Brad Wardell's site for talking about the customization of Windows.
Published on September 10, 2017 By Frogboy In GalCiv III Dev Journals

Last week I did a one on one test against the Drengin AI set at Normal. It was on a tiny map in which I rushed for invasion tech and killed them.

image

This time, let's bump the AI to Genius...

AI at Genius

The 2.5 GalCiv III AI is a lot smarter than previous iterations. It handles fleets better. It researches better. It builds up planets more intelligently.  But it is really difficult, even in human 1 on 1 games, to deal with a rush.

Now, at Genius, the AI starts out with more money so it's not quite a "fair" game but then again, I'm not playing fair either.  Let's see how it goes.
[[..]]

Differences from War V

This time, I know the map.  I'm playing on the same map as before so I know exactly where everything is.  The AI does not, even at "Genius".  So this should be interesting.

image

Because I know the map, I quickly claim the planets.

image

I also destroy all Drengin colony ships I can find.

The War

image

The AI responded by building Brindle's Observatory which gives it an extra home planet.  So 6 planets to 4 planets.  I picked a Malevolent ideology point to get a free Frigate.  That, with a Commander, allowed me to destroy their shipyard.

image

To counter me destroying their shipyard, they built 3 shipyards (one for each planet).  Normally they wouldn't do that but they correctly calculated that I would wear myself down destroying shipyards.

image

Having destroyed my fleet with his remaining shipyard, the Drengin then focus all production on the remaining shipyard.  I move my Survey ship, which has weapons, into the Drengin home system and pick off their ships.  Each battle wears my ship down a bit.  It's a question of whether I can get my transports over there fast enough.

image

The invasion begins...

I take all but one planet.

image

With only 1 planet, they station all their garrisons.

image

He follows this up by buying a rather nasty Mercenary which he then puts in orbit.

image

I can't take out this guy.

He's going to still lose.  But he's really fighting it.  If I didn't know better, I'd say he was griefing me.

image

No.  This isn't that type of game.

image

He buys a second Mercenary but I ambush it.

and then take my combined fleets:

image

To deal with his powerful Mercenary friend and escorts.

image

At last. 

With his last defenses clear, my transports make their way to the kill.

image

He buys another Mercenary but it's no use.

image

The invasion happens.

image

VICTORY!

AAR:

So I was wrong about the AI starting with more money.  He started with the same amount but at Genius gets a bonus to their economy. 

Probably the only way he could have realistically won is if he had simply been handed a bunch of money.  I do have some ideas on how to combat the rush but I'm not sure if it's worth pursuing since this is a 4X game and not an competitive RTS.  The goal here was to see how well the AI reacted to my actions.  There's no script in GalCiv. The AI instead looks at conditions and then values lots and lots of potential decision trees to decide what it should do.


Comments
on Sep 11, 2017

I wouldn't worry about defending rushing either.

Where the AI needs the most improvement at the moment is around resources particularly those that enhance combat capability.  I really enjoy the resources mechanics in Crusade, there is nothing similar in Stellaris, it's great!  But the AI is glacial at acquiring resources and even trade ships seem to get a higher priority.  While the AI is building trade ships, I'm happily acquiring all the resources I'll ever need to build fleets filled with Elerium beam weapons or Thulium Kinetic weapons with Antimatter engines and there is just no way the AI can ever put together a fleet that will be remotely close to being competitive.  

Some decision trees considering what resources are available would help the AI, integrated with research / design decision making.  No Elerium nearby?  Well then focus research and ship design around Thulium Kinetics (Antimatter is better saved for movement speed bonuses).  Decision trees which allow ship design / research without resources should be a last resort for crazy bad map conditions only instead of the default.

Now if the AI is just reacting, rather than using this as part of a base strategy, it's all too late.  It needs to have a strategy using resources in some way out of the gate.

Let's say AI had quick access to Elerium, well it can research to Particle Beams very quickly, any AI that can get access to this early with decent designs and research focus should be able to put up a decent fleet, with Thulium allowing a switch in strategy to Kinetics as required.

You could then repeat the scenario above but give the AI access to different resources combinations on the map and check that it can handle different situations.

In parallel there are some technologies that should always be a priority to any AI.  Specifically, logistics and capacity researched.  By all means given the AI flavour and personality, but some technologies just have to be done otherwise the AI has no chance.

I strongly suspect a lot of players would struggle against an AI with this capability so you might only want to enable it on Gifted.

I know you said the AI isn't scripted but the AI routinely going for 3 constructor ships with it's first pick would help immensely in getting key resources to enable this, particularly Pragmatics.

The other AI gap I see is with Precursor Anomalies which are just ignored.  I would have expected some race personalities would give these a high priority, particularly given the potential benefits e.g. logistics improvements etc.

Without this it really feels like the AI is groping in the dark, it can react all it wants, but has no chance against a competent player.

on Sep 11, 2017

I had posted somewhere a vague idea of Ideology choice modifications. 

One way to hard-crimp the rush would be a 'sue for peace' ideology choice that stops war (all war or only one civ?) for X turns.    Just a thought.

Off the cuff, it just popped into my head that perhaps /losing/ a planet should generate something - an ideology choice, or the flight of resources from that planet to the homeworld or the like.  This would make turning an early rush into an early complete-conquest an up-hill battle, as the remaining retreating forces would regather their strength and boost their resolve, so to speak.

 

just a thought, cheers, I like these little mini-writeups frog.

on Sep 11, 2017

As an aside, what ideology choices are the AI's grabbing in these mini-runs, any way to see them?

on Sep 11, 2017

tid242

Off the cuff, it just popped into my head that perhaps /losing/ a planet should generate something - an ideology choice, or the flight of resources from that planet to the homeworld or the like.  This would make turning an early rush into an early complete-conquest an up-hill battle, as the remaining retreating forces would regather their strength and boost their resolve, so to speak.

great idea, easy to implement. Developers love that.

on Sep 13, 2017

I really enjoy these action reports from Frogboy..  and I agree AI rush defense is not much important in this kind of game  

So far (in Crusade gold) the AI has a problem reacting on me stacking movement speed. No matter how much of it I have, AI does net seem to even put any engines on their ships. And if they do, they add them to fleets with slow ships. This lets me control large areas with single fleet, all the planetary defense stuff becomes obsolete and my worlds basically impossible to invade. The only chance for AI to beat me is early due to me having bad luck with planets and/or resources, and/or killing me before I manage to make my first good fleet (highest AI can do this to me on regular basis).  

Or is the problem local to my race or strategy? Anyways thanks for the patch, looking forward to see how the AI goes in 2.5

Meta
Views
» 2063
Comments
» 5
Sponsored Links