Plot and Parcel: Procedural Level Design in XCOM 2

[Music] welcome to the talk plot and parcel procedural level design in XCOM - here's some stuff you haven't heard this week please silence your cellphone's also if you would be so kind as to fill out your review forms I would appreciate it my name is Bryan Hess I've been a level designer for over 20 years I got to start making mods for first-person shooter games in the late 90s like doom and quake I've worked on multiple genres first-person shooters adventure games mmo's and tactical strategy games my most recent work is as the lead level designer on XCOM 2 and its expansion war of the chosen I design the procedural level system for these titles here's a brief outline for what we'll be discussing in today's talk first we will

go through the history of XCOM and XCOM level design then we'll walk through the development of the procedural level design system in XCOM 2 i'd like to then briefly touch on some of the tools we use to help us in development then speak about some of the challenges we faced with the system during production after evaluating that we will discuss extending the system in the word that shows an expansion and finally we will step back and draw some conclusions about what we've learned XCOM is a tactical strategy game that is the game has a tactical combat component and a global strategy component in playing the game you alternate between these two modes the ExCom franchise started in 1993 with XCOM UFO defense which was released by Microprose for exuses predecessor this is one of my favorite games in high school and I have to admit I was pretty bad at it until I downloaded the unlimited money cheat

from CompuServe I would never have imagined being able to work in a game like this at that time what I was playing it so when the opportunity to do so arose for me in 2012 I jumped on it for axises first XCOM game was enemy Unknown released in 2012 and its expansion enemy within in 2013 it was followed by XCOM - in 2016 and were the chosen in 2017 this talk is about the Firaxis XCOM titles more specifically the tactical combat part of the game at its core XCOM is a game about hiding behind stuff and shooting aliens here is a brief crash course on the components of an ex-con level three fundamental components are the grid cover and fog of war levels are created on a grid of square tiles objects in the world either occupy or straddle these tiles provide things to hide behind an effect line-of-sight

standing next to these objects causes you to take cover behind them providing a defensive bonus there are two types of cover based on the height of the object low cover the half shield which grants a small bonus and high cover the full shield which grants a large bonus player controlled units can only see a certain number of tiles in each direction high cover objects also block line-of-sight players are only aware of enemies that their units have line of sight to and the map is revealed as players navigate through the level exposing groups of enemies so what makes a good XCOM level map layouts have a considerable amount of control over the challenge level that a tactical mission provides the same mission with the same number of enemies played on two differently arranged bases can range from a cakewalk to a rage inducing fiasco the three primary factors that we need to consider our object spacing cover density and line of sight

if cover objects are too sparse then it is impossible to engage enemies with your entire squad while in cover and this creates a difficulty spike if cover objects are too dense then map readability becomes an issue this increases playtime and frustrating in addition we observe a one wide rule when constructing XCOM levels that is there must be two empty tiles between each cluster of objects this guarantees that cover even at maximum density still allows for freedom of movement and minimizes camera blocking issues a good XCOM level has a sea of low cover and a smattering of high cover a successful XCOM player will maintain high cover with all of their units towards the enemies they are engaging high cover also blocks line-of-sight too much line-of-sight restriction has negative effects players will often travel past encounters they were intended to engage this results in

engaging multiple encounters at once another difficult despite the occasional multi encounter engagement can prove to be a challenging and memorable experience but we'd rather have this happen intentionally that have them occur accidentally as a result of bad level design now that we know what makes a good map let's take a look at the maps that shipped with XCOM enemy Unknown enemy Unknown ship with around 80 unique maps and the enemy within expansion added around 30 more most of these maps were divided into two categories urban maps were abduction missions took place and wilderness maps where UFO missions took place vast amounts of custom art was created for each level custom sidewalks and roads were sized to fit maps had unique signage and cover props one of the most expensive aspects was that each building was custom build buildings were easily the most resource intensive part of tactical maps

internally we were pleased with the look and feel of these maps however they took quite a long time to put together the average EU abduction map took a number of man months to complete in a single campaign playthrough you could play so many tactical missions that you would start seeing the same maps over again on your second playthrough everything started seeming very familiar in a game with a lot of replayability repeat were one of the top reasons people burned out on the game the dlcs raised the art bar on the maps if we sold 80 maps for 60 bucks and we're charging seven bucks for three maps these better be pretty fancy maps and our expansion maps can't look worse than our DLC ones it was an arms race this just put us further in the hole timewise the image on this slide is one of our DLC maps called deluge it was beautiful and iconic but aside the vehicles it was

almost entirely custom one-off art and it took a considerable amount of time to create one of the positive advancements of the expansion came as a result of an experiment in an attempt to extend content cheaply an enemy with him we took some of the base game maps move the player star to the other side and altered the time of day enemy Unknown levels were mostly large enough that the player never made it to the other corner of the map before the mission ended you start in this space you've never been in before and eventually you get to places you recognize but from the opposite direction they played like totally new levels recognizes recognizing these faces again in a different way often added to the experience not taking away from it as we had feared both of these concepts that is randomized player starts and multiple times of day for each map were added to the pool of ideas for the sequel as ways to increase variety now that we've gone over the

history of XCOM previous to the introduction of procedural level design I'd like to shift over to its incorporation into the XCOM tactical game starting with pre-production on XCOM 2 in 2013 we establish some goals for what we wanted out of the procedural system before we started developing it and pre-production the first goal was to remove expensive environment art as a bottleneck from getting more maps in the game we're not staffing up to make the sequel but we need to ship more maps to the game this time around another goal was the pursuit of a concept we called play forever which is that you'd never see the same map twice while playing we'd like to remove Matt fatigue as a source of player burnout finally we wanted to switch to using modular building kits which allowed the assembly of structures using a generic library this is an old concept that has been discussed in detail and a number of excellent GBC talks this wasn't directly

related to the procedural system but we knew that we were going to need a lot more buildings and since they were so expensive we had to find some way to speed up that process this was the original idea of how to do procedural maps before we actually started working on it we would start with a large map and cut holes where the buildings were we would create the multiple buildings and separate mat files and stream them in were the holes in the other map work we called the master map with the holes in it a plot and the streamed in building maps as parcels in the image on the screen the blue and grey checkerboard boxes each represent individual parcels you're going to see a few more checkered board boxes during this talk as I mentioned earlier enemy Unknown shipped with two major plot types urban and wilderness we wanted to expand the number of plot types to five for XCOM - the plot types we wanted to

make for city center wilderness small towns shanty towns and alien facilities the images on the screen are what these plot types ended up looking like when they shipped with XCOM - the additional plot types were added to help flesh out the universe and tell the story of the alien occupation of Earth in the last 20 years since enemy Unknown we were betting that we could get all of this additional content done much faster than the previous game due to our shiny new procedural system even though we had not yet proven it could work now that we had our expanded list of plot types we came up with some initial rules and how these plots would be constructed because there was still a significant art cost in creating a plot even with the buildings removed the goal was to have as few plots as possible to support that goal plots needed to support all missions they also need to be large enough to support the mission that demanded the most amount of

real estate ballooning their size the original plan for enemy reinforcements was that they would walk in from a road on the edge of the map this required roads exiting the gameplay space at regular intervals we also plan to add large units which occupied multiple tiles and needed paths free from cover objects in order to travel around the map which meant no buildings in these spaces this means roads around every structure as well as the edge of the map large unit and reinforcements support made the maps mostly roads roads were in the plot and required an art pass in every plot more roads on a plot meant more art time per plot created the image on the screen was one of our first prototype city plots note the preponderance of roads and the significant square footage here are some rules that we came up with for our first round of parcels parcels were divided into three buckets based on size small

parcels were 12 tiles squared medium parcels were 12 by 24 tiles and large parcels for 24 tile squared on the right are three shipping city center parcels one of each size the idea behind this was that since our plan involved punching holes in the plot for just buildings we needed more granularity over the sizes the buildings could occupy so that we weren't wasting any space the buildings that we evaluated from the enemy Unknown levels were all slightly different sizes but could fit within one of the three buckets we could ensure the buildings could both occupy almost all of a given parcel and be consistent sizes by using our modular building kit which was based three now we had three separate buckets of buildings to make sure we had enough of because large parcels had four times the real estate of a small parcel it stood to reason that they would take four times as long to create in the interest of replayability we wanted to make more smaller parcels and fewer larger ones we

also wanted a large parcel to have the potential to appear on every plot to try and accommodate this we added the concept parcel subdivision a large parcel slot on a plot could spawn instead as two mediums or as a medium and two smalls this way we could worry less about which building sizes we needed more of since the building slots were somewhat interchangeable to maximize replayability we also had parcels loading into spaces at a random cardinal orientation so we've made our first set of plots and we have to evaluate our work before moving forward dressing up all of the rows on each plot was very time-consuming most missions required less space than the plots were built to this was bad because players use the edge of the map to help them figure out which direction to go and in a large map you never see the edge putting the onus on the mission UI to keep players heading in the right

direction it also made load times longer memory footprints higher and performance worse even with the building's cut out and able to be shared between plots this still wasn't a huge time savings and tactical map creation because the parcels were all roughly the same size and shape as the other parcels in that bucket combat wasn't much different when one building was substituted in place of another we weren't there yet we started looking for solutions to some of the plot problems that we were having what if we took the parts of the plot that weren't parcels and procedural eyes them in a similar fashion we were already building road networks from modular Road kits what if we saved off a few variations of each piece from the kit as a separate map rearrange the cover layout and sent it through our art pipeline thus the idea of plot covered parcels or pcps was born the rules were slightly different than parcels though unlike parcels it was

okay to reuse a PCP multiple times in a tactical map they were small and generic enough that unless two identical ones were side-by-side and facing the same direction you couldn't tell that we were recycling content furthermore by divorcing the art from the map layout we could create any number of plot layouts without additional art cost this was the first major breakthrough cedral development the images on this slide were the results of the first PCP prototype plot they were screenshots of the same section of a single plot loaded six times we were pretty excited about the variety we could get out of a single plot with all of these new systems in place PCP has resulted in a massive time savings for the environment artists as they could Metis make a set of one to three dozen pcps to populate in any number of plots it also allowed level design the freedom to experiment much more with level layouts as assembling a

plot of parcels and PCPs could be done in about an hour each plot type had its own PCP sets for example the small town Street set had 24 pieces four courts two corners eight straightaways of two different lengths and ten intersections of various types with streets boulevards and train tracks the picture shows one of our first stabs at a complete set of PC ps4 the wilderness plot type there is a set of dirt roads divided highways streams and woods DCP piece variations allowed for a variety of gameplay as well as visual variation as the cover layout was different in each very variation after the initial round of building construction we took stock of the buildings we had made and how they were working in game level designers were struggling to build building interiors that had enough space to sell a particular theme while obeying all of our gameplay restrictions on layout and the space provided it wasn't so hard to

fill out a large parcel but it was sure hard to make a coherent space out of a small or medium one because we had to leave unoccupied tiles on the edge of each map in order to observe the one wide rule a 12 by 12 parcel could only support a nine by nine building okay so let's retrofit all the parcels to be base 14 well we can fit a 12 by 12 building and a small parcel now so the buildings are easier to make coherent spaces out of but the gameplay still isn't there the picture shows one of our early prototype buildings using some old XCOM one assets and test pieces there's nothing to hide behind next to the building while moving down the street gameplay testing was showing that buildings played better with cover objects on the outside of them in order for you to be able to fight down the more open spaces adjacent to them in order to accommodate that we would have to shrink a building on a 14 by 14 parcel back from 12 by 12 to 9 by 9 with our base 3 modular building kit ok so

what if the parcels got bigger again so we went up to base 16 and this proved to be big enough the buildings could be large enough to convincingly sell a space and there was enough room and the parcel around the building to support cover arrangements as well another thing we noticed is that while our parcel subdivision tech was making deciding the right number of builds to buildings to build of each size easier it looked really weird to have all these unrelated buildings all adjacent in this way every time we loaded a map with the subdivided parcel on it it felt more like a hodgepodge of assets than an arranged space we never did find a way to do this in a way that didn't feel off and we turned it off it did however it get us thinking about the relationship of a building to its surrounding space we had stores coming in with their storefronts facing away from the roads and this didn't make much sense either in order to address these issues we came up with some ideas the first was that of a

parcel facing when a parcel was defined the side that was meant to be adjacent to the road was identified as the face each parcel and a plot had all of the sides adjacent to the roads marked as facings these were represented by yellow arrows in the image when a parcel was chosen to load into a plot it would only come in at the rotation where it's defined facing would match those allowed on the plot the original design was that a parcel could come in at any rotation to maximize variety while the new design was certainly less random than the original design the proper orientation of a building was pretty critical in creating a believable space we had a number of parcels that included parking lots or other vehicle access having these areas ringed in by sidewalk was pretty weird where all of these cars hopping curbs every time they wanted to park driveways existed in the plot not the parcel however so they would have to be spawned in only where they were appropriate so driveways became part of the parcel

definition as well and parcels would only come into a plot in a space that had all the driveways available that it needed as the driveway had to connect to a road in the plot the picture is what a parcel looks like in a plot in our editor excused the programmer art the arrows represent acceptable facings and the smaller squares are where the driveways show up after placing a parcel in a plot you can then toggle these settings per parcel based on the surrounding road networks this technology was delicate however as it was easy to define a parcel so strictly that it rarely if ever showed up if a given parcel had a specific facing as well as multiple driveway requirements there may be only one or two plots that would accommodate it and in those plots it had to win the role against the other parcels that would also fit into that slot there wasn't much point in building a parcel that no one would see so these concepts had to be used carefully in order minim to minimize this typically a

parcel either had a driveway if it had a parking lot or a facing if it didn't but never both we also added markup to each parcel and PCP in the game to provide eligible locations for the exocomp squad to faster up down from the sky Ranger at the beginning of the mission the number of player starts in each map was determined by its size providing a uniform density of start locations for the greatest amount of flexibility and variation another iteration on parcel development was to introduce the concept of zones now that we had a lot of buildings for each plot type they were coming in in strange ways small towns had two components neighborhoods and downtown retail stores it was weird to see a single-family home on the same block as a shopping center these buildings made more sense when they were clustered up homes around other homes shops all in a row by zoning parcels inside of plots we could guarantee this outcome zones facing and driveways

helped immensely with the believer ability of a space buildings now appeared on screen as you would expect them to in real life they were oriented towards the road were grounded with driveways and were surrounded with similar structures however this came at a cost we had significantly reduced the number of parcels that could appear at any given spot and a plot in order to achieve the same amount of variation in each plot we would have to create more parcels of each type zones were paying dividends and directing future future parcel authoring once we had added zones to all of our parcel definitions they generally fell within one of two categories per plot type and we focused on that zone specific plots were created for each plot type for example the image on the left is a small town neighborhood zone plot and the right is a downtown retail zone plot both of these zones were considered small town they shared

pcps and most of their art maps shrunk in size from initial estimates as missions came online and we experimented with how far we wanted the XCOM squad to travel in order to reach their objectives this kind of experimentation was much easier now that we no longer had to involve an artist in order to accomplish it through iteration our parcel sizes were now 16 by 16 16 by 32 and 32 by 32 nice base 8 numbers initial PCPs were sized to the dimensions of their contents however we had six tile wide roads ten tile wide roads and sixteen tile wide roads none of these numbers were compatible and it was impossible to assemble an asymmetrical plot with all of these elements we settled on base 8 and changed the sizes of all the PCPs to match our parcels this greatly eased plot assembly and allowed for significantly increased layout variation reinforcements that

walked in from the side often took a few turns to reach the combat minimizing their impact a troop transport system was developed so that reinforcements could now fly into any at any point on the map which made pacing combat easier troops could land at any of the player starts we had placed in each man because we had placed so many we knew there would always be a valid location nearby for this to happen it had a great side benefit for us as well as it eliminated the need for road networks around the edge of the map large unit support was also cut as we solved their pathing issues by having them travel through and destroy objects instead which made them feel more powerful as well without the need for roads around the edge of the map and around every building we could alter the layout of our plots to have more buildings in fewer roads this was a big improvement to both the visuals and the gameplay for these maps as visual interest and better gameplay was in parcels not road networks the picture on

the left was one of our older plots it had eight parcels and 40 PCPs the plot on the right was shrunk after the rules change it contained only four parcels and twenty pcps with fewer parcels and PCPs loading per plot we were getting more variation out of the same number of maps that we had created as players were only seeing half as many each time they loaded a map I've spent a lot of time so far talking about the development of plot and parcel maps I'd like to briefly shift to mention some of the tools that we developed in order to make development easier asset swapping was another procedural component that we developed during production that ended up saving a lot of our time as well we ended up shipping around 40 persons and around a hundred and thirty pcps for the wilderness plot type which was a temperate forest the original plan was to have a similar suite of maps for a pair of sister plot types the arid wilderness and the tundra

wilderness we have the arid and Tundra meshes and materials created already so we started by creating a set of temperate pcps manually swapping the materials and meshes in each map to the arid and tundra versions then saving them out as separate files this was really time consuming as was the maintenance of these additional hundreds of maps I managed to convince a programmer to write a system that would automatically do these predefined material and mesh wha during load-in game temperate arid and tundra were defined as biomes all maps were constructed as tempered and swapped on-the-fly as the game requested a level of a different biome now we only had to maintain a third of the maps we had originally scheduled for which saved man months of work while the environment artist got out of an art passed on hundreds of additional maps they had to make arid and Tundra versions of all of the cover assets and deco that we had

made for the temperate wilderness that were the same size and shape which was challenging the picture is the original test map that we used to plan out which assets we plan to swap and their equivalent for each biome the assets that we weren't able to find swaps for would live in a biome only deco layer in each map while this required the aforementioned dreaded comprehensive pass through all maps to add there's just no erred replacement for something like icicles you can't swap everything we eventually expanded this system to include map swaps as well we couldn't make our stream PC pset work in arid it looked more like a ditch however one of our environment artists created a beautiful riverbed set that swapped in when a wilderness map with a stream loaded into an arid biome the images on the screen are the same wilderness stream plot in all three biomes as you can see the arid version has a slightly different layout and there are elements

of biome specific deco in all three this system also worked on maps for other plot types now we had arid small towns and Tundra shanty towns for very little extra effort it wasn't part of the original plan but it came free with the technology this asset swap system was further extended to provide visual variation in a number of different ways recreated a number of color variations for each civilian vehicle that the game would randomly choose between unload that kept things looking fresh we were also able to swap out posters and signs and other miscellaneous deco our original method for testing new parcels that we created in game was to pick a plot that they were eligible to appear in and reload it until the parcel that we wanted to test appeared this wasn't terribly efficient as we created more and more parcels the strategy became too low the plot a few times to see if you

got lucky then get impatient enough to edit the parcel definitions locally to comment out all of the other parcels except for the one that you wanted our heroic systems programmer rose to meet this challenge and created tactical quick launch or tql tql could load the plot without launching a tactical game the plot would then load all of the component parcel and PCP maps as you watched on slower computers you could see the individual Maps winkin to existence this gave the people working on the tactical game an appreciation for what was going on under the hood one of the key features of tql was that after the plot had loaded you could select any parcel on the map and change it from a menu to any other parcel that would be eligible to fill in that slot parcel testing just got a lot easier eventually a rudimentary menu was added that would allow you to change a number of

variables before loading the map you could select the time of day the biome the mission you wanted to go on the squad you were playing with how difficult the enemy should be etc once you hit start it would launch the tactical game on the map you had just constructed this was a huge boon to testing and improving the game in general I don't think we could have pulled it off without tql now that we've had a glimpse at the tools I'd like to talk about some of the challenges we ran into later in production on XCOM tube once we had the entire game in place and were able to start doing complete campaign playthroughs we started picking up on an issue with how our tactical missions were playing out being able to go on any mission in any of our five plot types really genericized the spaces no place really had an identity our solution to this issue was to Paris Pacific plot types with specific missions previous to the iteration you could have gone on a guerilla Ops mission on any of

the six plot types now you would only go on guerilla ops in small towns this required rebuilding all of the plots in the game while this was a pretty major reconfiguration it was actually a good thing it allowed us to once again shrink our plots and tailor their size to be mission specific as some missions required more space than others and previously all plots had to be big enough to accommodate the mission that required the most amount of real estate this help with player navigation as it was much easier which was much clearer which direction to go when you started next to I ran up against the edge of the map it also reduced memory footprint loading times and prefer improved performance thanks to our map setup we lost zero art in this conversion it took me a week to reconstruct everything and the game was better off for it if this was done on enemy Unknown it would have required emergency surgery on every tactical map and it couldn't have been

done at that point in production without slipping the ship date being able to address this quickly was a good indication that we had improved our pipeline significantly now that we were doing regular full game playthroughs it came it became apparent that having all tactical missions be random left something missing an enemy unknown we had a series of what we called ten poll missions they advanced the story and keep places and usually involve cinematics we needed something like this for XCOM - but these were big-budget affairs and we were fairly late in production perhaps we can leverage our procedural system to cut down on the amount of work it takes to put one of these together what if we made the map lead-up procedural but the final destination was a static parcel with narrative elements including cinematics and vo this proved to be a good hybrid as the levels came together relatively quickly instead of making an entire map for these custom narrative missions we created a single 32 by 32 parcel for them and poured all

the effort into that one small space they went together rather smoothly the images on the screen are a black site our first Golden Path mission and the sigh gate which was our third apologies if either of these were spoilers when the final ratio of missions was established this put 1/3 of the missions into one of the five plot types people playing the game were reporting that they were seeing a lot of small-town way more than the other plot types what if we added another plot type to the game to split the load of guerrilla ops missions okay but our other plot types typically have three dozen parcels and six dozen PCPs to support them and we don't have enough time to build that many new maps the obvious question that we hadn't considered to that point was well do we really need that many maps to sell a new plot type we only had enough time in the schedule to build half that amount without crunching we decided to

cross our fingers and take a chance than the reduced number of maps being enough slums came in real hot but we got it all done in time to be place ested and shipped with the title as it turns out the more plot types you have the less content you need for each one there was enough slums content to support a single playthrough without getting any repeat content only our most hardcore fans noticed that the slums was a lest robots robust F and the others because we built the plot type with all the knowledge and experience of building the others we were able to take it farther more quickly and ended up being one of the most iconic spaces and was featured in a lot of marketing I was happy to have been wrong on this one okay so now that we've shipped XCOM - how did we do XCOM - shipped with around 80 plots a comparable number to the 80 tactical maps that ship with enemy Unknown however these plots had over 200 parcels and 450 pcps to support them while we

had technically the same number of map layouts between all the content loading procedurally inside of them each layout could provide a dramatically different visual appearance in gameplay experience watching people play and talk about the game tactical maps were generally well regarded and people were playing through the game multiple times without complaining about them which is one of our main goals also we were all playing the game over and over internally and I would occasionally hear from a co-worker about how he had come across some negative content in his third playthrough that he had never seen before and marveled at how much stuff we had made well that's cool but wait a second did we make too much stuff how many play throughs are okay before content is repeated we underestimated how much we could stretch our content and as a result we built more than we needed to we planned on playing through a campaign without seeing any individual building twice with all of the other visual variation going on the location of the

building inside the plot the rotation that came in at and the different objectives associated with it it was okay to see a building more than once sometimes it was cooler to recognize a building then play through it in a different way the takeaway is that future content could involve fewer Maps which was a valuable lesson for expansion planning not long after XCOM 2 was out the door we began planning for an expansion pack a traditional video game expansion pack is typically a discrete set of content that lives alongside the base game campaign what makes a Firaxis expansion pack so exciting is that they often involve revisiting elements of the core game the XCOM 2 expansion was no exception to this as a level designer I've had very few opportunities to alter content posts yet yet it's a fairly common occurrence to look back on old work with the benefit of distance and hindsight and come up with ways to improve upon it in addition no amount of

internal testing and evaluation will substitute for the feedback you get from reading forums and watching twitch post launch one of what I thought was the more valuable and actionable pieces of feedback that bubbled up from all of our post launch playthroughs and from observing the internet was regarding our plot types while the various plot types we had created from XCOM 2 were visually distinct they didn't very much in gameplay this was something that we wanted to address in the expansion but before we could address this we had a bunch of new missions plan that required a bunch of new plot types we now had a much better understanding of how everything fit together and we could use this knowledge take the procedural system farther than we had on the bass game I'd like to walk through a few of these now the idea for the abandoned city spawned from a conversation lamenting the cost of building interiors building interiors are the most expensive custom part of an ex-con map and took the longest time to create however the game encourages

players to travel via the rooftop in order to gain the aiming bonuses against units on the ground and fewer chances of being surprised by the more open lines of sight on the rooftops I postulated that we could address the problem without changing the gameplay mechanics or the art what if we built a city without any ruse obviously the building's themselves had to have roofs but they didn't have to be accessible to the player will make them go inside by going exploring this idea led to the discovery that it was tough to make a believable dense urban environment out of standalone buildings that were all 16 by 16 16 by 32 and 32 by 32 the building's felt like they needed to be larger but the best gameplay coming out of the plot type was the fighting into and out of buildings I wanted to procedural eyes our buildings but all of our visibility and other technical markup required the entire building to exist within a single map file we wrote some tech to allow a single building to

span multiple parcels this way we could have buildings the size and shape that we needed without losing the visual and gameplay variety of the procedural system armed with our previous conclusion that we could make the system out of fewer maps the entire block building system is only 16 pieces there were for each of building endcaps corners straights and T intersections because each plot had a differently sized and shaped block building no two buildings feel the same tunnels were another attempt to extend the game in a different direction tunnels is an entirely interior plot type with limited sight lines and negative space the visual focus of the tunnels is on the PCPs and not the parcels the parcels are mostly generic mechanical spaces and the art love went into the expansive subway tunnels cathedral-like sewers because they are procedural they share a lot of filler content which save quite a bit on our time the sewer tunnels are wide and open

and have a pit running down the center which cannot be entered but is regularly crossed by bridges which frequently became trip points this produces an interesting difference in combat from the limited sight lines and verticality of the subway tunnels chosen strongholds respond by the idea of creating a procedural dungeon crawl in XCOM continuing with the idea of creating as few maps as possible to support this how many unique 16 by 16 rooms do we need to support three dungeon runs of seven room dungeons after a lot of experimentation we ended up creating 15 rooms because the majority of the combat occurred between two rooms randomizing the order of the rooms appeared had a profound effect upon the gameplay this spawned an interesting conversation with the environment artist I was paired with on this plot type he was upset that the layout of the base made no logical sense he was expecting the player to start at the entrance to the base and fight to

the rooms in the order that a guy building a functioning base would place them in we could have built that dungeon procedurally but I expect we would have had to have make 3 times the number of rooms in order to have them in a sensible order and still get the gameplay variation we were looking for it raises an interesting question how much do these phases need to make sense I pointed out all the other shortcuts and liberties we took on the other plot types for the sake of gameplay has anyone noticed that there isn't a single bathroom in all of XCOM personally I think there's a lot of leeway available in the in-game representation of reality that you have before you actually break immersion for the player and no one playing the game will pay as much attention to it as the people who are working on the xena form biome was originally conceived as its own plot type similar to the idea of having arid and tundra versions of the wilderness just like the other two the procedural solution to implementing this was to

create a fourth swap type for wilderness assets we made a fourth set of rocks trees ground deco and terrain materials that simulated alien flora that had been released into the terrestrial wild corrupting it we created new times of day with unusual colors and effects that would only load with the new biome type furthering the visual difference this swap could be applied to hundreds of maps that we had created for the base game in this way the procedural system opened up new avenues of development for an expansion as we could alter base game content without editing any of it the two screenshots are of the same base game wilderness lake parcel the left is the temperate version as it was constructed in the editor the right is after going a swamp to this uniform biome fairly dramatic visual departure one of our goals in the expansion was to add gameplay variation to the base game plot types to match their distinct visual variation to solve this we added

a unique PCP set to each plot type to help change the way they played there was a focus on adding verticality and negative space to plots with each plot type getting its own unique take on the theme wilderness and Cheney got coastal cliffs city center got canals slums got washes retaining walls and modular overpasses and small town got ravines in an order in order to incorporate these new items of their appearance would be balanced I had to redo all 80 plots in the game it only took a couple of weeks and its benefits were twofold the plots now centered around the gameplay defining PCP sets for that plot type creating the gameplay variety we were looking for plots could now be shrunk even further due to Mitter mission iteration that took place after the original plots were constructed the new plots were as tight as we could get them and they were playing better than ever we also wanted to coax players back into buildings something we had addressed in the abandoned city but needed attention every needed attention

everywhere else building interiors were interfering with our concealment system jumping through windows an ex-con soldiers preferred method of building entry would break concealment a lot of new our new missions were on timers and players had to spend extra turns moving to an opening doors in order to make a concealed squad go through the interior we gain the ability to set the starting state of doors and took a pass to the base game buildings setting a bunch of them to start open this significantly improved the efficiency of concealed movement and building interiors now that you've heard where we've taken our procedural system it's time to take a step back and look at the advantages and disadvantages of switching to using this system I think the talk thus far has demonstrated the specifics of most of the advantages it's also worth noting the benefit that the procedural system had for modders the long war two guys our flagship total conversion mod were

able to rearrange quite a few things to suit their own ends without creating very many new parcels or PCPs they mostly focused on creating new plat layouts to suit the mission types they had created it's worth going through some of the disadvantages as well the environment artists almost all preferred making static Maps they felt like they lost a lot of control over the appearance of a seen through the procedural system and they were right for access prioritizes gameplay over visuals but this is not true from any studios and it may be a hard sell for a more artist centric environment cinematics were more difficult to create in a procedural environment a fun conversation to have with a cinematic artist is to tell them that the framing and background other shop will change every time they watch it we also never developed the technology to write out a procedurally generated map as a static map so cinematic artists often had no idea how much space they had to work with it was hard to fit a lot of visual storytelling into the smaller map chunks

this was a staple of the enemy Unknown Maps and we lost it to some degree with the switch this is not necessarily a disadvantage of going procedural it was just one that we ended up with as a result of choosing to use smaller map sizes the initial idea behind taking XCOM procedural was it's better because it's different every time for access has a long history of procedural content development with the civilization series but that doesn't necessarily translate into every genre outside of scenarios civilization doesn't even have level designers if I had to approach a procedural level design system again I would be more careful with what I was willing to leave to chance there's a lot to be said for handcrafted content and there's something to be lost for abandoning it entirely the question that comes out of it is how procedural does your game need to be there's a sliding scale between replayability and character more a smaller match provide

the most variety but can lack soul fewer larger Maps allows for more visual storytelling at the expense of replayability where do you want to land on the spectrum that's all I got so it looks like we have a few minutes for questions if anyone has one wants to come up to the mic hi I'm wondering of the admission contents of the replacement where the objective are was working with using only procedural generated map was it metadata in the platter first of themself or there was something that the level designer was adding afterwards all of the mission specific metadata was added to all of the parcels and PCPs each map new what's missions it was allowed for and that was added to the definition and so that was covering all all the potential variation of valid

mission in in in the map okay there was a lot of markup perfect banks were here how did you deal with more global issues such as lighting and performance well one of the downsides of switching to the procedural system was that in enemy Unknown we were using pre-rendered lighting that was baked out into the maps but that was no longer an option with the procedural system so we switched to a deferred system which you know have the lighting rendering in real time and that didn't look quite as good and it was more expensive but the variety that we got from being able to load any map with a different time of day I think more than made up for it over here did you support tools for say I'm a level designer and for this one story mission I want guys to spawn this way and run around the right side of this building did you support that kind of specific scripting you mean for

enemies yeah for enemies or allies we didn't do a lot of direct control over enemies via scripting most of that was handled by the system we could call for additional enemies but the game decided on the on-the-fly where it wanted to put them based on them where the player was gotcha thanks all right well I think we're out of questions thanks for coming you you