Attempting to improve the AI for Proportions 3.05

[PvK]

Range 20 on a starliner is fairly limited, but not useless (also considering the cost and that IIRC you may have no other way to build a ship that can haul a population unit at the point that you can build those starliners) - you'd just need to provide supplies for where you want to use it, which is also an intentional part of Proportions - you're supposed to have logistical problems to solve, and using tankers is a useful option (building cheap ships with a lot of supplies and fleeting them with ships that don't have much range themselves).

I don't expect the AI can build good starliners except by accident, though it should be possible to fiddle with the AI to get it to do so. What I'm thinking of doing is resorting to some AI racial techs to provide a challenging opponent rather than investing the time to get the AI to play by the rules designed for humans.

The number of AI empires isn't limited by the definition of a high number of AI opponents for randomly generated games. Also there's no reason the player can't change that setting if they want a full set of AI opponents. But, ya, I might as well turn up the number on that default setting.

The construction minister, when on, decides which things to build without your orders. The transport minister decides whether to order some ships to move units and population around or not.

[Arralen]

As the forum ate the rather lengthy reply of mine, and I don't have the energy to type all that again, here some more findings from my current game in more or less random order:

Xiati Empire does build Starliners, but has no use for them: On turn 2.6 they're still only at Rock Colonisation 5, putting too many expensive other techs into the research queue; On turn 40, researching Temporal Tech III (400k) while Rock Col 7 isn't even queued)

Cue Cappa does have colonisation tech in turn 40, but only 1 world colonised, 1 coloniser en route to the same system, 1 idle. On that colonised world (large domed), it's building a "Minor Spaceport City" what "only" will take 6.1 years ... instead of a simple Spaceport!

Xi'chung is colonising it's home system, and even builds Starliners - yet 2 of those seem to sit 1 move away from the HW in open space, at least 1 doing nothing... (both didn't have orders, but the minister seemingly moved at least 1)

Ruins: 3 times I "discovered advanced technologie" .. but got nothing .. yeah, I know it's in the readme, but I expected that those "dud" ruins do not generate a missleading message ?!

CSM I : damage resist 30kt // Damage Self-Def Cannons I 25, PD-C 25, Def Can 20; And things are bound to get worse, as CSM II is 80k resarch, PointDefCan II is 160k research ...
Who would design defense cannons which could barely not detroy incoming missiles?
Hmmm, okay, maybe make that question "Do we want that much realism in SE4?" instead ...

[PvK]

Interesting AI test results. Looks like the AI can use both colonizers and starliners, but isn't very coordinated about it in the early game, as you reported. Starliners moving speed one sounds like they are out of supplies.

Ruin messages aren't moddable. Sometimes they have ordinary techs which the game unmoddably randomizes, resulting sometimes in you getting tech levels you already have.

As for capital ship missiles versus point defense, I note your experience with them, but I've used them in competitive PBW games, and there they are balanced pretty much as intended. Yes a single PDC I will not shoot down on CSM I in one shot, but they can in two shots, and CSM I fires only once per three turns and the launcher takes 50 kT ship space and uses supplies, while the PDC I fires once per turn, even during enemy movement so it can get two shots in per missile, and it is only 20 kT. If you have 100 kT of CSM I's, that's two missiles, which facing 100 kT of PDC I is 5 shots at least, maybe 10. There are other factors too of course such as range, whom you're trying to defend, etc. I think it plays out pretty well and if anything, I'd say CSM's tend to lose out against point defense in a single tech versus single tech consideration - of course it becomes a guessing game how much PD you need, etc.

[Arralen]

Originally Posted by PvK

Interesting AI test results.

At least the Xiati need some tweaking of their tech research priorities: Investing 400k points (~40 turns) into Level 3 of their racial tech (Temporal, which does not even give ships hulls etc) while not having colonisation tech available is a no-go.

Looks like the AI can use both colonizers and starliners, but isn't very coordinated about it in the early game, as you reported.

It's only a guess, but to me it looks like that some AIs do not manage to beeline for Colonisation Tech-10. They're building lots of combat ships in the meantime, which seriously hampers their ability to colonise once they finally have the tech needed because of the high upkeep costs.

Starliners moving speed one sounds like they are out of supplies.

Oh, it's not 2 starliners moving 1 step - it's 2 starliners sitting in open space, both fully loaded and amply fueled, but only 1 is moved by the minister on the next turn, while the other just sits there doing nothing (2 steps away what could be the originating planet). Interestingly, both don't have any "standing" orders (like "move to XY, drop pop" or something).

As a player, I never got a minister to even touch a starliner, no matter what I did ...

Ruin messages aren't moddable. Sometimes they have ordinary techs which the game unmoddably randomizes, resulting sometimes in you getting tech levels you already have.

Ok, so as soon as I get the "tech discovered" message, it's a "standard" ruin, not one of your "dud" ones?
Then - how big is the chance that I get a tech I already have - with 3 ruins @ turn 40? Atm., it looks like ruins aren't working at all, but I'll check some more ...

[Arralen]

Ok, obviously there's lot of room for improvement on the AI for Proportions:

I did a test run, all 15 AI in a large galaxy, simultaneous game. I looked into the AIs turns on 2400.1, 2400.5 and 2401.0. I've attached my notes in pdf format.

- lots of AIs do not build proper scouts, but missile attack boats with 1x CSM and 2x std. ion engine: those ships run out of fuel before reaching the next WH ...

- most AI do not manage to research the colonisation tech 10, adding to many expensive techs "in between"; presumably to have some difference between the AI players, some of their research scripts have the same research topics listed in different orders. Sadly, it looks like those lists haven't been adjusted for the vastly different tech costs in Proportions.

Amon'krie are the only ones which (mostly) get things right (colonisation tech 9 on the 10th turn, 4 systems explored), showing that it is possible for the AI to be competitive - after all, the human player only manages to get to col tech 10 in turn 10 because he can rearrange the research queue, what the AI is unable to do.

- Troop transports and cargo tech shouldn't be build/researched before colonisation is possible, as those are basically useless before then.


=> Sadly, it looks like that adjusting the initial research of the AI will make all AI basically act the same during the first 10-15 turns, as there are limited options to choose tech because of prerequisites and costs.
On the other hand, interactions between human player(s) and AIs are very limited during this time frame, too, therefore this shouldn't be too much of an issue...

[PvK]

The AI files were I think ok for Proportions 2.x, but then weren't changed much for 3.x, and were changed in some ways that weren't tested much or at all, so your testing is great to have, thanks!

Agreed about the Xiati starting with 4 years of only temporal research.

The AI not giving orders to some ships sounds like a flaw in the AI that may not have much to do with the mod, or it might just be that the AI doesn't consider a ship with a "mere" one population unit to be a worthy population transport.

It's been years since I had fresh knowledge of what the messages for ruins meant exactly. I don't think it's at all likely that ruins are broken entirely. (We did play several games of 3.x multiplayer.)

Good observation about the 100kT missile boats with range 9.

FYI & IIRC, there is sort of a limit to the control possible over micromanaging what the AI does during the first few turns particularly without impacting what it will do during the rest of the game and with colonies on planets. So for instance it might be ok if it tries building a fire platform on turn 1 on the homeworld, if it means it also builds fire platforms early on colonies, since that's a useful defense to have there.

Also, in Proportions it's far less vital, nor even clearly the best move, to develop scouting and colonizing early on versus military. If an enemy seiges your home planet early on, it may be game over due to unrest. Also colonies are investments which only pay off after considerable time, and which in theory can be captured by invasion (though the AI isn't very good at capturing planets).