Improving AI - what are the priorities?

[kyrub]

What's really annoying and has the potential to slow down my advances is AAA combined with A-Complexes. Works best with sensors under the base tile (I do take time to bombard these), but that's probably asking for too much.

A-complexes, excellentd, we'll make them even more.
Sensors: In my opinion it is a design mistake (or possibly an unintentional mistake), that sensor array is not destroyed when city is built. I find the tactics annoying and never use it- though I understand that in MP everybody can use it. Still, does anyone in MP build sensors in other places that in a city square? (What kind of feature is this? It flattens the game.)

I'm not sure if I would recommend Trance as on par with AAA and ECM, though. The order I'd pick is AAA, ECM, then maybe some Trance.

Naturally, that's what intended.
I looked in the code only to see that I have left my work unfinished (noths ago), so the AAA is only half done (as many other notes in AI building). Next time, it will be better!
On the other hand, I tried to strengthen DROPs. No change? Also, artillery is already nerfed.

[TarMinyatur]

Some data from a nearly default game of SMAC(n):
(Accelerated start, Hive removed, Gains given Gatlings)

Faction vs Faction attacks: 100
Hurried attacks: 11
Hurried attacks at -33%: 11
Hurried attacks at -66%: 0
Hurried attacker won the battle: 8 of 11

The Gaians represented 10 of 11 hurried attacks. All of these were done by 5-1-1 infantry vs my (Spartan) 4-2-1 infantry on a rocky road. The other attack was a Peacekeeper scout vs a mindworm on the first turn.

The Gaians did not do a hurried attack from base unless the base had a 1-2-1 defender. Three 5-1-1's in an otherwise empty base seemed happy to sit there. Could be a quirk -- need more data.

I should do a similar test using the default terran.exe for comparison.

[ete]

Excited at the idea of playing with a more tactically competent military AI when I get home! Good to hear hurry attacks are starting to happen. That should make it notably harder to invade well defended AI locations with zero armor rover/infantry forces, maybe even make mixing some armored units in essential .

[Geo]

Sensors: In my opinion it is a design mistake (or possibly an unintentional mistake), that sensor array is not destroyed when city is built. I find the tactics annoying and never use it- though I understand that in MP everybody can use it. Still, does anyone in MP build sensors in other places that in a city square? (What kind of feature is this? It flattens the game.)

I usually construct them at the same time I plant a forest near a base, especially if there's still lost of fungus around. Early native life warning and such. In the beginning of a game (say first 50-100 turns) there's usually not enough formers around to delay a colony pod settling down.

[kyrub]

The Gaians did not do a hurried attack from base unless the base had a 1-2-1 defender. Three 5-1-1's in an otherwise empty base seemed happy to sit there. Could be a quirk -- need more data.

I should do a similar test using the default terran.exe for comparison.

Thanks a lot, Tarminyatur, for extensive testing. Great to see the systematic results. You don't need comparative testing on this.
Not attacking without another defender is a feature.


While working on another batch, last one for a while, I enabled ICS in a test for all faction, and... Morgan dominated over Miriam! I could not believe my eyes, hehe.

[kyrub]

I have analyzed AI crawler behaviour. AI uses all of its crawlers on SP once it starts building it. The crawlers normally search tiles outside of city radius, preferably rocky, convoying but 1 mineral. In late game, often before they reach it, they are eaten by worms, or redirected to SPs, not creating a single mineral.

This behaviour was changed, cralwers simply search for closest non-worked forest or mine tile. I recommended up to 3 crawlers per base. Crawlers will not be built during wars. SP focus stays, that is probably welcome.

[kyrub]

Well, that was it. It seems you could have answered in one sentence, to make it short.


As for bunkers, transports and encryption defenders, I asked because there is no way known to me (so far) to introduce necessary AI changes, to tackle
- AI defending roads and tubes at crucial points
- better and dangerous AI invasions
- AI defending against probes (what you suggest is easy on paper, but due to the architecture of AI movement, nearly impossible)

So the options would have sidestepped the trouble. Similar with Cloaked, but I think I'll get to it.
Also, bunkers could stay as an AI improvement with extra option, but without it, I will simply erase them, because they are a farce (but this was the lowest priority).

[Yitzi]

- AI defending roads and tubes at crucial points

Which is the hard part, teaching it to identify crucial points, or getting it to defend them?

- better and dangerous AI invasions

So there's no way to have it hold off on the invasion until it has an advantage?

- AI defending against probes (what you suggest is easy on paper, but due to the architecture of AI movement, nearly impossible)

What is the architecture of AI movement?

[ete]

hm, that explains a lot. I'm guessing the designers had issues with crawlers clogging up squares near bases (a big deal with the AI's old terrible forming) and sent them far away, which made them not very useful, so they made the AI not use them much. I'd recommend something like a sliding target number of crawlers per base, so as the base gets bigger it wants more crawlers. 3 crawlers per base is good early, but in mid/late it should have a lot more and this will help with SPs too. Hopefully the whole move all the crawlers to SP base thing will mean crawlers are rearranged often enough to not stifle development too badly, but if testing shows otherwise maybe something like if currently crawling and mission year is a multiple of 20 move to a new target square further from base would help (and allow AI to safely build lots of crawlers)?

For which resources to go for: bonus resources are notable. Crawlers are okay on forests and pretty good on rocky/road/mines, but an energy bonus with a solar or nutrient with farm also make excellent choices.

Do terraformers terraform outside city limits much? If there's spare former-turns putting r/r/ms on nearish squares for crawlers is good, though unless they're really serious about mass-producing formers there's probably not much spare.

Oh, and not crawlers in war? Does this include the kind of war where you declare war on someone on another continent who you're not actually fighting? Because I think going into war-mode for this rather than just building normally would be a mistake.

[Geo]

Oh, and not crawlers in war? Does this include the kind of war where you declare war on someone on another continent who you're not actually fighting? Because I think going into war-mode for this rather than just building normally would be a mistake.

Yeah. I think its a better idea to only let the military weaker faction forego building crawlers. In SMAC, basically all factions have an opponent with which an almost perpetual state of vendetta exists.

[kyrub]

Which is the hard part, teaching it to identify crucial points, or getting it to defend them?

I cannot see either solved, probably bunker defending can be done, but that is not easy at all and it would need many well thought exceptions (when the chips are down, better defend your base). Getting AI to build less roads is at least thinkable (and desirable). Still, too little.
 

So there's no way to have it hold off on the invasion until it has an advantage?

Oh there is. Just where the "there" is... If I knew the solution, if I saw the place with the decision, one bit of it, I wouldn't ask, be sure about it.

What is the architecture of AI movement?

Infowar units are treated separately, there are too many instances. Defenders are totally separate again; I cannot easily explain the AI that this defender "cannot defend", in fact, so the base needs reinforcing. Every other unit other than defender moves out of base. There are emergency defenders, maybe I could try to squeeze probes inside that code, but then I'd lost emergency counter-attacking from the base, which I won't do.

I don't say it is impossible, just the function is so complex that I have very little idea what to do. I still have no clue how to stop Infowar units from running close to other bases. Maybe, after that, I can stop some of them from running out of the base, if there are no probes inside. Maybe.

[Yitzi]

Which is the hard part, teaching it to identify crucial points, or getting it to defend them?

I cannot see either solved, probably bunker defending can be done, but that is not easy at all and it would need many well thought exceptions (when the chips are down, better defend your base). Getting AI to build less roads is at least thinkable (and desirable). Still, too little.
 

So there's no way to have it hold off on the invasion until it has an advantage?

Oh there is. Just where the "there" is... If I knew the solution, if I saw the place with the decision, one bit of it, I wouldn't ask, be sure about it.

What is the architecture of AI movement?

Infowar units are treated separately, there are too many instances. Defenders are totally separate again; I cannot easily explain the AI that this defender "cannot defend", in fact, so the base needs reinforcing. Every other unit other than defender moves out of base. There are emergency defenders, maybe I could try to squeeze probes inside that code, but then I'd lost emergency counter-attacking from the base, which I won't do.

So it sounds like the issue isn't one of possibility, but rather technical feasibility, i.e. it'd be too much work to be worth it.  Which is probably the most common sort of issue in assembly; very little is actually impossible there.

[kyrub]

This one would actually be really interesting, albeit potentially game breaking, which means it would fit better into a revamped Warfare mod rather than a general purpose mod. But it would be a good balance to the ubiquity of air combat.

That was my thought. I understand it may be seen questionable.
The thread title asked "what would you like", so I went for it. 

@Yitzi, it's as you say, thanks for understanding. Not impossible, but lot of work, I tend to take chunks I can handle due to RL.  (See the last bit I edited in previous post. Maybe the probes are possible, I realized while writing the response.)

[ete]

hm, I think at least two factors are more important than who has a weaker military in determining whether to go into defensive war-mode:
1. does the foe have significant forces near your bases
2. have you lost units or bases recently in battles

[kyrub]

I'd recommend something like a sliding target number of crawlers per base, so as the base gets bigger it wants more crawlers... Hopefully the whole move all the crawlers to SP base thing will mean crawlers are rearranged often enough to not stifle development too badly.

Nice. First we see how current version works, ok? My experience tells me that to move step by step before making things complex, is better.

Bonus resources are in, I forgot to mention them.

Do terraformers terraform outside city limits much?

Nope, and I'd rather not overcomplicate here.

Oh, and not crawlers in war? Does this include the kind of war where you declare war on someone on another continent who you're not actually fighting?

Not exactly sure, but I think the continent flag must be set on "offense" or "defense". So probably not, must be tested though.