Short Term Fix for Conquest AI's lack of defense

[DioAurion]

I was playing some wild mana tonight and seeing the usual AI doing their city trading back and forth and thought what if there was a quick way to make these conquesting civs far more dangerous while on the rampage? So I began to wonder if it would be possible to code AI in conquest mode to burn a city upon capture if the city had been captured and lost once before? I feel this would make players far more weary and more focused in putting up a solid defense if their hard work was on the line in defending these cities from total destruction.

So that's my short term solution. Any modders think this type of behavior is possible to code?

 

[Fafnir13]

Eh, best to leave it alone until a real fix can be implemented. Quick fixes can cause more problems then they solve and aren't worth the maintenance.

 

[[to_xp]Gekko]

I guess having them raze cities they know they won't be able to keep, just like a human would, would be great. of course, it's probably not as easy to tell the AI wether or not they're going to be able to keep it

 

[Sephi]

What "lack of defense" do you mean DioAurion?

 

[DioAurion]

I'll upload my scenario when I get home (standard sized Pangaea, 13 civs, monarch difficulty) . But currently Hyborem and the Khazad are fighting wars (efficiently I might add) on multiple fronts. However they do not have the spare military to properly defend these cities so there is a constant series of city flipping going on, causing the stack of dooms to have to keep u-turning and going back over and over again.

I've read plenty from other players to see that this is being used as an easy way for players to keep overwhelming stacks of doom busy by constantly sneaking speedy units behind the stacks and recapturing cities. If there was no longer a city there to be recaptured, the AI wouldn't constantly be u-turning to return to these cities. That way their large investment in that stack would more efficiently be focused on causing further widespread damage to their opponent's economy.

 

[Fafnir13]

If the AI could be taught to peal off a unit or two from their SoD, the problem would be taken care of. Let's say certain units are classified as city defenders (archers and the like). SoD's would want a threshold number of defender units before trying to take a city. Once taken, X number of defenders (based on how threatened the city feels) are left to defend. The SoD checks to see if they have the threshold number of city defenders in the city. If not, combat units (the weaker ones) are reasigned to defenders. At that point, the usual checks could trigger on the trimmed down SoD as to whether it marches or waits for reinforcements.
just a rough outline made without any idea of how the code itself works.

 

[DioAurion]

Yup that would be the long-term fix. The stacks already have large amounts of archers inevitably in them. Just have to teach them to reassign their AI upon taking a city.

 

[Sephi]

AI currently leaves two units (archer preferred) as defense in the city. Two units will never stop city switching though as they can't defend against a stack of doom.

 

[DioAurion]

I haven't seen that in practice. I've seen them be much quicker to get new defenders to newly captured cities than in the past as of patch 7.2. However the stack of doom will capture and move out of the city in the same turn in the game I have going.



Edit: Uploaded my game. 4.2 with a relatively up to date hot fix. Some back story. The town SW of Dis was in the hands of my ally the banner but was forcefully conquered by the infernals. The infernals were out of the city by the end of their turn and marching on my mercurian capital. I quickly rushed over a group of angels and took the undefended city. Immediately upon doing so, the infernals made a u-turn (even though my capital was less defended at the time than the banner city) and rushed back to reclaim this city again. I set up a trap expecting this and wiped them out in the pass with 2 combined stacks of mine.

My distant ally, the grigori has been losing cities constantly. Cities that are usually not defended or defended by a single unit at best (which was rushed there by our enemies ~10 turns after the city was captured). Yet the grigori remain complete pacifists and do not even bother rushing to reclaim these fallen cities which they are entirely capable of capturing with their mythril wielding adventurer dragon slayers.

 

[Lone Wolf]

Why doesn't this happen in vanilla BTS? In it, AI-AI wars are either one AI slowly plowing though other or a stalemate with few cities changing hands. Rarely I see city trading which you describe.

That proves to me that Sephi's AI is worse in some respects then BTS AI.

 

[Sephi]

read this thread http://forums.civfanatics.com/forumdisplay.php?f=245 and you might get an idea why the active defense from Wildmana AI is better than anything the BTS AI can pull off.

in general I find your post about the AI very annoying. You call my mod bugged even though you haven't played it, base your proves on a single critisim post.

Either bring it on and actually play wildmana and post screenshots/saves to proof your point or stop talking about things you have no idea of. Ofcourse this is an open board so you can write whatever you want but if anyone complains about things he has no idea of we wouldn't get anywhere.

 

[Lone Wolf]

why the active defense from Wildmana AI is better than anything the BTS AI can pull off.

Cases of massive city trading in wars indicate that something's wrong with active defense to me. BTS has no such phenomena, which indicates that BTS AI is better in that regard.

I expected to be forwarded to a post with an example of a brilliant active defense by Wild Mana AI, not BetterAI subforum (or even the 1game4suicides thread).

in general I find your post about the AI very annoying. You call my mod bugged even though you haven't played it, base your proves on a single critisim post.

Artifical Stupidity =/= bugged. I'm basing my posts on other people's evidence, as well as on your self-congratulatory attitude (apparently, an AI that for all intents and purposes doesn't build a navy is "better" at naval warfare then an AI that uses the navy half-competently late-game, because Sephi the Magnificient coded the former AI) and desperate references to BTS AI each time someone points out a problem.

 

[Valkrionn]

Personally, I find it more annoying when people are confrontational about their critisicm (This mod sucks! You suck!) instead of constructive (This is annoying me, and here's why), but yes, saying you dislike something without actually trying it, or basing your dislike on a preliminary version (Version merged to FfH is very different from the current Wild Mana) is annoying.

 

[Lone Wolf]

or basing your dislike on a preliminary version (Version merged to FfH is very different from the current Wild Mana) is annoying.

I'm basing my dislike on the attitude of the modmaker. When complaints are made with "well, BTS AI is not perfect too!", something's wrong in the core. When legitimately pointing out flaws is met with all efforts to deny these flaws (when AI building warriors non-stop in Always War was first pointed out to Sephi in one of the FfH scenario threads, his response was "so what, it's bad to make an AI for a specific scenario", only when the complaints grew louder he actually fixed it, when people complained about lack of a navy, he said that his 5.0 AI is better at navy then the default one, which was a lie, plain and simple), something's wrong at the core.

 

[Sephi]

Cases of massive city trading in wars indicate that something's wrong with active defense to me.

as I said before bring it on and post some screenshots/savegames where this massive city trading you talk about can be seen.

 

[Lone Wolf]

as I said before bring it on and post some screenshots/savegames where this massive city trading you talk about can be seen.

DioAurion called it "usual", and I trust him/her more then you. Your behaviour every time someone points a problem to you suggests that something's rotten in the state of Denmark to me.

I get the impression that if your AI disbanded their settlers in the first turn, you'd said "Not too fast. It may be rational behaviour. Sometimes, some units need to be disbanded. Besides, look how bad BTS AI is at with picking their city cites with their settlers! While I grant that the settler behaviour in my modmod may be occasionally subpar, in BTS it's even worse".

Honestly, it's not much worse then statements like "5.0 navy is better then what was before", "not killing the barbarians who walk around and pillage your improvements, though you can do so, is a legitimate "xp farming" strategy that is superior to what was before", "warrior spam non-stop is of no concern for me, because it had been pointed out in a scenario thread".

Using your statements, you can make a nice Attacko article.

 

[Sephi]

I'm basing my dislike on the attitude of the modmaker. When complaints are made with "well, BTS AI is not perfect too!", something's wrong in the core. When legitimately pointing out flaws is met with all efforts to deny these flaws (when AI building warriors non-stop in Always War was first pointed out to Sephi in one of the FfH scenario threads, his response was "so what, it's bad to make an AI for a specific scenario", only when the complaints grew louder he actually fixed it, when people complained about lack of a navy, he said that his 5.0 AI is better at navy then the default one, which was a lie, plain and simple), something's wrong at the core.

you can dislike my mod all you want. It's not nice that you accuse me of lying, but this is an internet forum so behavior like yours is something I have to expect. But it's good to know that you critize wildmana because you don't like me or my attitude or whatever.

also, I didn't fix anything because complaints grew louder. Your whole story is very absurd. Scenarios weren't even playable in Wildmana that time. I uploaded two python files with tweaks so Kael could take a look at them and since they also helped at the scenarios I made a post to link to them.

 

[[to_xp]Gekko]

I wish for this pointless arguing to stop and for the discussion to go back to post 9 by DioAurion. Lone Wolf, please take a look at that post and confront it to your posts. the former is helpful and can make the modmod better, yours are useless. you have the right to say what you want and dislike what you want, but please do realize that you're not being constructive at all, you're just using someone else's post as a means to show hostility towards a modder you dislike. or at least, that's how it looks like judging by these posts.

 

[DioAurion]

AI currently leaves two units (archer preferred) as defense in the city. Two units will never stop city switching though as they can't defend against a stack of doom.

I think I misunderstood this post earlier. The AI seems to leave substantially stronger defenses in the cities that they build up come mid-game. I actually have to plan out a warband ahead of time to take such cities, greatly slowing my conquest progress even with the Mercurians.

I think the war AI is quite incredible and just needs a small boost to how they handle freshly conquered/ freshly lost cities to complete a near full AI improvement package.

One thought that comes directly to mind is that I noticed the conquest AI civs tend to be dragging around these near death catapults late in their campaigns that never have time to heal. Perhaps if the SoD could be taught to stay within a conquered city until the initial revolt ends, you could kill 2 birds with one stone. The stack would full heal. The catapults would be repaired. Quick city grabs by human players would be prevented. And the AI would have time to rush new defenders to the city.

 

[Lone Wolf]

It's not nice that you accuse me of lying, but this is an internet forum so behavior like yours is something I have to expect.

"5.0 AI is better then the pre-merge one at navies" is a blatantly false statement. After that was pointed out to you you said that navies aren't important, so you don't care about them.

The evidence is clear.

also, I didn't fix anything because complaints grew louder. Your whole story is very absurd. Scenarios weren't even playable in Wildmana that time.

You deliberately make incorrect statements for me to go berserk and post something stupid in capslock mode, right? Fact - after the 5.0 merge, scenarios were definitely playable (as in "load-able").

I uploaded two python files with tweaks so Kael could take a look at them and since they also helped at the scenarios I made a post to link to them.

Good for you (and for FfH). But you've done a nice job of completely ignoring and dismissing all complaints before that.

to go back to post 9 by DioAurion.

I'll deal with it myself, Sephi-style:

was forcefully conquered by the infernals.

See! Wild Mana AI conquers cities. It's waaay better then BTS AI, that had trouble efficiently conquering cities.

Immediately upon doing so, the infernals made a u-turn (even though my capital was less defended at the time than the banner city) and rushed back to reclaim this city again.

Sometimes, it's better to attack a more defended city, sometimes, it's better to attack an ordinary city, not a capital. I don't see a problem here.

Yet the grigori remain complete pacifists and do not even bother rushing to reclaim these fallen cities which they are entirely capable of capturing with their platinum wielding adventurer dragon slayers.

I don't see any problem here. Look at any succession game to see how BTS AI manages its stacks. Face it: the AI has trouble counterattacking its cities, so I gave the AI a very low priority to do that. How is giving a lower priority to something the AI doesn't understand bad?

It seems a lot of complaints around here are just "the AI plays different than me, it must be stupid". BTS AI was really bad at counter-attacking, mindlessly throwing kitchen sink at your stack, and after it was defeated, it did nothing. In your save, I see evidence of how brilliant my AI is - it wisely considers the benefits and the negatives of counter-attacking before deciding to counter-attack. Remember: AI uses a different logic than a Human.
That's true there are some issues which hopefully'll get fixed over time. But what you see is a massive improvement over BTS AI.

I just prefere an AI that uses all the things that make FFH unique, like spells or heroes (which is why I don't care about counterattacking cities that much). If you want to play a game of counterattacking cities, just load classical BTS. Sheesh.

And now, I'm really curious, what will Sephi really write. Will my imitation have something in common with his actual answer?

but please do realize that you're not being constructive at all.

Today is my Trolling Day. Though admit, I do have a point...

When Kael will merge the 7.x AI, I promise to post a thread with screenshots of the new AI. Depending on my impression on the new AI, that thread will be either very complimentary, or it will be written Attacko-style, with each screenshot labelled as "fact" and words "the evidence is clear" in the end.