MEGATON...... Well, nah, not really. But it seemed people are having trouble getting a feel for the character of this new format, and there were a few requests for number-crunching, so I ran some simulations and
WHAT HAPPENED NEXT WAS SO INSPIRING.
First up, before presenting my results, I should be up-front about what my computer program is and does:
- This was a quick job, and could be all wrong, so please check the code if you're going to somehow use this as a basis for further MEGATONS. (However, in defence of my couple hours of work, I'll remark that the code gives the kind of curves I was expecting to see and so should be at least in the right ball park.)
- It simulates completely random decisions. Random leaders, random lynches, and random infections.
- Only the human, infected, and alpha roles were taken into consideration, as there is no plausible and tractable way to teach a computer program (e.g.) how a seer would act or what might happen if one were infected. (I'm a scientist, not a game AI programmer!)
- I obviously can't account for any special events that Moggy might run.
Here is the complete code to the program so anyone that wants to can critique, fact-check, or update it. In the spirit of good science, it would be great if
any player with the ability to read Python looks over this and checks it seems sensible. Please,
don't take my word for it: verify it for yourself. You will need the two main Python scientific computing libraries, Numpy and Matplotlib, to run it. Good results can be had from as few as a thousand iterations, but I used one million for the sake of certainty.
The results are as follows:
Probability of a human victory: 38.1%
Probability of an infected victory: 61.9%Again, this is a rough-and-ready approximation to the actual game and certainly shouldn't be read into too deeply. It should however be remarked upon that, at least in this model, the long-term odds are against us: most games in the simulation ended with infected victory after
24-26 days. However,
we are not random actors. With intelligent play over the next three weeks, and by taking care to utilise any special roles or events, I'm sure we have every chance.
Further work: I may update this soon to have more realistic lynches, i.e. if the random leader is an infected then only a human should be targeted. This assumes that the infected will all be informed actors, though, and I wanted to keep it simple for the first attempt.