By beta 5, it should be pretty clear that GalCiv III is an entirely different beast from GalCiv II. This post won’t go into the game play differences, which are pretty vast. This is about something a little more dear to many of our hearts: modding.
Why GalCiv II was the way it was
The thing about GalCiv II that I would like you guys to understand is that I wrote nearly all the economic, diplomatic, AI and other code personally. I enjoyed doing it but it also meant that *I* wrote it. It wasn’t data driven in any way. If I wanted the Drengin to focus more on getting some tech, I modified C++.
GalCiv III, by contrast, has a much bigger team and while I designed (or helped design) the high level way that the economic, diplomatic and AI works, I didn’t program it. And since the talented folks on the team are aware that we will be making GalCiv III expansions into the next decade, the entire thing is data driven. This is good news for modders.
Your first mod
If you hit the ~ key you can bring up the game’s console. From there, you can type “soak” (no quotes) and “fow”. Soak has the AI play itself once you hit the turn button. fow unhides everything.
In GalCiv II, if the AI was doing something stupid, well, too bad. Because you’d have to wait for me to fix it and between my various opium walkabouts and general laziness who knew how long that would take? But in GalCiv III, you can make the AI better the same way I am, through changing the data.
What makes the AI decent?
Right now, in beta 5, I’d say the AI is…adequate for most people. Which is to say, it’s not very good by my standards or the standards of those reading this post. This isn’t due to the AI being badly written, it’s all about data and strategy.
So what makes an AI in a strategy game decent?
- Having a focused tech path
- Knowing what techs it must get
- Understanding timing attacks (i.e. by this point you should have X)
- Being good at macro strategies (building out, expanding).
- Being good at micro strategies (make this planet do this, make that planet do that)
- Being good at having good ships and at the right time.
- Focusing your attack on one player at a time and not attack everyone
- Organizing the tip of your spear to be as sharp as possible (no death trains)
- Focus its spending on what its current strategy is effectively.
- Build a network of friends diplomatically and make good trade.
At this stage, the AI gets an F at 1, and F on 2, and F on 3, a B on 4, an A on 5, a C on 6, a B on 7 and a B on 8 a D on 9 and a C on 10.
This might be a bitter pill for some of you but for release, we only need to get everything up to a C. If we do that, the AI will satisfy nearly everyone. Post-release, me and you guys will likely work together in ways we can only dream of to bring all these to an A.
Where GalCiv III beta 5 currently falls down:
- What to spend its money on
- What techs to pick
These are not “AI” as much as scripted strategy based on playing. It’s purely data driven.
The Terran Alliance: Techs
First 100 turns the Terran Alliance needs these techs:
- Xeno Industrialization
- Planetary Improvement
- Soil Enhancmeent
- Environmental Engineering
- Interstellar Travel
- Orbital Manufacturing
- Orbital Speiclaization
- Zero Gravity Construction
- Weapon Systems
- Defense Systems
- Universal Translators
- Xeno Commerce
Now, a player can try to b-line over to Invasion tech but unless you’re playing on a really small galaxy, that’s not idea.
Modding the Terrans: Techs
So open up TerranTechDefs.xml. You can find those techs and find AICategoryWeight. Make these numbers big across the board because they are strategic. The higher you make them, the harder it is to trade for them so keep that in mind. I’m going to make mine all worth 25 in every category.
Modding the Terrans: Personality
Now open up FactionDefs.xml and look for the Terran Alliance. You can see the AI category weights which align up nicely with the ratings on different techs. So you can pretty obviously see what techs matter to the Terrans.
These are great numbers. Unfortunately, I’ve learned something about random number generators. Do you know what the difference between a 12 and a 14 is as far as random is? Nothing. If we want the players to really play differently (differently as far as the player is concerned) we’ll need to make these numbers a bit more extreme. How extreme? That’s the fun: Try it for yourself and find out.
Here are my numbers:
So I made the scale from 0 to 25. As long as you’re consistent (your range is basically the dice roll size). I like to have very distinct differences in the numbers.
Test #1: First 100 turns
So before, after 300 turns, none of the AI had even medium sized ships.
Let’s see how the two sides are doing now.
Terran Flag Ship:
Attack of 20, defense of 18. 63 HP.
For comparison, let’s look at what the Drengin has:
Attack of 12, defense of 18, 39 HP.
So clearly no match, one on one, with a Terran ship. They’re both using rail guns oddly enough. Not sure if that’s a coincidence or not.
Terran techs after modification
- Xeno Industrialization
- Advanced construction
- Institutional Research
- Planetary Improvement
- Environmental Engineering
- agriculture ADaptation
- Xeno Biology
- Interstellar Travel
- Orbital Manufacturing
- Zero Gravity Construction
- Interstellar Logistics
- Weapon Systems
- Missile Weapons
- Beam Wapons
- Kinetic Weapons
- Militarization
- Defense Systems
- Shield Systems
- Armor Systems
- Universal Translator
- Xeno Commerce
So it got all the techs and then some.
However, not all these choices were good choices. It should have picked a single weapon system to focus on.
Drengin techs (unmodified)
Let’s look at what the Drengin have (their tech tree is whatever is already in beta 5)
- Xeno Exploitation
- Xeno Slavery
- Persuaive Research
- Planetary Exploitation
- Agreiculture
- Enviromental Engineering
- Interestellar Travel
- Orbital Manufacturing
- Interstellar Logistics
- Weapon Systems
- Beam
- Missile
- Kinetic
- Militarization
- Defense Systems
- Shield Systems
- Universal Translator
- Xeno Commerce
So some easy progress there.
Next up: Spending.
