Bankruptcy.

Bankruptcy is just about the only thing which pushes the AI's global settings around aggressively. Money production is, for some reason, hugely overrated in power score terms (while money on-hand isn't). So if they ran out of cash for maintenance, hit their spend-til-broke limit, and then had to go 100% econ for a turn or two, that's cause big swings in power score from turn to turn.

Check their econoic power graph; if it's very 'spikey' then they're stuck in a bankruptcy cycle where they run out of cash, go 100% econ, then have enough cash to go back to normal spending for 1-2 turns, bankrupt again, and continue like this until their losses go down. 

I don't know if this has changed for 1.3, but in 1.2 the AI only has an invasion range of about 16 tiles.

Hahahahahahaahahhahahahahahahaahahahahahaha. Good one.

Oh, you're serious?!?

No. No chance. Not in a million years of soaks.

The scripts are timed wrong. It thinks the opening 30 turns are the 'colony rush'. After that, it diversifies a lot and it's colonization rate goes off a cliff. I didn't want to re-write the AI scripts completely for IAB. I had to rewrite them from scratch. The scripts are basically best-guesses at what the AI ought to be doing at turn X. There is an enormous difference between what it should be doing on turn X on a small map to what it should be doing on an insane map. This is completely unaccounted for in the triggers, and so the AI does the same thing regardless. 

As I say, I've not been banging on about the AI scripts being kind of high priority for no good reason.  The enormous difference in scale between a tiny map and an immense or insane one requires the AI to think on entirely different time scales, and there's presently no way to tell it to do so.

It's probably a bit more than 16 tiles, but it's still very short ranged indeed. And I'm not sure if it has been changed for 1.3.

Thankfully, it's easily fixed. In your galcivAIDefs.xml, find <WartimeFocusSectorRadius> and change it to 256. There's a comment noting that it should be multiplied by map size... but this doesn't appear to have been added yet as of 1.2 (or if it has been, it doesn't seem to work). And do not go over 256 - it makes the AI go funny. 256 leads to the AI suddenly becoming quite aggressive at invading, if it has transports.

You can open it with notepad.

Next question: how do I find that file?

you could probably use search to find it but it is in your steam folder, default would be

c:/programfiles(x86)/steam/steamapps/common/GalacticCivilizationIII/data/game

you would copy the file you want to modify(in this case galcivaidefs.xml) into your mod folder which would be

c:/users/(username)/mydocuments/mygames/galciv3/mods/(folder name of your choosing)/game

then open the file with notepad and edit away, make sure use mods is enabled and should load up barring any errors from editing

Thank you both. I have to try it. I will try the "Mod" approach as I assume that won't make a permanent change. I'm pretty discouraged right now. I ran a SOAK game on the map. As I said, Insane, very small clusters, 5 or 6 planets per faction and 48 opponent factions. It ran between 400-500 turns. I came back to the screen to see "Defeat." No problem, I saw that the Dreadlords colonized a dead world in my cluster. They must have taken me out. What bothered me was that ALL 48 other factions still existed. NO one conquered anyone! There were some wars. I checked this from time to time. 

Couple of questions for anyone:

1) Do the Galactic events that create factions always happen to the player? Brain'yacks and Dread Lords BOTH in my cluster, not 1 star away. Each time I've played someone always appears in the dead system near me. 3 times Brainys, once each Pirates and Dread Lords. In GalCiv2, the Dread Lords appeared everywhere. They simply sought out Starbases.

2) Do the AI factions actually take each other out?

3) Is there any difference below Gifted? I had some set at Normal, some at Easy and some at Beginner and STILL no one was taken out. I purposely gave the Drengin 2 opponents set at Beginner and they never seemed to conquer them. The Drengin were set to Normal.

4) Is it simply a case of too much space between factions on an Insane map? Should I drop to Immense?

Thanks to anyone who might offer advice. 

Not familiar with the soak process. How does one initiate it?

Naselus and Moogle, thank you both for your help. I followed both your suggestions and changed the file from 9 to 256. I booted the game with no problem. I will give it a try tonight or this weekend. Perhaps I will try another Soak game. I must admit, it was a bit unnerving for me messing with files for the first time. By the way, the actual file read "GalCiv3AIDefs.xml." I gambled that the 3 was left out by each of you. There was no "GalCivAIDefs.xml."

On a side note: Naselus, I tried your Insane-Abundant mod with no success. I received an "error box," and the game would not boot. I'm using 1.3, so I assume you haven't updated it yet? I read that you were waiting until the official release. I will keep an eye open for that.

Thanks again to both of you.

Would you mind explain me the "invasion range of about 16 tiles".

I understand that if the AI see a planet within 16 tiles of one of is planet, it may try to invade it.

So, if the invasion brig it within 16 tiles of another planet, will it try to invade it too?

I hope !

As far as we can tell, the range is measured from the present location of the transport - not the planets. So if the transport moves to a rally point within 9 tiles or so of an enemy planet, it will attack it. If it's more than 9 tiles away... well, it doesn't seem to do very much at all in that case. 

This is largely work that Vid did for his AI mod. The 256 range check is ideal for larger maps, but may make the AI attack somewhat odd targets on very small maps. SD are aware of the issue - they actually reduced the number from 16 to 9 for 1.2 - and intend to make it multiply my map size. But as I say, if they already added that it doesn't appear to function... and there's been no change to the map size defs xml that would imply the change was ever made.