Reorganized some information in the primary post to make it more visible (eg. the AI's "tech roulette" is now fully detailed under the Research category instead of hidden away inside the bullet point for the priority parameter).
First the problem. I have a generic faction AI that loves to keep its capitalships guarding its homeworld for the entire first part of the game. This is extremely weird as I've heard of no one else having this problem. Eventually after the AI builds several capitalships it can get unstuck and will start its first cap orders, but by then it is usually at an extreme disadvantage already.
Strange. However, since I do not know precisely when the AI ticks over from the faction-specific one to the generic faction one, there's not much I can do to pinpoint possible causes. Does it still do the same thing if the capital ship's roleType is set to "COLONY" as opposed to "Invalid"?
The second problem I had with this AI is that I noticed in the dev.exe it got stuck in a black market buy loop. Basically it was using all the credits it could get to buy metal from the black market, even when it had 60k metal and 30k crystal. Eventually around when it got 100k metal it stopped this behavior and actually started to play really well as it sold its metal reserves to buy tons of ships and research, in a real game it would long since be dead. The problem seems to have been corrected by adding a BlackMarketBuyAdjustment to an AI only research it gets at the beginning of the game that makes the by and sell price on the market equal. Even for the non-generic AIs I think this might be a useful way to boost performance, since the AI often has issues using the market correctly, and it would help a lot in situations where the AI is not able to get much of a particular resource.
I haven't actually analyzed how the AI handles the black market. Thanks for pointing this out, I'll see what else I can find out on my own.
To add to Goa's post, putting in a HyperspaceBetweenSystemSpeedAdjustment modifier will encourage the AI to attack other star systems. With SoGE's heavily modified AI, there have been Easy AIs that have done this ~2 hours into a match. Further, giving the AI ExperienceConstantGainRate and ExperienceConstantGainLevelCap modifiers that make capital ships be insta-level 3 upon construction will get the AI to build capital ships en-mass, as it values levels. Another addition that gets the AI much more competent is giving it PlanetSlotsCivilianIncrease modifiers, in increments of 4, as this allows it to build more structures and increases efficiency per planet. This effect is particularly noticeable in the game expansion.
Yes, I think I discussed the hyperspace stuff before: instantaneous interstellar travel will give the AI more flexibility when handling ships, but it does not affect the AI's target designation. Once the first steps of an expansion are achieved, the AI will handle being in multiple solar systems much better, since its fleets spend less time waiting to arrive at their destination, unable to accept new orders. However, the AI will still be just as slow with its first steps as before, at least according to my observations. The main thing I still point to is wormholes: wormhole travel is instantaneous in vanilla Sins, yet the AI is still very hesitant to use them unless they are the shortest way of getting to a gravity well the AI has already designated as its attack/colonization target.
I really need to look into the AI's handling of capital ships, you guys say the darndest things about them, and the only things I really looked into were ability usage, capital ship composition, and combat strength evaluation (for aiRetreatThreshold).
Increasing logistical slots would make any player a lot more capable, not just AIs. On that note however, I'll double check to see if the AI tends to ignore logistical slot upgrades the way they ignore population upgrades. My bet is that they only queue up a logistical slot upgrade if they want to build something but are out of logistical slots, but this means they have to wait an extra few minutes every time (so they don't queue up the upgrade when they run out, they queue it up when they already want to build something but have run out).
Quick question: does this mod need to be updated in any way for the 1.82 release (1.82.5006) as opposed to the BETA (1.82.4976)?
Most likely yes; BETA compatibility was simply a case of performing the edits on a newer Gameplay.constants file, but since I'm guessing the 1.82 release will encompass a lot more changes, there are a lot more files I'll have to update as a result. If the .exe was updated, I'll also need to run a lot more test games to see if the AI was modified in any way (doubtful, but you can never be too sure). I just wasn't aware the release version came out, since Steam still said my version was BETA until I launched the main .exe.
EDIT: Nope, no update needed, though I'll still rename the files and remove the 1.80 version to avoid confusion. The only difference between the BETA and the release version are an updated Galaxy Forge, an updated Particle Forge, and different particles on Strip to the Core. The .exe's were still changed though, I'll either have to run a lot of test games to see if the AI was altered... or open them up in a hex editor first and see if enough bytes were altered compared to a 1.80 .exe to know if I should bother with test games (unless the older .exe's are available somehow, I'm probably going to need to borrow a "quasi-legal" version for this).
EDIT #2: Hex comparison between newer 1.82 exe and an older 1.80 exe shows quite a lot of differences, though most are sections that were replaced with something smaller, causing the remaining bytes to shift up a few addresses. I'll need to run more test games nevertheless.