The Sims vs Rimworld

There's a new genre hiding in plain sight. A genre about people.

Base building games are including more and more "personal simulations" about the people in those bases. For example, in Rimworld each member of your crew is simulated in great detail. Not only their stats and skills, but also their moods, personalities, traits, and so on.

But it's difficult to feel the "story" of these people. Despite getting sad over losing their dog, or having a bad case of the plague, or dating someone, the character stories don't really have a strong impact on the player.

This is because those events are framed in terms of how they affect the facility rather than how they affect the person.

If someone loses a pet and goes into a long mourning slump, the player has to try and keep them from going berserk or spending days just wandering around. It doesn't even really feel like "the story of that person", it just feels like one of the cogs in your machine is wobbly.

Because it is.

A base building game like Rimworld is about building bases. It's about creating an ever-more-complex self-sufficient machine. The inhabitants, no matter how diligently simulated, are cogs in that machine.

In contrast, consider The Sims.

It's a different setting, but The Sims has most of the same kind of setup as Rimworld when it comes to people. Each sim has specific traits and skills and goals, they tend to have specific jobs, things can go wrong - it's very similar to Rimworld in that regard.

But the house they live in is not a complex machine.

Although you can build your house with astonishing attention to detail, it is not self-sufficient and doesn't need to be. As long as there's some source of money, everything else is optional, and there's no real need to optimize your performance.

The role of "day job" is less about optimally making money and more about providing a scaffold for life experiences - it shapes both the character that goes to work and everyone they share the house with. If things go wrong or get delayed, the house will not collapse just because the day job person is being suboptimal.

Compare this to Rimworld, where it's very likely that half your base will catch sleeping sickness and then whoever is awake will get too moody and start lighting things on fire, at which point a crowd of monsters will attack. Rimworld is about creating a base that can survive all these bumps in the road, even if they pile up. So anything that goes wrong is a bump to your base, even if on paper it's the story of how Anna is depressed and Bob is sick.

The Sims is just the opposite. If something in the house goes wrong, like a sink exploding, the player will naturally contextualize it as part of Anna's crappy day rather than a systemic setback. Even death isn't a sign that the house is doing badly, it's a sign that someone's particular story has come to an end. That may be very upsetting, but it's contextualized as that person dying, not your overall facility degrading.

I think nearly all of the difference in this contextualization is simply because the 'facility' in The Sims is not a high-stress facility. You don't have to worry about droughts or animal attacks or hurricanes. It's pretty easy to establish a baseline habitability, and then everything else is just improving things more or having fun with side tasks.

The lifestyle options for Sims characters are all side options. Some are very dense, some are not, but they are all optional. They provide a stable, steadily-progressing scaffold for the characters' life stories while also providing a steady diet of life events. For example, gardening: you don't have to garden, but if you choose to, it's a steady task that moves forward day after day. Your garden will never be attacked, and even if your whole garden was destroyed, your sims would not die: the baseline habitability would not degrade that far.

If we want to make a game that focuses on the lives of the characters, it's critical that their support system is straightforward and robust, so that setbacks can be judged as affecting the characters rather than the support system. Similarly, the progression system needs to be character-centric rather than facility-centric.

As an example, Rimworld's research system unlocks construction options for the whole base once someone researches how to build it. A character-centric approach would be to allow only that character to build it.

In addition to being character-centric, a character's chosen lifestyle/career/hobby needs to provide a stable, steadily-advancing scaffold, needs to exert pressure on their life and the life of those around them, needs to provide small random events and schedule burps, and needs to respond to a character's own personality/pressure/situation outside of the lifestyle.

For example, rather than the farming system Rimworld currently has (identical to other games such as Dwarf Fortress and Minecraft), we would instead have the farmer work fields on a schedule. The more fields, the more hours of the day are required, and time spikes at harvest and planting. The result is a nice, predictable, slowly-changing curve as to how much time they have to work and how much time is available for other pursuits. Their early-morning schedule would bump up nicely against someone who gets up later - for example, a researcher or carpenter.

In theory, the fully-simulated fields work this way. But in practice, there are too many variables. For example, just walking from point A to point B can take a lot of extra time. Also, the characters aren't great at schedules, and frequently get distracted by their overwhelming need to pick up a meal halfway across the map or whatever.

By reducing the simulation fidelity, we can allow for a more 'readable' character lifestyle. This will help the player to 'feel' the character's life, personality, needs, and so on.

And I think that is where a genre is hiding.

Right in there.

Expectations of Control

I keep thinking about control in games. Because I'm a simulationist at heart. I want to know that if I push this to happen, then other things happen descending from it.

I remember reading Choose-Your-Own-Adventure books, which fascinated me even if I never followed the combat rules or anything else that required stats or dice. It didn't bother me that I couldn't "choose the things I wanted", because it was presented in such a way that it was obvious there was no real simulation involved. I was just navigating something that was already written.

Similarly, I don't feel disappointed by adventure games or other very heavily-scripted games, because it's obvious there's not any kind of adaptive simulation. Funnily enough, as soon as these games include even a hint of adaptive simulation - such as giving you choices between two distinct plot branches - I start to get itchy.

"You gave me a choice, which means you CAN give me choices, which means you SHOULD give me choices, but you aren't, which means you suck!"

Obviously, I understand intellectually that most of these choices are of the choose-you-own-adventure variety and are not simulated. But I grew up with so many simulators that the itch is very strong.

This is the reason I dislike Knights of the Old Republic, and anything else that gives you "good or bad" choices. At first it seems like a simulation to my instincts, but even when I realize it's not, I still want more detailed and reactive control over my goodness and badness.

Do any of you have this problem?

"True" Simulation

I have found that, once I tackle a subject for a few years, there is suddenly no "best" rule set. An example that is coming up more frequently these days: social simulation.

Most people come into the field thinking that they're going to try to discover some underlying truth, some supreme rule set which will be able to do all sorts of social things. Most people, when they first try, come up with some ultra-generic system of addressing some kind of "social atom", like "I'll track how much they like any given thing and how much those things like any given thing and then use a logic chain to determine what the reaction to any attempt to affect any given thing is..." and so forth and so on. The end result, when applied, is labyrinthian and never works.

This is true of other fields as well, of course. Newbies tackling any system tend to come in with a simplistic approach that will solve all the problems of the world.

But there is no "best" approach, at least in games and simulations. There are variety of approaches, but the "best" approach is whatever simulates exactly what you want to simulate.

For example:

If I approach a game based around a social mechanic, I don't think to myself, "what's a good way to simulate social mechanics?" Instead, I think, "What's a good way to make the player feel the pressure and rewards of obligations?" or "What will make the player feel the contrast between the rash relationships of youth and the staid relationships of older people?"

See, the theme of the game isn't "relationships". That's like saying the theme of a game is "blue" or "culture". Instead, the theme is something very, very specific. I generally use a comparison of some variety, because it gives me a basis for the game's core challenge.

Understanding the theme, you can then decide on the rules of the simulation. You don't need to solve strong AI for this, and you never did. Trying to put strong AI into a game is like saying that you're going to build a go-kart, but it's going to have a cold fusion reactor in it. Just build the freakin' go-kart.

Simulating Characters

My computer hasn't been behaving. In the process of writing my new Big Theory, I've lost the document four times. I'm not kidding.

It's okay, though - each time I write it, the thing gets shorter. The first time, it was more than twenty pages long (and unfinished). Now I've got it down to three. Of course, a lot of content was shaved off - the original essay included a scathing review of some misleading experiments, a section about simulating interplayer complexity, and a full game doc. Those will have to come some other time, I suppose.

Anyhow, I think this is an important theory, so I want you to read it and tell me what you think:

Improving character AI.

What do you get when you...

Took me longer to get back on line than I thought it would.

One of the things I've been thinking about is how to simulate players. Not as in "how to make AI which does what a player would do", but as in "how to simulate someone who gets excited about specific things and wants to play with the other players".

Ha! It's tough. Any thoughts?

I'll post mine once I remove the various stupid things in them...