The thing that makes the Left 4 Dead series so special isn’t the zombies or the shooting, it’s the AI Director. Originally created by Michael Booth (now at Bad Robot) and later expanded at Valve, it translated the tense mix of intense action and quiet anticipation from the best zombie movies into video games. It didn’t create that tension by throwing endless waves of zombies at you, but by holding back, giving players those creepy, anxiety-ridden moments that bring truth to the cliché–it’s quiet, too quiet.
When I was at Valve working on Left 4 Dead, we knew that the Director would create game experiences that were different every single time. After all, players were used to action and down times being tied to triggers in specific areas of the map, and we didn’t do that. They could happen anywhere at any time and it changed dramatically from session to session. Players immediately picked up on this and they loved it.
The Director created another effect that we didn’t anticipate. It was less noticeable but more profound. The down time was new for a multiplayer game. It transformed the game into a place for players to hang out, to chat, and to be social. And just like people do in high-intensity situations in real life, players replaced the tension with camaraderie and jokes. Because the game was flexible enough to accommodate a broad range of skill levels and play styles, more people were able to play together and share those moments. What started as “the game where players went to shoot zombies” became a place you could hang out with your friends.
This basic idea—that games could challenge you and simultaneously be a fun place to hang with your friends—has shaped most of our design decisions as we built The Anacrusis. The first code we wrote for the game is what creates that cadence, the code that spawns all enemies. It was our take on Left 4 Dead’s AI Director.
We also knew that knocking off a ten-year-old game wouldn’t fly with players. We needed to do more. Under the hood, the Left 4 Dead Director works due to a combination of careful tuning and random chance. The chaos that is Left 4 Dead comes from getting lucky (or unlucky) rolls at the right time to spawn specials, hordes, or… nothing at all. It knew next to nothing about the player or their experiences.
The Brute is capable of dealing a ton of DPS with his short range attacks, but his charge will scatter a team that doesn't scramble out of the way in time.
But it worked. The very best moments in Left 4 Dead were ones where the Director pushed the players right to the edge of what they could handle, then it was up to them to either fail spectacularly or triumph against all odds. But those moments happened because of chaos and randomness in L4D’s Director. They may have been infrequent, but those relatively simple rules, combined with random chance brought a great deal of complexity (and fun) to the game.
To start building our game, we needed something people could play. In the beginning our Director randomly deployed aliens on the map. When we first started playing the game this way, we killed pretty much every play tester we put in the game for months. At one point, testers couldn’t even finish the first level. (To our friends and family who suffered through these playtests, sorry about that! Your suffering helped make the game better!) For a game that wasn’t supposed to be about being difficult, this felt really, really bad.
But we learned from these early steps and iterated on the game. We learned that the most fun games weren’t the ones that were the most difficult, they were the ones that felt the most intense. We knew the traditional ways to make a game difficult — you increase enemy HP, reduce player resources, lessen player health, increase enemy damage, spawn more enemies, or some combination of all that. But we wanted to control difficulty in a new way and we wanted to decouple that from intensity. And intense games are what we wanted our Director to create.
We realized that almost every aspect of the game contributes to the intensity—the number of enemies you have to fight, the feeling of being surrounded, the pressure that comes from being low on ammo or health, the soundscape, the chaos of grenades exploding and fires burning on your screen, how many players are incapacitated, and a whole lot more. But, we needed to figure out how to deliver those high-intensity moments across the entire range of player skill levels.
Goopers slow teams that are moving too fast by incapacitating players and covering the ground with sticky goo.
So naturally, we made the game even harder. This let us understand how the enemies and specials we had already built affected both difficulty and the intensity. This testing also helped us learn the difference between fun intensity and frustrating difficulty. For example, just being mobbed by massive hordes wasn’t fun.
We added backspawns–hordes and specials that would spawn behind the player, on the already-cleared path. Backspawns are ass-kickers for new and lower-skill players. While most teams could handily dispatch attacks from the front, only the most cohesive teams could take the added pressure of enemies approaching from behind. But the backspawns added great intensity to players who were on their game.
In order to know when players could cope with backspawns, our Director needed to understand the direction players move through the map. So we built a tool that scans each map and figures out the golden path that players will take from the start to the end of the level. That lets our Director know which direction players are moving and which direction was safe to spawn aliens.
But we still couldn’t tell which players would be able to handle the more challenging spawns. Our Director needed to distinguish good players from bad players. What is a good player? What is a bad player. The engineer architecting our Director, Amy Ackermann, started collecting data, experimenting, and iterating. The first thought was that if you do more damage to enemies and do lots of headshots, you were probably a good player. It turns out, it’s more complicated than that. A headshot god who takes a ton of damage because they charge headfirst into every horde was a liability compared to a player who played smart, conserved ammo, and utilized choke points and grenades well. We kept tracking new metrics and testing until our Director had a reliable understanding of player performance.
So now our Director knew if players were good or bad, and it knew which direction they were supposed to be moving across the map. Without lowering the intensity we’d established earlier, we made the game more manageable for lower-skill players, but we still needed more ways to increase the challenge for higher-skill players.
We also knew that the longer players stayed in a level, the more difficult the game got for them. Playing longer meant they fought more, which eroded their resources. Eventually the attrition of never-ending hordes of aliens would overcome everyone. This held true for our early prototypes, even with our Director dropping health and ammo as it was needed.
Experienced players instinctively understood this, they found the golden path and stuck to it. The more they played, the more they stuck to that path. We needed something to lure players off the path. We tried building special enemies that encouraged players to make bad decisions—the Spawner does that by hiding off the path and spawning an endless stream of turrets designed to annoy the player into leaving their friends to shut down the source of the turrets. But in our tests, the Spawner didn’t slow players as much as we wanted. So we added more grenades and started hiding them off to the sides of the map, away from the golden path.
This helped, but it didn’t get players excited in the way that we hoped it would. Experienced players would break away from the path when they were out of grenades, but they stuck to it when they were flush. We needed more, so we built a bunch of perks and had our Director spread them around the level randomly. We knew the perks needed to change the way you play the game, marginal buffs like “Your gun does 3% more damage” wouldn’t cut it. Our first batch of perks turned grenades into area-effect heals, enemies you land a headshot on into exploding corpses, and turned semi-auto weapons into full-auto killing machines. Many of the perks synergized with each other, so by the end of an episode, players experienced a power boost that was closer to the skill tree in a more traditional game’s campaign.
Playtesters loved the perks.
Some players actually loved them a little too much. We wanted players to find perks in the world, talk through the options available, and make sure that each perk went to the player that would benefit the most. Instead, some players would run out in front of the group, gobbling up every single perk they found, leaving nothing for the players trailing behind. These loot goblins had a great time, but it had the opposite effect than we wanted. Instead of slowing down high skill players, it actually encouraged them to race each other, moving faster through the levels!
We found the solution in the Matter Compiler. Matter Compilers, or MCs for short, are perk dispensers. A player finds an MC, then each player walks up to it. Each player hits a button and is presented with a handful of different perks to choose from. Each player picks the perk that they want, then moves on down the path. The Director places the MCs on the golden path for teams that are struggling and off to the side for teams it wants to slow down a bit.
Matter compilers distribute perks to players as they progress through the level.
The Matter Compilers, along with the perks they dispense, add more excitement to every game. Every time a player finds a MC, they’re excited to share it with their teammates. We found a new way to make the game more intense and exciting as well as another way for our Director to increase the difficulty without resorting to the traditional means.
Now, our Director had access to information far beyond what we’d been able to expose in Left 4 Dead. It knew how players were doing, what their relative skill level was, and whether they were moving quickly or slowly. It had access to tools that let it increase and decrease both difficulty and intensity independently of each other. Our Director was matching the game’s difficulty to the player’s skill reliably in both internal playtests with people who had spent hundreds of hours playing the game and in external playtests with brand new players.
No matter who was playing, they could still wipe and they could still lose, but our Director was able to push groups right up to the edge of what they were capable of. It was able to make the game intense for all players, without making it oppressively difficult. And then, it let off, giving them a chance to rearm, recuperate, and talk about what had just happened.
And we started to see teams succeed, to complete episodes. They loved it.
The Director was giving play testers the kind of games that you remember. The ones that you can’t wait to share with your friends. These are the games where everyone else on your team is down and you save the day with a perfect grenade or a series of kinda-skillful-but-maybe-also-kinda-lucky dodges. They’re the moments where things go so bad that you barely squeak by a fight and then race to the safe room, slamming the door shut inches ahead of the horde.
One of our external playtesters summed the experience up perfectly, “We barely made it. Just barely… but a lesser team could never have survived that.”
Skill wise, they were an okay team. But they’d played the game of their lives to beat the first episode.
In most games, you only get those moments when your team’s skill level happens to match up with whatever difficulty settings the developer set when they shipped the game. In The Anacrusis, our reimagining of the AI Director precisely tunes the difficulty of each game to put every team in those magical situations. Where players have to go beyond what they thought they were capable of in order to succeed.
And that’s the difference in The Anacrusis. That’s why we think our Director has earned a version 2.0 label. The Director 2.0 knows everything about the game. It knows how you’re doing, it knows whether you’re tired or you’re ready for more. It knows when you need healing and when you need a horde. And it constantly puts you in those situations that you’ll want to share with your friends—where you’re under-the-gun, with seconds left to live and only one chance to succeed. But the Director can only push you up to the edge. Whether you live or die is still up to you.
- Chet Faliszek