AI behaviour

The AI is still more on the placeholder side of things, although they have done a couple passes for it. I've noticed that aggressive builds more ships first and defensive appears to spam planet defenses. I'm not sure the research/economic do anything unique yet.

In RTS games overall, or at least the ones I've played, once a player "get a foothold" it is always straight downhill for the defending player. RTS games have that as a weakness (snowballing).. I'm not sure how the AI would be able to counter that?

Identifying choke points may be a bit difficult, especially if planet orbits is turned on. However, it may be possible to bias the AI to buffing it's defenses over "key" planets first such as homeworlds & planets with either artifacts or specific planet bonuses (inherent research, extra planet slots, or something silly like space ponies).

It'd be cool if they set it up to bias the unit/component compositions by AI behavior types, after they add whichever systems that are still missing from the game? A couple examples of what I'm getting at are:

• Economic TEC > max tradeports, bias towards akkan, starbases have tradeports, builds 1 exotic refinery/planet after tradeports maxed, diplomatically is the easiest to cease fire/ally.
• Defensive TEC > bias towards garrisons, starbases have factories, ships prefer components that improve durability/survivability, maxes starbases in each gravity well, difficulty to ally but easy to cease fire

Getting the AI to mimic top competitive players is a bit of an expectation, but getting it to behave in accordance with it's flavor settings should be viable. Same as for using faction-unique capabilities (It'd also be nice to see Vasari AI jump their starbases around, or to have TEC-loyalist AI build both starbases. Seeing a fully planet-less Vasari Exodus AI would be cool too.). Eventually, it'd be cool to see defensive AI who are allied build starbases at each others worlds as well as build other remote structures at stars or other uncolonizable locations.

If the AI isn't fully opened to us, it'd be cool if we could at least set biases within each player file by AI behavior type. I'm more bothered that the AI tries to play every faction the same (at least it felt like it did in sins:rebellion, for me)

I thank you for your addition to this topic. You are right that simulating a top tier player behaviour is super hard - and I am not expecting that. But if you think about it, the SOASE star system is a simple math problem with graphs, due to the phase lanes. To correct my previous sentiment - I want the AI to identify nodes that have one egress line and a lot of ingress lines (i.e. a choke point) and fortify it accordingly.

But true, there were serious improvements, and we haven't seen the Advent yet (which had the most devastating starbases with meteor shower in SOASE1) - so we cannot expect a full balance without that.

Tec-loyalist build both star bases.

They max the star bases at the edge of the empire and as you advance they build/upgrade. But it's true they seem to choose the upgrades at random. Haven't played with Novalith long enough to see if they build a shield once i start to bombard... let me check at least that.

Every AI seems to have the same flavor at the moment. Not sure about the ones who are eliminated early before i get to conquer them.

They've all stopped building culture, though which is a bit of a downgrade.

Neat, I haven't seen the TEC-loyalists build both starbases yet but I'll see if I can get them set up to do it in my next game. Maybe I need to have my game run longer than 6hrs, or maybe I need to increase the difficulty so that they have enough resources to build them. They seemed to be constantly replacing ships (my last game the loyalists were losing to the rebels while I gobbled planets in unclaimed space). Rebels were offensive AI & loyalists were set to defensive AI.

I've also noticed the no-culture thing to in my most recent games [patch 1.18.14 on test version] now that I think about it.

Yeah, it's strange they leave planets unoccupied. Maybe the bug is that they only get enough planets to build their research then they stop expanding on unoccupied planets.
I also noticed they don't balance well enough exotic factories either. I watched for a long time how a star base had the missiles in que.

Don't get me wrong, the AI is much better, especially early, compared to how it cheated to be difficult - like having 40 Civilian Slots on an asteroid, but there is still improvement needed, especially after the first hour or two.

Excuse me for the slow reply - I had IRL things to deal with.

So, Vasari are not adept at fortifying starbases, unfortunately. And also, they do live some planets unoccupied, especially on the bigger maps. Like next to their home planets. This is definitely a bug.

I would like to see an AI that does not have to cheat to be competitive, just optimise the APM values and follow a proper stragegy - because if I can repel the first 2-3 huge assaults (which I definitely can), then there is no more challenge.

I still haven't encountered a TEC Loyalist with both starbases built yet, but I did see one with some culture generation from the trade port culture research I think. It didn't build any culture centers and the culture spread was a bit fast too just be from lvl4 commerce on its homeworld.

-=-=-=-=-=-=-=-

Don't get me wrong, the AI is much better, especially early, compared to how it cheated to be difficult - like having 40 Civilian Slots on an asteroid, but there is still improvement needed, especially after the first hour or two.

As for AI cheating, I don't mind a little bit if it's flavored. I.E:

• Economic AI has extra logistic slots on high difficulties
• Research AI has faster research speed on high difficulties
• Aggressive AI has larger fleet supply on high difficulties
• Defensive AI has additional defense slots/larger hangar squadrons/extra starbase slot on high difficulties.

I've always been outvoted when it comes to this though...  :/

-=-=-=-=-=-=-=-

I would like to see an AI that does not have to cheat to be competitive, just optimise the APM values and follow a proper stragegy

What we have currently looks like it intends to be a generic/high-level playstyle imitation. Just to play devils advocate though, how would you define proper strategy? Would it be something like having a set of AI scripts broken down such as:

By map size, because the 'proper approach' can vary with proximity to other players:

• Small Map AI
• Medium Map AI
• Large Map AI
• Huge Map AI

or by faction, because each faction has a unique set of tools:

• Exodus
• Alliance
• TEC-L
• TEC-R
• Advent-L
• Advent-R

or by another set of criteria that you had in mind?

A few month  ago AI difficuly was changed.... and it became super booring to play, basically it broke single player   
No matter how many AI I play against and of what difficulty they are, it is literally impossible to lose