Brad Wardell's site for talking about the customization of Windows.

image

From time to time I like to do a full AI evaluation to see where a given game is.  The reason I started programming in the first place was to write AI.  25 years later, we have Crusade.

So how does it stack up?

[[..]]

I've scored the other AIs Stardock (or I) have developed over the years and I suspect my literal score changes slightly due to faulty memory.  But here is where I see the various GalCiv games. 

BTW, while the score is 1 through 10, it is not a linear quality bar but rather more exponential in nature.  That is, a 6 is a lot better than a 5 and so on.  A 5 would be what I'd consider "game industry acceptable for the genre".

So here we go:

  • GalCiv OS/2: 8
  • GalCiv I: Windows: 7
  • GalCiv II: 1.0: 5
  • GalCiv II: Dark Avatar: 7
  • GalCiv II: Twilight: 5
  • GalCiv III: 1.0: 5
  • GalCiv III: 2.0: 6
  • GalCiv III: Crusade: 7

I may have some bias against GalCiv III's AI since I didn't write it.  I consulted on it.  I did work on the GalCiv III 2.0 AI which is where I started to learn the architecture.  With Crusade, I implemented a multi-core AI (AI jobs get spun out to the least busy core which is why turn times are fast...and why you get stuck turns).

Strengths:

  • Can beat most players on normal without cheating (and no, it can't see everything, 18% of its turn CPU time is spent figuring out what it can see or guess things based on my knowledge of how the map generation is done -- something an expert player would notice too).
  • Is REALLY good at trading overall (and yes, I just fixed that ship cost thing, very annoying).
  • It is quite good at military engagement.
  • It is quite good at colony construction

Weaknesses:

  • Incredibly wasteful with administrators.  No one needs refunding from constructors more than the AI. 
  • Not good at finding detours to get to a destination to route around problems. I wish GalCiv had a way point system. hint hint.
  • Still new to the promotion system, doesn't use Commanders well yet.
  • Still new to the global economy, doesn't use leaders well yet (still better than most players).
  • Needs to use starbases better
  • Still learning the new tech tree system. I would like to have the game send me your data but the PR people say we could run into issues. But if I knew what paths 100,000 players were using, the AI could start using it.

Turn 500

Let's check in on an AI vs. AI game on turn 500.

image

So Jar Jar is allied with the Engineers, the Vorlons and the Irridium Corporation but at war with the Arceans, Federated Suns, and the Spathi.  So there's a big galactic civil war going on here that my race, controlled by AI, has stayed out of.

image

Many of the good asteroids are not being mined.  However, he only has 109 billion credits (bc) to work with so there may be a reason.

image

I'm playing as  a Xenophobe, Colonizer which, in my build, means a 100% boost to research and social construction and a 50% drop to ship construction and 1 admin point per colony.  Here, he's moved his leaders in a way that seems reasonable and is making 79 bc per week.

image

Definitely did a nice job laying out the planet.  I don't know if I would have built the preparedness center and probably wouldn't have built the market center.

image

Currently stealing tech from the Federated Suns.

image

Gungan fleet is no joke.

 

So overall, it's looking pretty good.  But...it can get better. A lot better.  And over the next two years, in free updates, it will continue to get better.


Comments (Page 3)
on Aug 21, 2017

Frogboy


Quoting BenjaminLight,

I'm rather let down with the difficulty settings. The AI needs lots of procentual buffs to be a challenge.  Which I strongly dislike.

 

I really want the battles to intellectual. If they have just lots of advantages it feels very cheap.

Currently playing on incredible about 400 turns in (24 players, 24 player map). The early game was very rough, where I had to fight off multiple AI's. Later in the game, whilst I'm long way from finishing. Haven't seen a fleet that is a challenge. Even the strongest AI that is supposed to be 4x stronger than me, has not sent a fleet that would be a threat. 



 

The AI, even on incredible, doesn't get any boost to its galactic resources.  This is a tough one because super killer ships really need a lot of galactic resources at once.

 

One possible but "graceless" solution to this issue is to reintroduce the stepped research for ship components.

In GCII  you did not just have warp drive,  you had warpdrive 1,2,3  etc before getting HyperWarp 1,2,3 etc...

Build a branch of of each one of the ship techs and call it Improvements.       Have either a maintenance, a build cost/resource, or some other stepped improvement.   Make them expensive to research..   but perhaps make it so if you do research them, the next tech down the path is less expensive??   


If you have limited Resources these become very very valuable...     if you find yourself with excess resources, then you would want to just continue down the normal path for the next tech.     The AI would in theory be even better at figuring out the "value" of lowering resource costs over going to the next tech...

 

on Aug 21, 2017

Then the AI needs to gain that skill.  Also, it needs to brag about it when it does something strategically significant.  "Ah ha! I am dominating the Anti Matter market to supply my Hypercrunch level of Missile warships!"  Let the player know the AI is doing something clever and let the newbie player know some clever ideas to improve gameplay as well.  Makes the game look overall smarter, like it knew what it was doing.

 

on Aug 21, 2017

erischild

Then the AI needs to gain that skill.  Also, it needs to brag about it when it does something strategically significant.  "Ah ha! I am dominating the Anti Matter market to supply my Hypercrunch level of Missile warships!"  Let the player know the AI is doing something clever and let the newbie player know some clever ideas to improve gameplay as well.  Makes the game look overall smarter, like it knew what it was doing.

 

... the forum won't let me give you kudos... but yes.. the overall diplomatic taunting can be much improved.

on Aug 22, 2017

@Frogboy

thanks for your continued work on this game.  I'm new to the series but it's nice to see devs responding to the community, and you seem really passionate about improving the AI and gameplay. 

Though I've had only a little experience so far, I think there some changes required from a balance standpoint, and that many of these would help streamline the focus for the AI. 

The most abusable power differential right now, as I noted above, is move speed.  This I have found gets even more egregious later into the game, due to the ability to stack support modules (another questionable decision.) 

In my current game, I am only in the midgame, but my planet cracker fleet has 50 move (as well as outgunning all opponents by a factor of 10) due to stacking move speed enhancing fleet support modules.  This should not be possible, except maybe on Ludicrous sized maps.  50+ move in a turn based game like this completely destroys strategic balance.  The enemy cannot plan and respond to threats, because I can literally hit all his unguarded starbases and planets in a single turn.  Positioning and scouting are rendered irrelevant.  Even if both sides were taking advantage of this exploit, the game would not be balanced, as tactical decision making in a tile based game relies on the ability to predict your opponent's options and respond.  You cannot do this when your opponent can cross the whole map and hit any target in a single turn. 

So, step one is to remove fleet wide support module stacking, and tune move speed bonuses based on map-size.  for most map sizes this should mean a severe nerf to available move speed bonuses.  

This will help the AI out in several ways, as I'm sure you know.  First, it will penalize them less for not having the correct ship designs that abuse move speed.  Secondly, it should improve tactical movement decision making, as fewer options=easier AI optimization.  Finally it should free up system resources from being wasted on long pathfinding algorithms (problematic on larger maps.)

___ 

The second major change I would suggest is around invasions. 

The fact that you can capture planets costlessly if they have no legions stationed there is not good for game flow.  It is much too easy to get an early lead with a blitz attack and a single transport, especially given how expensive it is to station a legion on a planet in the first place -- you have to create a special structure and build a legion or spend a citizen on a general, then spend more production to garrison them.  Defense is more costly than offense, AND less flexible.  This shouldn't be the case.  

My suggestion: give population some resistance value equal to some fraction of a legion, say .25.  However, they don't gain benefits from Soldiering technologies (meaning the ratio of population to legion favors the legion more the later the game goes, this prevents large populations from being too hard to capture late game). 

In addition, make legions cheaper to build.  750 is too high, given they are necessary for early game defense.  Instead, increase the cost of the transport module.  

Overall, the impact of this is to make planets harder to crack, as they will have higher defensive value from population, and more legions available for defense.   Additionally, casualties on the attacker's side will slow down momentum rather than allowing rapid fire jumps from system to system. 

Combined with the lower move speed of ships, blitz rushes of enemy civs will be severely slowed down.  In all my games so far, I've been able to wipe out an enemy civilization in just a few turns with 1-2 transports and one planet cracker fleet.  this should not be possible.  It should take a couple turns per planet, per transport.  

This will make combat more of a tactical slog planet to planet, rather than a smash and run blitz. 

These changes are significant, but necessary in my opinion.  The game as is has so much promise, but is utterly broken to the point of being trivial for anyone with 4x tactical expertise.  

I hope to see your continued support of this community and game pay off!

Cheers!

Meta
Views
» 37287
Comments
» 34
Sponsored Links