DISCUSSIONS
- << previous
- 1
- next >>
So, you guys are developing a historical-geopolitical simulation game. i'm curious, what sorts of tests do you run to ensure that the thing runs quasi-realistically? i expect you play test-games, to see if for a player things tend to go in rational directions (e.g. Japan will usually invade China, Britain will usually side against Germany) - but if you let the AI play itself, how much time does it have before things go totally haywire?
a real simulation would be tested in this way : by running it over and over again, recording each run, and then observing the typical behaviors. you could turn off graphics and sound and other user interface stuff, and just run the game-with-AI for 10,000 5 year games, recording each game as a list of turns containing all important information. then, look at the average behavior of the game.
you probably would find that in the first months, behavior was pretty predictable, following your preset WarPlan likelihoods (i.e. Germany invades poland ~70% of the time because that's the value given to Germany etc. etc.). but as things went on, they'd get more and more complex, and less and less predictable. with a method like this, you could actually estimate the likelihood (given your simulation) of the Actual Historical course of events taking place - it would probably be very low, depending on how lenient your terms were.
if I had a fast enough computer, and if i knew how to read relevant info out of the savegame files, i'd do it myself - but i'd never figure that out. anyways, I propose a Monte Carlo simulation of mid-20th century history using the MH2 engine. it already sounds like a Science Magazine article title, doesn't it? i mean, if you guys want to pitch this thing as a learning/education tool (is that still a part of the pitch?), why not do a little science project with it?
Why is it a problem if events don't go historical? Its not like what historically happened is the only logical course of action.
Is it a logical imperative for the Front of Decent People to successfully be able to give the Fall gelb warplans to Belgium, thereby precipitating the adoption of the far higher probability of success Manstein plan which allowed Germany to successfully conquer France? There are clearly many logical scenarios with France not losing to Germany.
Is it a logical imperative for Hitler to trust the leaked Rainbow Five warplans which convinced him to declare war against the United States? Shouldn't there exist the possibility of scenarios where the person who historically leaked it, never thought of the idea of leaking it? Clearly there exists scenarios where Germany does not declare war on the USA
Is it a logical imperative for Mongolian calvary to cross the Manchukuo border on may 11th, 1939 precipitating the Battle of Khalkhin Gol which convinced Japan that USSR is not as viable target as southeast Asia? Is it also not a possibility that Japan would do better than they did historically, allowing the later possibility of a joint German/Japanese invasion of the USSR?
As long as unrealistic events don't happen like France starting WW2 with a invasion of Belgium, Netherlands, and Luxembourg or US declaring war against Germany over Poland....
For MH2 to successfully model the world of the 1930s-1940s
There would have to infinitesimally large number of possible scenarios, otherwise it wouldn't be realistic
- Carl
- location: United States
- joined: Aug 10, 2009
- posts: 241
the intention of the simulation wouldn't be to 'ensure' historical behavior by the game; instead, it would be using the game as an historical simulation, where pivotal events are modeled simply as uniform probabilities, battles as relatively complex dice-games, etc.
so, the predictions of the simulation (e.g. "the general historical course of WW2 in Europe had a probability of 20% of occuring") would be interesting to talk about, in so far as the game is an accurate model.
also, if the simulations predicted that the historical course of the war was extremely unlikely (like less than 1%), we might conclude that the simulation is a bad one, or that the course of events actually were extremely unlikely.
i would tend to argue that history is relatively likely to unfold in the way that it does, so you shouldn't get very small probabilities in a good simulation. just my opinion, i am not a historian!
P(Historical WW2)=P(Event 1 occurring historically) * P(Event 2 occurring historically)....*P(Event N occurring historically) etc etc
Given the gigantic number of events in WW2, the probability of the historical WW2 occurring would inherently be very small even in the best historical simulation
A good historical simulation would be created if all the nations caused and responded to events in manners that they would in real life with realistic causes and effects
Lets take a Spanish civil war historical simulation (its much simpler than WW2)
If all nations refuse to aid either side, Republican Spain should win most of the time, as their initial position is stronger than the Nationalists
But if nations act like they do historically, Italy/Germany (but mostly Italy) and the USSR will aid their favored sides. The winner of the Spanish civil war would in most games be determined by which side quits aiding their favored side. Historically the USSR stopped aiding the Republicans as they needed the equipment for their war with Japan. The Nationalists continued to receive military equipment from the Axis and won.
Of course there would also be the probability that one of the sides would win even though the aid war was still going on. There would be nothing wrong with that, even though that's not what happened historically as there was always the probability that it would occur that way.
- Carl
- location: United States
- joined: Aug 10, 2009
- posts: 241
- << previous
- 1
- next >>
