The way the AI designs its ships is, unfortunately, not geared towards being adaptable. The AI cannot choose to make a ship with more of one defense or another, without there being a specific blueprint for it. The AI cannot choose to put more engines or sensors on its designs to deal with a larger map, unless there is a specific blueprint for it. The AI cannot choose to put fewer support modules on its ships even if the map is tiny and its range is already 3x its radius.
The best thing the AI can do at this point is copy player ships. I like that it does that, but it does not make a smart AI, not s particularly adaptable one - how is the AI to recognize what purpose a player design was good for?
I believe the AI needs to gain the ability to prioritize components, rather than follow a strict list as in the current blueprint system.
Instead of the current 'beamweapon - beamweapon - life support - armor - shield - ecm - engine - fillwithbeamweapons' designs, I would hope to see something like this:
beamweapon
defensive system<CheckStrongestCurrentRivalForMainAttackType>
engine until <CheckAverageDistanceBetweenStars/balancingvariable>
life-support until <checkRangeFromBorderToNearestEnemyPlanet+10>
fillwithbeamweapon
Honestly, I understand what you are saying and, conceptually, it seems solid.
But I feel that the problem with the AI is it has too many choices and it cannot adequately use all its options.
I feel like the best way to make the AI challenging is to make it more basic. Give it fewer, reliable ship designs and make key changes like forcing it to only send out fleets when they are at max logistics.
Having the AI use suboptimal ship designs is acceptable given that they get buffed manufacturing speeds at higher levels. The key is don't let them use their resources foolishly. At this point the AI would be more challenging if it simply stacked 60 logistics worth of its most advanced ships at every planet. As is they build fleets that they then send out in small groups away from their real assets.
I really think simpler is better.
Goatmeal,
Not necessarily, the mass limit on a chassis/hull is for a ship what logistics is to a fleet. The difference on a ship is the components add up but in a fleet you get the slower speed (and lower range maybe(not sure haven't checked)) the code for making a good fleet, like picking ships with the right weapons and defenses would look similar to adding the right weapons and defensive modules.
Well, like what goatmeal said, if every fleet you encountered by the AI was at full logistics, you would certainly take 'some' damage, perhaps lose a single ship , and even perhaps halt your advance.
But when they dont focus their firepower, and all defenses "recharge" after each battle, you can just lay waste the stacks of ineffective, low logistics fleets.
As well, one way to prevent single fleet dominance in a single turn, is to not allow defenses to recharge on subsequent attacks. Right now, its too easy for one of my megafleets to attack 3-4 fleets (which are only 1 ship) in 1 turn, laying waste and take ZERO damage.
I'm not saying any of these improvements are bad ideas. All I'm saying is that the AI has a rather long ways to go and sometimes fails at the most basic tasks. On suicidal the only time I ever lose anything meaningful to the AI is when they sneak a small invasion fleet to one of my border worlds. And this only happens if I choose to play with the handicap of no sensor boats. Other than that there are a few slow death fleets, as you mention, but far far more single ships that just get picked off. I'd really start with ensuring they group their ships before spending too much time making each ship better designed. Even a slow moving death fleet is better than a bunch of well built solo ships. And I've also found that only a few AI end up with super slow fleets based on their preferences. Most fleets will have upwards of 6 movements points which, while not great, are still good enough to not be ignored.
It was my understanding that eventually there would be data mining of ships on beyond what was on my disk drive, that it would be looking at ship designs from everywhere. At one time, I thought it would be through the workshop, but uploads to that say they strip off the components. What do you know about any larger choice of ship designs that might be availible to the AI?
I like your approach. Whether for original designs or for borrowed designs, there has to be a finer granularity to the evaluation decisions. And it really needs the awareness of opponent ship designs. I hope a lot more of this gets exposed to XML so clever people can help the AI get even more clever.
There are many great features available to you once you register, including:
Sign in or Create Account
