A game I made in which zombies have to form up by the door to run past a turret, most being mowed down in the process.

I remember my father once telling me sometime before the age of ten that I’d regret all the time I was spending on video games, because years down the road when my son asked me what it was like for me growing up all I’d be able to tell him is that I spent it as a blur in front of my TV. Admittedly, there was some truth to this. I wanted to be a game developer when I grew up, on the basis that I knew what made a fun video game and a lot of the people making them didn’t. My enthusiasm for video games has chilled quite a bit since then, and probably the largest factor in that was the rise and fall of Halo 3 multiplayer. Halo is what got me into online multiplayer, and ultimately Halo is what got me out of it.

It let me live the dream of being a game developer, without ever touching a single line of code.


Me killing three people with one burst from a Spartan Laser

The first striking feature of Halo’s multiplayer is bright, vibrant colors. In a gaming landscape dominated at its release by dull grays and browns Halo stands out simply by deviating from palettes that include Crimson as their only primary color. The shiny, sci-fi almost-cartoony aesthetic of Halo paves way for the necessary suspension of disbelief at gameplay aspects like personal protective forcefields, ‘energy swords’, ludicrously high jump capabilities, and halo’s version of the classic arena shooter ploy of firing your rocket launcher at the floor to get air on your jump: grenade jumping.

From a gameplay perspective, Halo combines the best aspects of arena shooters with their modern military counterparts such as Battlefield. Small quick paced maps, large vehicle brawls, and everything inbetween are represented, part of what keeps you coming back to halo is the diversity of gameplay on offer. Halo could be realistically marketed as a game for four to sixteen players. It can handle as few as four and easily provides structured entertainment for sixteen. Barring a LAN party involving multiple xbox 360’s and a good deal of in person finangling to get together enough people, the only way to fully experience Halo 3’s multiplayer is through online play.

Paying sixty dollars annual for the privilege, Xbox Live’s answer to organizing tedious LAN parties is the matchmaking system. It’s pretty simple, you alone or with up to fifteen other people comprise a party, this party can enter various playlists, which are sets of maps and gametypes, a map being the environment you play the game in and a gametype being the rules of the game. Certain playlists have party size restrictions, which means that even though technically your party is allowed to be sixteen players, this will exclude you from almost every playlist. More realistic party sizes for matchmaking are one to eight people.

A party has one party leader, who chooses the playlist and has the ability to kick people out of the party. (Referred to as ‘booting’ in game.) Some examples of Matchmaking Playlists:

  • Lone Wolves - Ranked free for all matches, with a party size limit of one.

  • Team Slayer - Team versions of the ‘slayer’ gametype, where you kill other players for points.

  • Team SWAT - Team games where players start with headshot capable weapons and no shields, meaning that you’re always vulnerable to a headshot.

  • Big Team Battle - An unranked playlist consisting of games oriented for large sixteen player parties.

At the end of a game, you can press X to ‘party up’, letting you retain players from one game to the next. For most people who played Halo this was probably about the extent of their experience. They’d go into a matchmaking playlist alone or with friends and make new ones.

It was also the breeding ground for early custom game parties.

Custom Games

My first introduction to custom games was from browsing youtube videos about Halo 2’s online multiplayer. One I remember particularly well was a comedic take on the popular ‘Infection’ game type variant of slayer, where instead of caring about points at all the game ran on honor rules. Everybody except one person was on the red or blue team. The one person is on the green team, making them the ‘zombie’. Their job is to go around with an energy sword killing other players to convert them into ‘zombies’. There was no actual enforcement mechanism for this, players killed by a zombie would need to change teams manually. Anybody killed by a zombie has to change to the green team. The game ends once everybody has switched teams to green. This was one of many such variants that ran on honor rules such as ‘Lava Man’, and one of even more variants that were popular but not necessarily honor rule reliant like ‘Tower of Power’.

In Halo 2 custom games were limited to changing aspects of the gametype. (Ruleset) The core map would be largely the same in terms of scenery, placement of objects such as crates, etc. The major changes that could be made to the environment through game types were the removal of weapons and grenades. One major addition in Halo 3 was the introduction of Forge, which let you edit maps to tailor them towards a specific gametype. We’ll cover Forge in more detail later, but if you’re completely unfamiliar with the subject this video (The discussion of forge starts about four minutes in.) should be illustrative.

The summer leading up to the release of Halo 3 was frustrating, I was eleven years old and had only gotten to experience Halo’s online multiplayer vicariously through youtube and trailers, finally getting into the action at the tail end of Halo 2’s prominence as the console shooter. I was thirsty for more and at the place we’d just moved the Internet was horrifically slow so online play wasn’t really an option. Moreover it was a lull in the video game release cycle, at least from my perspective. This was before I’d realized that there existed websites dedicated to video games to keep me in the loop. I spent weeks on end with bated breath reading and watching everything I could on Halo 3.

When I got the game, my first instinct was to blast through the campaign on easy, in one long non-stop marathon gaming session. It would take months until I beat the game on the hardest difficulty, and it wouldn’t be years until I used the hardest difficulty as my default. The first time I booted up forge I did what I can only assume everybody else did the first time: I made a giant pile of explosives, and then blew it up. To make a pile of explosives in Halo 2 one was forced to scrounge around the map and slowly corral them together without accidentally blowing any up, and even then what you could get was a pittance compared to the abundance on offer in forge.

Bungie shipped nine different gametypes with Halo 3. (Not counting variants.) I could take the time to list them all and how they work, but in this case there’s no need. Generally in any given custom game party, people were playing two things:

Slayer, and Infection.

In case you missed it the first time, slayer is pretty simple to explain. You kill other players, and earn points for it. In team slayer you kill the other team, and earn points for it. It’s almost disingenuous for me to list slayer alongside Infection as ‘the only thing people played’, because that implies people played slayer and infection at something approximating a 1:1 ratio.

Part of the reason they didn’t was an anti-feature of the custom games approach to team games. Teams would be chosen by each individual player deciding what team they would be on. The party leader had no say, and there was no good way to coordinate who goes on what team, espicially in a large party of say sixteen people. This meant that even when people were trying to balance the teams in good faith they would almost inevitably end up with too many people on one side, and then have too many people try to correct for the imbalance resulting in too many people on the other side. How it needed to work was the party leader either setting the teams, or the party leader being able to flip a switch which would freeze team switching so that he could set an equilibrium and start the game. Instead this single issue largely kept me from taking advantage of the palette of gametypes on offer. Admittedly if I had been willing to boot anybody that changed teams and demanded they only change teams when I tell them to I might have been able to keep enough order to make them work, but that was a lot of effort. Infection was easy because the rules were all part of the game type, I could load the map and pick the game and it’d work with no interruption.

In reality, Infection was the custom games gametype. Bungie was so taken with the popularity of Infection in Halo 2 that they added it as an official gametype in Halo 3. This single descision is a large part of what led forge games to evolve beyond Bungie’s vision.

Besides the actual draw of the game itself, Infection as a gametype offered certain features that made it very useful to repurpose for creating ‘minigames’:

  • Asymmetric team traits. (Eg. Different shields, different starting weapons.)

  • The “You die you become one of them” mechanism.

  • Fine grained control of spawning (where people show up when the game begins and after they die) for both teams.

  • Asymmetric traits between the people who start as zombies and the zombies infected by them.

The custom games that went above and beyond, turning Halo into completely other things made extensive use of these aspects of the Infection gametype. (As an example, this game for three people would be functionally impossible to create without the use of these features.) What gametypes really needed was a scripting language so that you could create your own and not have to try and shoehorn Infection into so many things it wasn’t intended to be. Interestingly enough, in Bungie’s sequel to Halo 3, Halo Reach the dev team created such a language for internal use called Megalo.

Initially there were no matchmaking playlists that offered Infection, the only way to satisfy the demand for it was custom games. I personally had a great interest in playing Infection from all those youtube videos I’d watched. So right away there was a demand both from the community and myself for Infection. The only problem was that most of the games were mediocre at best, outright frustrating or boring at worst. In the very early days it was common for maps to follow the pattern of a ‘big base’ for the humans, which they must defend from the zombies. The base would of course be stocked with an obscene amount of arms and ammunition, so there would be no worry about running out. Games would be balanced so far in favor of the defending humans that the challenge was absent, leaving bored repetitive killing for the defenders and bored frustration for the attacking zombies.

Better maps came along.

The first times I played Infection there was this sense of anxiety, a sort of shiver or chill that added a lot of atmosphere to the gameplay that came along with the entire concept of your buddy being there one moment and the next him being gone, with being snuck up on. It faded with time, but I can still feel it a little when I make an effort to remember. The games that caused this feeling were the ones where there was ‘real’ danger, a sense of inevitable helplessness. Games that got infection right.

After playing so many mediocre games, I decided I could do better.

The first time I met Patrick I was twelve years old. I’d been playing matchmaking and every so often during a game I’d recieve strange notifications that people I’d never met had invited me to play a game of halo 3. The first few times I accepted one I’d find myself playing something that was quite strange compared to the standard fare of halo. Whether it was running around a map full of explosives or a gametype in which everybody kills in a single shot, I’d find myself simultaneously intrigued, disgusted, frustrated or bored. The games were almost always terrible, but the potential was obviously there. Most of the time the issue was that the person who made the game had no sense of balance, or no sense of taste, or some other mitigating factor that was stopping them from creating something good.

I don’t remember if I met Patrick in matchmaking or in a Custom Game, but we were both fascinated by ‘customs’, and I lamented to him how I had all these interesting ideas, but I wouldn’t have anybody to play or test them with. He knew all these people that hosted custom game parties, and I started getting invited to a lot more of them, building up a network of other people who were also interested. Patrick was around my age, with the maturity level to match. Together we discovered this method of getting a large custom game party together that only required you to have about three or four people. How it worked was you and your friends would join somebody elses game and demand to be made the party leader. The idea was that if they refused you’d all quit, leaving the party at half its new size. Out of loss aversion I would usually end up party leader. I only had to resort to such extortionate tactics at the very beginning, soon my parties were gaining a following of their own and I could just draw from my stock of friends to get a game going. Having this network was important.

There were a lot of ‘missing features’ in Custom Games and Forge, stuff that you really needed in the game but just wasn’t there. One of these was the ability to let people other than the party leader set the game and map. The only way to let somebody in the party play their map was to make them party leader. This came with the risk of them not giving it back, or booting everybody so that you’re forced to scramble to rebuild the party.

Probably the biggest missing feature was the ability to search through active custom games in the same way that one could choose a matchmaking playlist to find others for a ranked match. The largest challenge in providing this would be figuring out how to let people filter through custom games in a meaningful way. The obvious method would be to let you search by gametype, but since the sort of game you can get out of a category like ‘Infection’ is so wide, this isn’t really good enough to let you find what you want. One way to do it might be to let you search for particular maps and game variants. The biggest immediate problem with this is that two map or game variants are allowed to have the same name. Confirming that two files are the same file is not a terribly hard problem. 1 In Reach Bungie didn’t add matchmaking but improved the file sharing solution so that you could search for maps and game variants through various criteria.

The lamest aspect of Custom Games from the hosts perspective was the arbitrary hundred item custom content limit. This limit wasn’t even one hundred items per category. (eg. 100 maps, 100 game variants, 100 saved films.) It was 100 custom items in all categories total. So if you saved a film clip from the games internal theater system (the other subject covered in the video I linked earlier) that would count towards your total. What this meant in practice is that eventually if you spent enough time making games you’d hit a wall where you couldn’t make any more without deleting old ones. Further it would lead to a situation where you wouldn’t know if you had a hundred items or not before beginning a forge game, and if you already did you couldn’t save a new map, your only options were to save over your blank slate variant of the map or abandon your work. This disincentivized the creation of new maps while you still had a large amount of existing ones.


A map I started, spent a lot of time on, but nothing ever came out of. Note the extensive interlocking.

Bungies original conception of forge was as a tool to enhance the core multiplayer experience: Shooting people and blowing stuff up. Their vision is plainly illustrated in a pre-release video I linked earlier where the dev team talks about moving the flag in Capture the Flag over a meter. More ambitious uses are hinted at, reworking entire maps so that the reworked versions become used as Bungies official versions, creating tall buildings for ones amusement. Even having the foresight to know that their players would surprise them, Bungie didn’t expect what actually came out of Halo forge.

The first hints of what were to come were maps like TrueDarkFusions “Hobo Heights” (A word of warning, the Wayback Machine mangled the formatting a bit.) in which a group of humans tried to avoid being killed by a zombie with a laser.

My friends would ask me “How do you sit for eight hours and make a map? I get bored after a few minutes.” One significant factor is having already made a successful map. When you know you’re capable of producing something good, it motivates you by allowing yourself to imagine how cool it will be when you’re finished. The social aspect of craftsmanship in this case cannot be understated. ‘Thirsty for more’ describes the feeling well, when you’re not so much willing yourself to make a new game as you are compelled. When you’ve reached the point where you feel this compulsion towards something you can spend a good deal of time at it without getting bored.

Custom Games and Forge had a sort of complimentary rhythm. During the day you could invite your friends (and their friends) to play quirky, fun and funny, intense matches, then slowly segue into the solitary workshop of forge at night. In this way it became a sort of lifestyle, the natural die off of parties as people slinked away to bed gave natural pauses for forge, which itself was a calming, focusing activity that let one reflect. Just as you find yourself tiring of death and explosions, forge is waiting for you at the end of the day with a cool glass of solipsism. And when you’ve started to really wish you had other people around to test this map you’re making they slowly start to come online.

Exploration was a huge part of Forge. In Halo 2 maps didn’t have barriers preventing you from leaving them. This meant that if you could find a way up (eg. The infamous ‘superjump’) you could explore the exteriors as much as you wanted. In Halo 3 with the addition of flying camera mode in theater and monitor mode in Forge Bungie could no longer make the assumption that players couldn’t fly. They built invisible barriers to stop people from just waltzing out of the map like they did in Halo 2. In spite of their efforts it was still possible to break out of maps using Forge tricks. This meant that you could spend a significant amount of time trying to get past an invisible barrier to explore an area outside the map. Some of these areas actually made interesting environments for custom games. I remember a lot of my early maps being centered around these locations, in large part because I was fascinated with them.

The other thing to look for was new tricks and glitches. At the beginning there were a lot of maps centered around the concept of the ‘sky castle’, which was basically a big base style of map where the base was in the sky. I remember sitting in a forge match with somebody and asking them how you put things into the sky like that. They told me it was simple: If you hold a teleporter in the air and then save the game, teleporters don’t move so if you exit the game and come back it will be fixed where you put it. This let you put objects in the air by holding them up with teleporters. As it turns out, there are other ways. Forge glitches and Infection are two of the components that let Custom Games get really good. The sort of maps you could make with glitches versus without was like night and day.

One common place to find inspiration for Forge maps was in exploring Halo’s physics engine. As an example, on one of the last maps Bungie released, Longshore, they included an orange dinghy that could float on water. This was exciting for the ten seconds or so it took before the boat begins to sink. Fans had been asking Bungie to include boats for ages, and it felt like a tease. As it turned out however, if you shot the boat with a weapon it would reset the timer before it sinks. This meant that you could take an automatic weapon such as an SMG and shoot it at the boats engine to propel it forward, ‘driving it’. (In fairness, I didn’t actually discover this on my own. It was pointed out on the wikia site for halo.) It was while playing with this mechanic in Forge that I realized you could make a map out of it, because ‘driving’ the boat was so difficult that it provided challenge all on its own. The seed of this idea developed into a map where humans spawn on the second floor of a dock overlooking the water while zombies try to ‘drive’ the boats to the landing point of the dock and kill the humans. It was a game where the zombies almost never win but managed to be really fun anyway. I named the game variant ‘landing party’ after a term for lander ships I’d heard playing the demo to a flight simulator as a really small kid, and I named the map variant ‘Sharkey’s End’ as a subtle nod to Lord of the Rings, which I’d just finished reading.

This sort of whimsical inspiration was Forge.

People familiar with PC game modding might react to forge with “That’s lame, I can do all that and more in Gmod.” They’re not wrong, but part of the magic was that you couldn’t just reskin the game to be Mass Effect or change all the character models. Because all forge games had to be fundamentally built out of the components available in the Halo environment they all felt like Halo, even when they were far removed from standard Halo gameplay. Another aspect of Forge is the simplicity. Forge is powerful enough to make interesting games but still simple enough to capture the interest of somebody who is twelve years old.

Powerful options for gametypes were only half of the equation that let Halo 3 stay fresh long past its shelf life. After Halo 3’s release Bungie put out Map Packs that would give you new environments to play and forge on. The first big game changer in Forge came when Bungie released the first map pack. One of the maps, Foundry was a complete departure from the maps that shipped with the game for a few reasons:

  1. Unlike the shipped maps, Foundry let you move and place immovable objects.

  2. Foundry was a giant square flat space, which meant that you no longer had to battle hills or worry as much about running out of room in the few flat spaces you have.

  3. Foundry included new items like bullet-blocking energy shields and ‘mancannons’ that shoot a player across the map when they walk into them.

  4. Foundry gave you lots of blocks to build a map out of.

Foundry let people do things that were previously impossible. Where forge art had previously been making drawings out of weapons, suddenly people were coming out with jaw dropping 3D models. Minigames, which had been sparse in the original generation of maps were now much more plentiful. The addition of Foundry is what started really taking Forge beyond the core gameplay of Slayer and Infection. But even then there were still problems, you could only use blocks of the sizes given, and there was no way to shorten a block or only use half of one to fill a gap.

Remember those Forge glitches I talked about? It turns out all the tools people needed were already in the game, they just had to be discovered. The first big discovery was interlocking. This pretty much fixed most of the “Only able to use blocks of a certain size” problem, by letting you place blocks together so that they overlap. It also let you remove the seams between items by interlocking the edges, improving even the look of maps. If interlocking was the only thing people discovered it’d probably have been enough, minus some problems when you ran out of space in Foundry and really wished you could put an item into the wall.

Geomerging of course, let you do just that. By using an immovable object to push another immovable object in the two or three seconds before it becomes immovable, you could stuff it into the ground or into the wall. This was very tedious, but it worked and let people start making maps where items were in the ground, in each other, and totally seamless in build. Still, the problem with Geomerging is how much work it took. The holy grail would be the ability to just make it so items transparently merge into walls and other boxes when you pick them up, and a lot of people got onto Bungies forums to beg for this in the sequel.

That was actually the last big Forge glitch to be discovered. Once people did geomerging was largely deprecated. People still used interlocking because it was strictly easier to set up than a ‘ghost merge’. There were other map packs too which added more, including Sandbox. Sandbox was essentially Foundry on steroids, including a Foundry-like interior basement along with an open desert middle floor, and a ‘skybox’ on top that let you build floating maps with a killbarrier if you fell off. You could also use teleporters to move between them. This was the complete Forge workshop.

People did such a good job squeezing more out of Halo 3 Forge that when Bungie went ahead and did the Forge for Reach they just formalized what people were already doing in Halo 3. The real upgrades beyond that were the ability to move items by discerete units, the ability to slowly move an object by clicking the trigger, equipment such as jetpacks, more objects and bigger budget, little things like being able to make a Warthog the hill in King of the Hill. Reach was probably on the whole better than Halo 3’s forge, but it did miss one major component. The little moveable objects such as crates and concrete barriers and pallets that were available in the earliest shipped maps were slowly phased out in each map pack in favor of adding more immovable objects. This meant that a lot of the interesting physics aspects of Halo used in the early maps were no longer possible in later ones.

While writing this I booted up Halo 3 to play with forge again. Even with the knowledge that anything I build will only be seen by me it’s still fascinating. On a flat surface I laid down 32 (The map limit) explosive barrels and set them to instantly respawn when destroyed. This means that when you throw a grenade into the cluster it will explode, and then keep exploding as new barrels respawn in their old positions. In 1943 computer pioneer Alan Turing proposed the following problem as an analogy for Nuclear Fission:

“Upon Turing’s return from the United States in March of 1943, he discussed “a problem about bags of gunpowder at the points in a plane with integer coordinates. Given the probability that the explosion of one bag will cause adjacent ones to explode, what is the probability that the explosion will extend to infinity?”2

While not a direct 1:1 analogy (Of a problem that is itself an analogy.) if you lay out a flat grid of explosive barrels in Halo 3’s physics engine, and then blow them up the explosion eventually stops, implying that the probability of it never ending is low. Interestingly enough the length of the reaction does appear to be probabilistic.

I guess I’m not quite into the same things now that I was at the age of twelve.

The Decline

There was no one thing that killed Halo 3, it was a death of a thousand cuts.

If I had to name a single thing, it would be the Call of Duty franchise. One of the things I didn’t realize until later was that I would make friends in Halo 3, and then those friends would quit the game and go play COD. So I’d have large parts of my friends list that weren’t actually active Halo 3 players. This wouldn’t be a big deal except that like custom content, the Xbox Live friends list had a 100 person limit. This meant that from my perspective the number of people I could get together for a Halo 3 custom game shrank and shrank and shrank. In reality what I needed to do was just remove people who didn’t actively play Halo and replace them with new people who did.

Regardless though, Call of Duty consistently took out a chunk of Halo 3’s player base with every new release. Of course to say all the drop off was caused by Call of Duty would be disingenous, the game that took the most players away from Halo 3 was Halo: Reach. A lot of people just got bored and stopped playing. No matter how many accelerators you can point to, the basic fact is that Halo 3 died in large part by getting old.

One major interruption to the Custom game ecosystem was the introduction of Xbox Live Party chat. To eliminate confusion, the term ‘party chat’ has nothing to do with Halo 3 parties, as this is a generic xbox live feature. The idea is that you have a communication channel that lets people all connect as long as they’re online. Intended to allow friends to communicate across games, Party Chat filled a major gap by no longer isolating people who wanted to go play something else while their friends are in matchmaking. The problem was that party chat was such an attractive option that most serious players had friends who would insist they use it. Only eight people could be in a party chat at the same time, without even subtracting spots for the people playing other games you’re trying to include in the first place, only half of a full custom games lobby can be in a party chat at the same time.

What this meant in practice is that if you went into matchmaking almost all the players would be in party chat, which made it much harder to make friends you could play customs with. In a custom game a lot of the players would be using party chat, which meant you couldn’t cooperate with them if you were in ‘game’ chat. And what that amounted to was a general social breakdown, where before it was hard to play a game with honor rules, now it would be all but impossible.

I joined back up with Xbox Live in 2013 briefly to play Halo 3 again. The general atmosphere was one of an apocalyptic aftermath. Custom games were haphazard and scattered, rarely having sixteen people like they would in Halo 3’s heyday. The regular gamut of maps I was used to were absent, instead strange creations (some of them quite good) filled their place. Almost nobody had mics, as they had probably bought the game for cheap and could not afford one. Players who did have mics were almost all in party chat, I particularly remember joining one custom game where the only person with a ‘working’ mic had some sort of static going through it that added an eerie feel to an otherwise silent room of people. The friendly jubilent atmosphere I was used to had been replaced with silence.

A caricature in place of a masterpiece.

Notes and Bibliography

  1. In particular, you could have the party leaders maps hashed when they register with the Custom Games server, which would then let people search for maps that they already have, to find people who have similar tastes in games. Another option would be letting you search by what maps are popular. This also gives you part of the infrastructure for a recommendations system. 

  2. Dyson, George. Turing’s Cathedral. New York: Vintage Books, 2012. Print. Page 257.