By Evie Moriarty, Lead Game Designer
Photo by Fatima Martín Pérez

With the new edition of Wasteland Warfare focusing on solo experiences, we knew that we needed to make sure the AI was the best it could be. The first edition system had many positives we tried to learn from, but the feedback we received from players was that it was quite complicated, and often generated behaviour that wasn’t all that smart.

Sitting down to design the new system, we quickly concluded that ‘smart’ shouldn’t necessarily be the goal. When you’re playing a Fallout video game, you’re rarely blown away by the intelligence of the enemies you face. In fact, the joy of playing the games was often coming up with ways to trick the AI; lure it into traps, draw enemies away from their allies, fortify bottlenecks, and take them down as they came through, and so on. What was fun about the AI was the way you could outthink it, play with it, and also how flavourful it was to the various denizens of the Wasteland.

We realised that the goal shouldn’t be to make an AI that pretends to be a player (we were always going to fail to do that), but instead to make it feel like you were playing a Fallout video game on the tabletop. The enemies you run into aren’t trying to do the same kind of thing you are, outside of heavily scripted moments. They’re mostly just hanging around, maybe scavenging things, or defending their territories and themselves. We realised quickly that making an AI system that evoked that in a fun way was actually a lot easier, and set to work turning it into a reality.

The outcome is a completely new AI system that doesn’t try to create a fake player that you face off against. Instead, it tries to create a living environment that reacts to you and that you can trick and interact with. Instead of a player-versus-player experience, where one player’s actions are driven by dice rolls or card draws, we wanted to create a player-versus-environment experience, where the AI system mediated how the environment responded to player actions.

Every enemy model in the game now has a behavior. There are six standard behaviors, which means you can have them on a single reference card and never need to look anything else up when you’re playing against the AI. Here’s an example of one of them, the Skirmisher, for enemies that are primarily ranged and will seek cover to shoot from:

*Design not final

When you activate an enemy model, they will take one to four actions depending on how much threat it has built up (more on that next time). Each time it takes an action, it’ll start at the top of the behavior list and try to do that action. If it can’t, it’ll move one step down and try the next action, and so on.

The symbols next to the actions listed show the limit on the number of times in a single activation that a model can perform that action. So for our skirmisher, if they had three actions to perform, they’d do Take Cover at most once in that activation before beginning to Open Fire instead.

What these actions involve specifically will have to wait for the rulebook itself, but they’re pretty self-explanatory. The Skirmisher is going to prioritise getting to a position in cover, then opening fire on an enemy, and it start to advance towards the enemy, but only if it is threatened or has no choice. Finally, if it doesn’t know what else to do, if all the other actions are impossible, it’ll default to taking cover to avoid attacks.

But how do you know when to activate each enemy model? How do they pick their targets? That’s for the next diary entry! See you next week!