What to improve next on Unnamed's Stock AI |
Obviously there are plenty of things the AI could do better. I'm wondering what specifically does the AI need the most improvement on to be competitive/challenging/fun to play.
A couple of improvements that are in my head currently are:
1) De-couple the unit demand system from the empire wide demand system and make it specific to each planet. This would allow me to improve the AI by having it build more defenses in specific situations. For instance, the AI colonizes a planet in a enemy system, then it would immediately begin building some defenses before building facilities. Planets in core systems wouldnt build many units except to load onto ships (fighters, drones, troops). Border systems would build units to defend its warppoints to unclaimed solar systems and enemy solar systems (especially fighters, mines, sats, etc). This should improve the defensive capability of the AI. Currently it more or less randomly builds units on a empire wide demand basis.
2) Create a primary enemy selection function. This would determine a primary enemy for the AI to focus on based of proximity, score relationship, log files of combat and intel, allies' enemies, etc. Once a primary enemy is selected the AI would focus attacks on that specific enemy first and all other enemies secondly. This would focus both military and intel attacks to truly try and destroy a specific empire instead of merely attacking all enemies equally or randomly.
3) Work on the AI fleet scripts to try and improve its attacking capabilities to really hit home, and follow up on attacks. (not sure specifically what needs to be done, would figure it out as I go).
4) Work on the intel scripts, so that the AI adjusts how much it spends on defense and how much it spends on offense depending on the logs it receives each turn. This would allow the AI to defend and attack with intel more efficiently.
Anyone have preference for which one I work on first or ideas on what else I should improve. Research is still being improved slowly but surely, and the resource management/colony management is pretty decent now.
-unnamed

Re: What to improve next on Unnamed's Stock AI
I'd have to say number 1, along with the population scraping noted above.
i've also found an issue that annoys the beans out of me, tis one thats transfered over from IRM, it annoys me so much i've had to stop using IRM. the ministers force you to have a spaceport even when not needed due to racial trait, tis most annoying as i cant stand space ports as they arnt usefull enough imo, might have to change em a lill when i start making my own hybrid mod ^_^ but thats a long time coming. (will probly end up using this AI as i cant code well enough >_
Re: What to improve next on Unnamed's Stock AI
I haven't seen any supply-less AI ships in my current game. They don't seem all that good at repairing their ships though.
AIs still starve themselves out very quickly when using Finite Resources. It seems better than stock (?) but still too easy for them to deadlock their empires. Much higher emphasis on building to value improvement tech? (or other green/renewable techs - I love my Nature Shrines) It's either that or they need to start remote mining. Aggressively.
Can the AI's use a resource converter? Could they be taught to?
How do AI's determine where "good" worlds are to colonize? It seems like they need to explore for "good" worlds (their atmo & planet type) more/faster and colonize them more quickly. I routinely find uncolonized systems with worlds that would be ideal for the species in question, just a couple jumps away from the AI's existing colonies, and no signs of traffic headed out to colonize them.
I, too, turn off intelligence and would like to see the AIs not waste points on it in games where the projects aren't allowed. Guess we wait for MM to code in a function for it.
P.S. great job with the AI mod, looking forward to more improvements!
Re: What to improve next on Unnamed's Stock AI
Here's how I remember which systems have unclaimed worlds in them still, maybe the AI could do the same thing.
If a new colony is in a system that I haven't had a colony in, I have that colony build a single satellite first. It's a spotter sat. At the start, it is nothing more than a sat computer core and the best basic sensor I've researched at the time. It lets me see the planets near mine so I know whether they're still available for colonization.
Later on, after Tachyon sensors' range catches up to the basic sensors, I might switch to those...but early on, it's all about the cheap peep. In my current game, thanks to some early ruins my missile tech was a bit better so I put a single missile, a supply crate and an ordinance crate on the second generation sat design as well. Quite a bit more expensive but it kept the tourists away, and I had one AI species that kept sending scout frigates to a few planets...one at a time...and that lone missile was quite effective for a surprisingly long time.
Re: What to improve next on Unnamed's Stock AI
Unnamed,
I haven't played your mod enough to find any extreme problems with the AI but I would just like to say thanks for making it. All us guys that want to play the standard game against a stronger AI very much appreciate it. The fact that you are also here looking at ways to improve it, and responding to suggestions is great. Thanks again and best of luck with your continued modding, I'm looking forward to it.
Innuendo.
Re: What to improve next on Unnamed's Stock AI
There has been no effort to teach the AI to handle finite resource games so far. I'll do some improvements for the next release that increases priority of value improvement facilities in finite games (research and use).
Currently the AI doesnt use the resource converter and I dont think the function is yet available to teach them to. Kwok, if you see this, any thoughts?
The AI colonizes the closest planets first and works its way out, with some preference for breathable atmosphere and planet value. Originally IRM had AI go for the best planets first, but since colonies are not remembered, the ships would travel several systems, only to find the planet colonized by an enemy. This slowed down growth considerably. I dont think there has been a ideal inbetween solution found yet.
-unnamed
Unnamed:
Check the Ship Orders file I sent you, It has a start at a decent priority based queue system on where to send colony ships, taking into account breathability, size and ruins, and planet valued factoring in distance to travel, It's a start which I'm still working on, but once the forumla is worked out, I think it would suit the purpose you're looking for.
If you need help implementing #1 or want me to break down the function some just let me know, I'm still fidding with it and construction queues in general.
Re: What to improve next on Unnamed's Stock AI
Unnamed,
The support for that already is in my scripts, all that is needed is that before calling the script to add vehicles with a queue of 0 your process the conditions and situation for the planet / queue and call it with the planets queue.
I just never added any specific conditions to check yet, but you can see the setup where I have it place Star bases and Weapon platforms on planets that need them (and more where the planet is under threat) and had it skip straight to "Fill queue" mode of filling all remaining queues with ships first then units (hence the low priority values)
If it's about keeping the seperate spending values, such as 50% to ships and 30% to planets all that would be needed is just creating seperate resource values to check.
The script itself is just a framework to hold the logic, anything situational based can be added straight in using priority values and condition checks as I intended

Re: What to improve next on Unnamed's Stock AI
Great work thus far, I look forward to playing a intelligent AI..for once 
- United we stand, divided we fall...but dam it you touch my cheez its and not even god will be able to save you!

Re: What to improve next on Unnamed's Stock AI
Not sure if you've heard about this bug yet, im guessing its quite likely as its affecting all my large/huge planets (havnt anything smaller atm). no matter what i try (turning off relevent ministers), the game scraps a random facility when ever i near the limit of my planet, say with my homeworld, i have a limit of 22 facilities, i have 19 built, then the next turn the ai scraps one randomly... the last straw for me was when my only organics farm was scraped.
oh and this is with the latest version of the AI mod, no other changes. it even affects domed colonys.
_____________________
Grand Lord Exelsiar of the Azra-Dun Commonwealth, at your service
Re: What to improve next on Unnamed's Stock AI
Might be related to the check for space ports, do you use the natural merchant ability? cause stock often scraps a facility randomly to make room for a spaceport then never builds one because it's not needed by a natural merchant.
Might have been fixed in stock or the mod, but I fixed it for myself long time ago so not sure, but sounds very familiar to what was happening, at some point the check for the natural merchant ability got removed

Re: What to improve next on Unnamed's Stock AI
aye it does sound like that, Unnamed's sent me a message and i've since emailed him a saved game. I was gonna try a bit more testing after speaking with him as i gained a few minor ideas that are unlikely to work, but alas the game crashed on this laptop, it would appear ive found the limit to what my lappy can do 
_____________________
Grand Lord Exelsiar of the Azra-Dun Commonwealth, at your service

Re: What to improve next on Unnamed's Stock AI
Better and faster way to find the limits of a laptop:
Run the official Second Life client:
How many FPS and how long before it crashes?
If your answers are "over 15" and "at least three hours" then you've got an amazing laptop.
Re: What to improve next on Unnamed's Stock AI
use a CPU speed controller to control the CPU down to reduce heat generation. I use Notebook Hardware control. see:
http://www.pbus-167.com/
It also increases your battery life 
Re: What to improve next on Unnamed's Stock AI
Unnamed I was going to add that in at a later time, it should just be a simple matter of scanning the cargo of the vessel, finding the population in it and comparing:
Sys_Get_Space_Object_Cargo_Item_Population_Atmosphere_Breathed vs Sys_Get_Planet_Atmosphere_Type
of each planet in the colony list rather then calling Is_Planet_Breathable
To check for ruins I had to actually mod AI_Lists to double compare the planet lists, since there is no function to check if a planet has ruins, you can find the code I used to generate the list in that file.
Still tweaking the constuction script, have this nasty overspending bug that I'm trying to locate but it's illuding me, but I've gone and implemented the script for placing res_mod facilities on planets that are already established prior to their invention, as well as the eventual conversion of organic and rad facilities to monolith facilities in the end game.
debating weather to implement a script to convert research plenets to resource planets when low, and resourse to research when producing too much, but haven't decided on the benifits of it, and it's still on my todo list.
Right now my focus is still on trying to get the AI's research to be compeditive, it still only averages 10k research per game turn: IE 50,000 at 50 turns, 100,000 at 100 turns
It's a monster of a expansion though, it will swarm the entire quadrant and by turn 100 will have all 3 colony techs if not sooner. Once I can get the spending bug squished it should be highly compeditive
Re: What to improve next on Unnamed's Stock AI
I had that overspending issue in my AI until this latest version.
It got fixed as a side effect of the scripts I made to let each empire compare its relative economic strengths. This gave me a global variable for its net resouce income. I have it check this value, and switch gears from military buildup to economic expansion as that gets low.
"Only by being constantly at war with evil in all things, including yourself, can you truely know peace."
Download my mod here: GidMod

Re: What to improve next on Unnamed's Stock AI
i appreaciate the help with the laptop speed guys, but i doubt i'd get second life to run 
cheers for the fix unnammed, i'll test it when i get home ^_^
_____________________
Grand Lord Exelsiar of the Azra-Dun Commonwealth, at your service
Re: What to improve next on Unnamed's Stock AI
Exelsiar:
For a laptop try setting the video usage memory setting from Auto to Safe, that's what I use on my laptop and it makes a huge difference. The other thing I would suggest if you haven't is update directX to the latest version and your video card drivers.
I see no reason why it shouldn't play on a laptop unless it's really really ancient 

Re: What to improve next on Unnamed's Stock AI
its a dell T20, so ya quite ancient, thankfully not as old as the laptop i 'generously' gave to my brother 
as for the memory thing i'll give it a shot, not sure if i've already done it tho
_____________________
Grand Lord Exelsiar of the Azra-Dun Commonwealth, at your service




Re: What to improve next on Unnamed's Stock AI
Fleet composition and resupply IMO is very important.
Intel however I find very uninteresting in SEV and I play mostly with intel disabled.
Attacking more frequently and continuously would be another important point for the AI in my experience.
Scrapping non breathers on colonies with mixed population (as Balance Mod does now) seems to be extremely helpful for the AI.