the fact that encounter 3 is set in a closed off, but covered filled area and encounter 4 is basically a fortress with a fairly open killing field makes encounter 4 that much harder.

However, using Clockwork, a player may find themselves not staying wholly in one track and are expected to move from track to track. This is a concern with some DDA systems as players who get lucky in one or more encounters may accidentally find themselves in a difficulty level that they just cannot defeat. While there is the potential for this to happen with Clockwork, the structure of the encounters follow an upward trend, pushing players to become better over time. And should a player face say option 5 instead of option 4 which would better fit their skills, their poor score in the option 5 encounter would help correct the difficulty curve to something more in their favor. The player may experience a dip in difficulty, dropping them down a level, or even a plateau. If this is the case, the player may rise to challenge or fail again. This may mean that a fight might not have as much impact as originally intended, but the impact and gravitas of an encounter is relative based on the player’s experience up to that point. As such, through Clockwork can only set up these encounters and spawn them based on the player’s experience. Through averaging out the player’s scores and keeping their previous encounters in mind, Clockwork can help to mitigate some of the potential troubles, but not in extreme cases.

After testing, iterating, building, and testing again, the dynamic difficulty system titled “Clockwork” was successful in multiple playtests in spawning several different encounters as different players played through the level. Some testers played through it quickly, gaining the more difficult challenges but still having fun. Others struggled during the first encounter or two, but soon found the experience more to their speed and completed the level with a sense of accomplishment.

V. CONCLUSION One of the most difficult things for a game developer to accomplish in their game is balance. Whether it is a first- person shooter, an adventure game, all games require some measure of balance and tweaking to give the player the Figure 28: Encounter Table experience the developer promises. One small aspect of this balancing is the difficulty levels of single player games. The

17 developer wants to challenge the player enough that they, the encounters, the system can effective track player skill not just player, stay interested, but never overpower them and push through the player’s current level, but also between levels, them into frustration. On the other hand, the developer does augmenting the first encounter of each level appropriately and not want their game to be too easy, then no one finds it even potentially augmenting aspects game wide to provide a challenging enough to be worth finishing. It is a delicate powerful, yet subtle impact on the player’s experience. Yet, balance and one that Dynamic Difficulty Adjustment can help while it is simple, it works meaning that a system does not to solve. have to be some grandiose, 14 stage behemoth that assess itself and acts similar to a pseudo-A.I. to be effective. The The purpose of this thesis was to build a four-encounter fact is, a game system only needs to be as complex as it needs level built in CryEngine 3 using assets from Crysis 2. As to be to provide players the experience best suited to them and players played through the level, the system gathered data their playstyle. Also, figuring out a system like this was an about the player then changed future encounters so players enjoyable experience. It was frustrating at times, confusing in could have an experience that better fit their skill set. This many ways, but ultimately truly worth it to understand code multi-encountered approach, endeavored to provide players a logic and development. better experience as they played. Development of the system began like many projects, A system such as this has many applications and many testing to see if the tools would work. With the proper tools designers and academics have explored and even developed found, soon came layouts of the level, a testing boneyard, and similar systems. From Einhorn’s Hamlet that could assess, several attempts just to get the core idea of the system make a change, then check to see if the change was enough; to working. While promising at first, time, playtests, and Left 4 Dead’s A.I. Director that actively moved pieces of the iterations proved that time alone was not a final solution and level and spawned different enemies; there were plenty of the system required more. Iterations of various health systems potential examples and proven concepts to learn from. There came next until one was found that provided consistency. was also the surprising amount of different types of DDA Throughout all this, the level progressed, buildings and systems from Mario Kart’s rubberbanding system to Resident encounters evolving not just because the pieces were not there, Evil’s difficulty range death counter, each one handled level but to also make the best use of the experience and the system. augmentation and difficulty adjust differently. Then came the simplification, the breaking down of the system into its smallest part and rebuilding it to something Taking influence from some of these successful digestible and reusable. implementations, Clockwork began as a simple system that worked to off scoring the player’s times for the four central After many hours of testing and compiling of data, the encounters. The system would then take these scores and “Clockwork” dynamic difficulty system was rather successful. spawn different enemies based on those scores. At first, the While it does not track the accuracy of the player or adjusts system proved useful and showed some promise. However, as the actual physical spaces of each encounter, it still gets the development continued, it was clear more statistic tracking job done and with very little set up. The system works off two was needed in order to create a better system. Soon the idea basic variables, compares them to previously gathered data, of storing previous encounters and incorporating them into the and puts out a result that often challenges players, but does not scores came, then further refinements to the spawning system, mock them with its ease or beat them down with its difficulty. and lastly the health counter was added. And along with the It is a simple system that could be the foundation for a much system came the level, adjusting and developing based on the more complex system or implemented into more modern assets and knowledge gained through playtests. games with relative ease.

If time allowed, there are definite improvements that could be made to the system, some of which might be fixed just by using a more modern game engine. As discussed previously, being unable to assess a player’s accuracy or dynamically adjust enemy load outs greatly impacted the final design of the system. By using say Unreal 4, not only could new ideas be rapidly prototyped, but even the visual scripting language of blueprint allows for more freedom than Flowgraph. As for the Figure 29: Final Equation for Clockwork System system’s implementation, it could be very beneficial for the system to set up the core part of the encounter, then alter only Ultimately, the system is fairly simple. As seen in the portions of the encounter. For example, a player enters an equation in Figure 29, all it tests is how long the player takes area and normally must face off against three snipers and a to get through an encounter and how much damage they take few riflemen across a small ravine. The foundation of the during that time, compares them to averages gathered by encounter focuses on long range combat while staying mobile playtests, compiled, added to any previous data, then assessed. so riflemen cannot pin the player down and snipers cannot It analyzes two aspects of the player’s style and fixes future zero in. If a player is particularly skilled, Clockwork could encounters to try and match their abilities. Giving time and adjust the difficulty by spawning three snipers and four riflemen, giving the player a true challenge. However, for the 18 less skilled players, the system could adjust the encounter to Available: have only one or two of the snipers active at any one time, http://www.hindawi.com/journals/ijcgt/2012/625476/ stretching out the encounter, but keeping the encounter’s [6] A. Eishorn, (2015, 7, 31) “Reducing difficulty dynamically foundation intact. There is plenty that could be done to make and invisibly,” Gamasutra [Web Article], Available: the system more robust and figuring out the system’s http://gamasutra.com/blogs/AsherEinhorn/20150731/250041/ relationship to player deaths and checkpoints would be a great Reducing_difficulty_dynamically_and_invisibly.php first expansions. The system could also track the player over [7] E. Adams, (2008, 5,14) “The Designer’s Notebook: multiple levels, ensuring that certain aspects of the player’s Difficulty Modes and Dynamic Difficulty,” Gamasutra [Web experience were consistent. For example, if the system noted Article] Available: that the player was overall performing below average between http://www.gamasutra.com/view/feature/132061/the_designer the first few levels, Clockwork could adjust the later portions s_notebook_.php of the game appropriately by either setting the player into a [8] S. Baron, (2012, 3, 22) “Cognitive Flow: The Psychology different difficulty track, much like Resident Evil 4, or by of Great Game Design,” Gamasutra [Web Article], Available: taking whatever encounter the player is supposed to get and http://gamasutra.com/view/feature/166972/cognitive_flow_the bringing them up or dropping them down by one in order to _psychology_of_.php provide a better experience. It would also be interesting to [9] S. Gaynor, (2009, 16, 2) “Basics of Effective FPS further expand on the system with the use of different arrays, encounter design (via F.E.A.R. and F.E.A.R. 2) Fullbright averages, and times for each potential encounter. However, [Web Article] Available: this last suggestion would increase development time http://www.fullbrightdesign.com/2009/02/basics-of-effective- significantly so it is unlikely to see further development. fps-encounter.html

Overall, “Clockwork” is an example of a complex system [10] http://slyflourish.com/images/typical_encounter.jpg made simple. There is plenty of potential for using this [11] A. Saltsman, (2009, 5, 7) “Game Changers: Dynamic system in further projects to provide players enjoyable, yet Difficulty by Adam Saltsman,” Gamasutra [Web Article] challenging experiences. Available: http://www.gamasutra.com/blogs/AdamSaltsman/20090507/8 3913/Game_Changers_Dynamic_Difficulty.php [12] REFERENCES https://upload.wikimedia.org/wikipedia/en/5/5b/Left4Dead_W [1] Crysis 2, and EA, 2011 indows_cover.jpg [2] [13] Left 4 Dead, Turtle Rock Studios; 2008 [14] Left 4 Dead 2, Turtle Rock Studios; 2009 http://image.slidesharecdn.com/christophersonrobertcovertand [15] https://upload.wikimedia.org/wikipedia/en/d/d9/Resi4-gc- overt-110726112206-phpapp01/95/covert-and-overt- cover.jpg measures-of-engagement-within-an-educational-multimedia- [16] Resident Evil 4, Capcom Production, 2005 environment-25-728.jpg?cb=1311679443 [3] R. Hunicke, (2014, 8, 25)“The Case for Dynamic [17] Resident Evil 5, Capcom Production, 2009 Adjustment in Games,” Northwestern University, [18] Fallout Series, Bethesda, 2008, 2010, 2015 [19] Elder Scrolls, Bethesda, 2006, 2011 ResearchGate. Available: [20] Cryengine 3, Crytek, 2009 https://www.researchgate.net/publication/220982524_The_cas [21] e_for_dynamic_difficulty_adjustment_in_games https://upload.wikimedia.org/wikipedia/en/a/a1/Crysis_2_cove [4] http://gamasutra.com/db_area/images/feature/166972/figure2. r.png png

[5] G Hawkins, K. Nesbit, S. Brown, (2012, 4, 27) “Dynamic

Difficulty Balancing for Cautious Players and Risk Takers,”

International Journal of Computer Games Technology, Volume 2012.