Civ-like history simulator

[Koshling]

http://www.newscientist.com/article...on-game-shows-impact-of-war.html#.UkMjn4ashcY

 

[Hydromancerx]

Ha! They could have saved time and effort with that simulator and just played C2C to learn that.

 

[n47]

William Thompson, a political scientist who researches conflict and international relations at Indiana University in Bloomington, says Turchin's model offers insights into the dynamics of societal change.

I hope they will not take it too serious. There is too much entropy they can't take into account, to take serious results form it. Especially, as the history showed even a single man can change the world. And they certainly do not simulate every human being.

 

[Koshling]

I hope they will not take it too serious. There is too much entropy they can't take into account, to take serious results form it. Especially, as the history showed even a single man can change the world. And they certainly do not simulate every human being.

66% historical accuracy isn't bad though (provided they didn't back-propagate parameter setting using the same training data of course!)

 

[pepper2000]

I've thought that it might be interesting if Caveman to Cosmos, or something similar, could be used to actually show how history played out. It would be an interesting way to see how realistic the models are in the game. Some things don't fit very well--wars and exploration do not necessarily occur in real time--but C2C is getting ever better at capturing the broader trends of growth and stability.

 

[n47]

With my historical knowledge, this 66% is hard to believe. If you have a proper state of situation in Europe and Asia, I believe you can have 66 of accuracy about the state in next 50-100 years. But when I think about all those turnarounds caused by silly luck and single persons, I can't take it seriously in 3000.

Another thing is, 66% is nothing for serious reasoning. You possibly have 34% chance your result will have nothing to do with reality. In compare, physicists need 99.999% (or more, I can't properly remind) to take their results seriously. -- I know, they do not need to deal with the entropy of past time, ... but 34%? Come on. You can fit even mice domination over the world in it.

 

[Thunderbrd]

Yeah... a dubious test that may be as designed to produce a special interest supporting result as any real science.

C2C has made it abundantly clear that there are simply too many variables to consider. But results can certainly be rigged.

 

[raxo2222]

Too many variables... I think even 2050's best supercomputers won't go over 99% lol