It. Is. On!

Our story so far…

As an AI test, I see how well the Drengin do in what would normally be a multiplayer game. I have this map where I play as the Terran Alliance run by those crazy primates from Sol 3 (aka Earth) from the year 2245.

My opponent are those brutal, skin eating bastards from Drengi who recently wiped out their own sub-species the Korath because, they’re just that nasty.

In the previous tests, it’s been a sad sad story of me just wiping out the stupid AI.  But now a month of heavy duty AI coding has gone in.  Let’s see how they do.

Notices

First off. I’m turning off FOW.  I’m doing this to better view what they’re doing.  So we can all see what’s going on.

Second, I’m playing on normal. So the AI gets no advantages, bonuses, etc. whatsoever.  So overall, as the human, I have a slight advantage.

Starting Locations

In this corner the humans get their watery planet that they have ironically named Earth.

The main stats:

• Manufacturing: 7.7
• Research: 8.3
• Net taxes paid in: $6.7
• Approval: 50%

 

Now, let’s look at that polluted crap hole called Drengia.

The main stats:

• Manufacturing: 12.5
• Research: 10.8
• Net taxes paid in: 0
• Approval: 40%

So they’ve already adjusted their spending and rush built a work camp.  The Drengin are naturally disconent because, again, their home is a crap hole.

Turn 0

This is the whole map.  The AI cannot win through attrition here. It has to play intelligently.  There’s no opportunity here for shenigans since I can see everything it is doing every turn.  The goal isn’t to be fair but to see how well the AI plays.

Early Choices

I am going to play as Benevolent.

I chose to rush colony ships a bit faster than the Drengin so I take an early lead in planets: 4 to 3. 

The AI makes a mistake: Tries to colonize Mars.  I don’t have an API yet that lets me make a good judgment (without hard coding which we’re avoiding) on whether a planet is too far in someone else’s influence to bother with.

Expand: Complete

By turn 30, all the planets are filled. How we doing?

They’ve caught up in power.

And in total production:

 

Exploit and Exterminate:

But I have planetary invasion and in fact, I have 9 techs they don’t have. They only have 3 I don’t have.

I am also going to see how much I can exploit the game’s influence system.  Have we balanced this properly?

AI mistakes

The AI should be able to take notice that I have invasion tech and start putting defenders in orbit. Instead, they’re building constructors

Which is a shame because I’m totally going to exploit it.

And so I do. And they die.

Where did  they go wrong?

That’s easy. We intentionally make it easy for a player to see what techs other players have. It’s vital. And the AI simply didn’t respond to the fact that I had invasion tech.  They should have switched to building some defenses. If they had done that. It would have been a very different game.

UPDATE: THE REMATCH

The rematch

After a day of coding I think I’ve gotten them to be much better in the area of situational awareness.  I am confident I can beat them in a fair fight but I expect to have to work a bit more.

Let’s find out:

Turn 0:

This time I’m optimizing from the start so it’ll be even harder on the AI.  I’m going to b-line to warfare and conquer them.

• Manufacturing: 6.7
• Research: 17.7
• Net taxes paid in: 0
• Approval: 50%

vs.

• Manufacturing: 12.5
• Research: 10.8
• Net taxes paid in: 0
• Approval: 40%

Now, the thing to remember here is that the particular game I’m playing is not representative of a normal game. I am b-lining in tech so I have my tech way up at 18.  Plus, the Drengin tend to focus more on manufacturing. 

The other thing to remember is that the Drengin are not as good as the humans economically but are really good later game with their fleets (they pay slightly less in maint). This means that in a short game, I have a slight advantage by picking Terrans.  Once I’m satisfied with the AI, I’ll probably do Terran vs. Terran games.

Exploit and Exterminate!

I got invasion in only 36 turns. I bought up research centers instead of taking an extra planet.

This means we both have equal numbers of planets (4 to 4).

and this time..

The Drengin have started building token defenses just in case.  The system is actually reasonable accurate in that it can tell whether I’m in range of their planets (i.e. if someone is off on a huge map they won’t care as much).

He’s definitely keeping an eye on me.

Here we come…

The Drengin have a Mark II defender class ship (they call the Tho-Altha) in orbit. That…is unfortunate.

Let’s take a look at the home worlds..

• Manufacturing: 17.3
• Research: 18.2
• Net taxes paid in: 6
• Approval: 58%

The Drengin are putting some effort on getting some money.

• Manufacturing: 23.1
• Research: 11.8
• Net taxes paid in: 1
• Approval: 51%

We are relatively even.  The AI is being punished a little because approval doesn’t currently matter as much as it probably should. The AI is taught to keep its approval decent.  I think eventually high approval will translate to production bonuses.

Too late

He saw the threat but was too slow to do anything about it.  He had defenders on his border planets but not back at his core worlds.  Admittedly, lack of FOW made that possible.

I won’t be able to get Kona though.

Meanwhile, I took one of his other planets.Not sure what happened to that ship that he had.  I know the AI can get a little aggressive in decommissioning “old tech”.

What he does have are pretty decent ships. His battle cruiser is giving my battle master a work out.

However, and this is where GalCiv III shines: I have a Prototype ship called the Sherman. Prototype ships are the result of owning special resources. In this case, I have an anti-matter resource.

Anyway, it’s only a matter of time.  I need to get the AI to surrender at this stage since it’s hopeless.

What mistakes did they make?

The first one is something that might be a bug. One of its shipyards lost its sponsors because they were in my territory.

The second issue is that the AI probably should have a cheaper defender ship. It shouldn’t be about having a great defender but rather an inexpensive one that does the job. Will have to think more on it.