Thursday, September 12, 2013

Polishing the Mad Science

In the last post I made, I talked about how you might be able to create a mad science game. The approach is that you create experimental apparatuses that allow you to run variations on the same experiment over and over, and you get insight points based on the observations you make while doing so.

The gameplay is about designing progressions which create variations.

As an example: you start with a sample group "combustible matter", which contains samples of things like wood chips, dung, oil, charcoal, and so on. By simply placing a sample dish on the experiment table and linking it to that sample group, you are creating an apparatus where each variety is placed there and examined in turn.

Of course, you can't learn that much by just looking at and smelling the various samples, so you might add in a process so you can observe something more interesting. So you add a match to the apparatus, meaning that you put each sample on the table and then burn it, observing the different ways they burn.

The first key to constructing experiments is multiplying variations. So you have the fundamental variation of the different kinds of combustible matter, sure. But if you really want to spike up the variations in the experiments, you can add another factor. For example, let's put the same in a small vacuum chamber. Now we can ignite it with different levels of air pressure. So not only do we have the 50 or so samples of combustible matter we start with, but for each of those we have maybe 10 different levels of air pressure to test. These 500 trials will tell us a lot about the different properties of our samples, and the way in which atmosphere affects combustion.

From the player's perspective, the specifics don't matter. The player doesn't need to know precisely how well oak bark burns in various pressures. The scientist in the game just makes some notes on his imaginary science-notebook and the player watches the insight points tick in.

Of course, 500 variations is a fair amount, and time will tick by as the scientist does the testing. Even if you can accelerate time, it's generally better to pack more punch into fewer variations. For example, you could choose to put the combustable things into water of varying depths. While there is some to be learned on the interactions of water and combustion, the difference between putting the sample in an inch of water and a foot of water won't give you any significant information.

Inspiration comes from measurements that differ, see. So if your experiment wastes a lot of time doing things that result in precisely the same results, you're just wasting time. Similarly, choosing what kind of measurements to take at what points in the experiment matters, because those are the things that will vary and therefore the things that will give you the inspiration.

For the burning tests, a sensitive thermometer would be a valuable measurement to take, and that's an instrument you start with. Carefully weighing the sample before and after would also be valuable - another instrument you start with. Obviously, analyzing the resulting atmosphere for various kinds of chemical and particulate matter would be extremely insight-producing, but that sort of measurement tool is not something you start with. Measurements take time as much as anything else does, so useless measurements actually make the whole thing go slower.

So a new player might play around with a bunch of different measurements before starting to realize which measurements would be useful, and in time would be building pretty polished apparatuses to maximize insight.

An example of a complex apparatus you can make with just starting materials?

You could use those combustible samples, but create two inputs instead of just one. Grind each sample into a powder (at several different levels of fineness), then mix the two ground samples together, light them up in a variety of air pressures. The end result is something like 50 * 5 * 50 * 5 * 10 different experiments (625,000), so it'll take a while to run through them. Simple heat, speed of burn, particulate exhaust, and mass measurements would probably be enough to get a ton of insight.

Some of the samples will fail out of this test, though, because not all of the samples can be ground up. Wet samples such as oil cannot be ground into powder. So you could make your experiment a bit more complex by adding in either a drying system (to dry all the samples) or a gating system (to treat wet samples along a slightly different path that ends up in the same final process chamber). You can see how your apparatus is beginning to grow.

Want more? Instead of burning the powders, mix them into water, then harvest the three kinds of results (sediment, scum at the top, and the tainted fluid) and do experiments on that. You can even try more advanced methods, such as spinning the flasks, heating them, treating them with acid, whatever.

Want more? 625,000 experiments will take a long time, especially if we're using a drying chamber. We can accelerate that significantly by automating most of the process and/or setting it up to run several parts in parallel. For example, having 10 vacuum flasks, each set to one particular pressure level, means you can do 10 of those variations at once. Of course, monitoring all ten for heat, length of burn, brightness, particulate exhaust... that's not something one human can do. So you either need to assign ten scientists to the task, or you need to automate the sampling using things like photosensitive film, sticky tape (to catch particulate exhaust), and a temperature blotter which draws a line to the hottest temperature reached.

So... just to be clear, this is a machine with a dozen scientists working in tandem. They are drying, grinding, blending, measuring - FLASH FLASH FLASH the samples go up and someone hurries in to take the various slips of paper and film used to record the results, while another one removes the remnants, cleans the chambers, and resets them...

Yeah, I kinda like this idea.


Ellipsis said...

My main question at this point is how you incentivize players to build more efficient labs (unless you already clarified and I just missed it). Is it simply that they'll eventually run out of resources and fail if they don't hit another "demo"? Is it that they need to wait in real time for the insight to trickle in?

One of the reasons combat is popular is that it creates a distinct small gameplay loop within the larger game progression where players can excel or fail and get feedback on their strategy without failing at the entire game (unless it's a roguelike). Is there a smaller loop in the Mad Scientist Game?

Craig Perko said...

Like Kerbal, there's no "fail state". But your level of insight is the maximum of your experiments, rather than being added in, and it also unlocks new instruments.

I think that'll be enough, even minus demonstrations.

Soyweiser said...

I have been thinking of the same idea, with the large size of samples and possible experiments.

I realized that while 600K experiments is pretty high in a SP game. It might work in some sort of MP environment. Some sort of shared world.

Probably best to provide some sort of logging system to show players which experiments have already been tried.

A sort of randomized skyrim alchemy system. (Now that I think of it, the current skyrim system is all players sharing the same world (all possible reactions are now know and searchable in a wiki)).