PDA

View Full Version : Research Methodology



Tamur
10-11-2004, 21:02
Not a topic for the faint of attention-span...

I know there are a few around here who are doing some serious research into the SP game's mechanics. Would anyone like to share tips of the data-generation and interpretation trade? A couple of specific questions:

1. How do you define the variables that could be skewing your numbers? That is, for battlemap unit research, we know about unit size, fatigue, armour & weapon upgrades, elevation, tree cover, morale, and flanking. Understanding those eight factors and neutralising them helps to keep unit battles even and makes the data meaningful.

Economy, on the other hand, is proving a tough new nut. There's building upgrades, port/road upgrades, squalor, Trade Rights agreements, blockades & brigand-age... but there must be others because those five don't explain the number's I'm coming up with. Diplomacy is the one I'm focused on, and the same problem applies there.

2. How do you limit the environmental variables that MIGHT be influencing your data? (and yes, I already tried writing a Perl script that goes something like { for each $env in @ENV } :clown: ) For diplomacy, for example, the number of factions involved and the possible "moves" each side can take is making a big jumble out of what might or might not be good data.

Err, thanks. Kind-of a hefty question, and I don't expect anyone to give the entire answer. But thoughts from those who are active will be very appreciated.

therother
10-11-2004, 23:37
Not a topic for the faint of attention-span... Are you going for the "Understatement of the Day" award or what! ~:)


I know there are a few around here who are doing some serious research into the SP game's mechanics.I suppose I qualify then...


1. Economy, on the other hand, is proving a tough new nut. There's building upgrades, port/road upgrades, squalor, Trade Rights agreements, blockades & brigand-age... but there must be others because those five don't explain the number's I'm coming up with. Diplomacy is the one I'm focused on, and the same problem applies there.A lot of patience, repetition, observation, then theorising on the data, predicting an unknown result, then testing to see if your prediction is accurate. Repeat. Repeat. Repeat. Much the same as any scientific investigation.


2. How do you limit the environmental variables that MIGHT be influencing your data?Modding, use of RomeShell, editing memory addresses, and basically anything my fiendish imagination can come up with!

It that the sort of thing you were looking for?

andrewt
10-11-2004, 23:49
1. Quicksave, end turn, quickload.

2. Change one thing at a time.

3. For example, I noticed that trade is affected by population earlier. I needed Alexandria's grain for a few more turns to grow Antioch into huge city. However, Antioch switched from trading with Alexandria to trading with Halicarnassus because the latter became more profitable. I experimented with changing taxes, population, etc. and found out about the link. What I eventually did is build tons of peasant units in Halicarnassus for a few turns and canceled Alexandria's peasant building and disbanded a few. I managed to upgrade Antioch to huge city.

Tamur
10-12-2004, 04:15
Yes, thanks andrewt & therother.

I'll have to develop the patience to do one turn several times over varying little things here and there. It was nice to see (just now, after reading the responses here) that if I reloaded a turn and did the exact same sequence of actions, I got the exact same results. I was worried that even that would vary widely, apparently not.

I did get a memory editor today. How to use it for research, that's a different matter entirely!

therother
10-13-2004, 18:48
A good little utility for your research may well be RomeSAGE. Along with the ability to change the difficulty levels, it can turn off the battle timer, change the default unit size, and has an experimental feature to change the length of campaign.

You can get it here. (http://www.twcenter.net/forums/index.php?showtopic=11393)