Quantitative Spring Economics

[Argh]

He said that my answer wasn't addressing his question, so I answered his question.

In terms of establishing evidence, it might be useful to see if there is much coorelation between people's win/loss and their economic stats in typical 1v1 match play. Maybe get with Gota, organize SA matches all played on the same map, in the name of Science.

My guess: you will see surprisingly-large variations, you will see a point where outcomes really start to branch, somewhere around the chaos point, and the economics will not be a reliable predictor of outcome.

[squeakycleaners]

Heres a book that could be of use for modeling spring:
http://books.google.com/books?id=xEpaT6 ... ls&f=false

Again I repeat, the goal is not perfection, but useful results. Even the above book on war modeling states that. You would use the results as tools, not as guides. I reckon that these types of models could be of use to AI programmers too.

[Dragon45]

It's so damn obvious that half this thread has no training in anything approaching advanced mathematics that i just want to stab my cortex with a steak knife. not saying which half



gg all

[Caydr]

I was going to say something similar, but I wanted to wait for someone to do all the work for me. I'm efficient that way.

[Argh]

If you want the math...

[EDIT]Holy shit, the probability section, lol...[/EDIT]

[Dragon45]

If you want the math...

I said *advanced* math.

[Argh]

Read ... hmm ... page 317 onwards. Personally, I found the author's description of what he thinks the actual utility of mathematical models pretty amusing, though. As he put it, it's highly reproducible and easy to manipulate, but has little correlation to actual outcomes.

[Dragon45]

[/troll]

Okay, math looks very comprehensible, but I'm not sure how much of this is actually relevant... I'm seeing some stuff here related to target acquisition, application of some probability distributions, random variables etc to target picking, anti-air penetration, and so on...

Looks like Chapter 11 and 12 are the most relevant here but I can't exactly access them via this preview. wot Something tells me they wouldn't be terribly useful anyway, I don't think this book is meant to address meta-strategic growth and expansion.

I think we'd be better off deriving own basic stuff and throwing LP, discrete event simulation, and calc at it, as I said. And by "we" i mean anyone except for myself.

[troll]

[Argh]

I think that it's a fairly rigorous treatment of modeling approaches to actual combat, myself- just awfully hard to apply when you're dealing with heterogeneous combat mixes that you see in a RTS. I think that, for most situations, it's massive overkill, since running those models to even arrive at loose approximations of kill ratios will fail to take into account all possible variables.

"Yeah, your super-tank is awesome, but I can kite it to death with my cheap little robots!" IRL, this stuff doesn't happen, because you only rarely have ranges and speeds where it's remotely possible to even try such a thing, but in RTS games, it's a constant problem.

But yeah, it's not strictly suitable for the economic problems. There, the OP's probably on the right track, by using the phrase "opportunity cost". If you think about it, it is almost certainly possible to build an ideal set of moves to reach an optimal outcome X for resource Y at time-stamp Z, until you hit the limit of the probability space.

You can even be fairly sloppy there, and simply make a few assumptions, and arrive at reasonable results, like, "on this map, its actually better to go from the nearest of the three metal patches to the farthest, then back, because you won't choke on M in between then".

I know that phrase must make you cringe, but it's the best concept I've come up with, to see the black hole where the models always break. I'd explain it further, but I see no point- you either get the idea that there is Chaos that wrecks any sufficiently-complex game played by people, or you don't.

And then there's always stuff that breaks it that you just don't anticipate, frankly, because people do weird stuff you didn't bother to model or test yourself. Just par for the course, tbh.

[squeakycleaners]

Well all we can do is speculate until we get empirical evidence. Does anybody know of a debug/simulation/god mode in which you can control both sides at once and place units around?

[Kloot]

Code: Select all

/cheat
/godmode
/nocost
/team <teamnumber>
/give <amount> <unitname>

Have fun.

[CarRepairer]

Well all we can do is speculate until we get empirical evidence. Does anybody know of a debug/simulation/god mode in which you can control both sides at once and place units around?

Code: Select all

/cheat
/godmode
/nocost
/team <teamnumber>
/give <amount> <unitname>

Have fun.

This evidence may be empirical, but it is still only experimental and meaningless in a sense. Licho has created a website that uses data gathered from real games to determine how each unit balances out against every other.

Unfortunately it's currently broken (you can still see the link). It also could use a bit of modification where it should take into account the elo ranking of players to weight their games more than newbie games. However it's still lightyears more advanced than spreadsheet formulae and /cheat - /give units - make them fight.

[SirMaverick]

Unfortunately it's currently broken (you can still see the link).

The site itself is not broken. SpringDownloader is broken somewhere (it was used to upload that stats).

[JohannesH]

I found a decent method of testing stuff, I play some games