AI wishlist

[Foogsie]

Example: My pathetic 10 regiments are watching 40 enemy regiments perish in an adjacent, scorched province. Logic dictates that the enemy home in and crush me and then disperse throughout the immediate vicinity to take my land.

Although it should certainly get the heck out of scorched land, I'd try to avoid having it simply run deeper into enemy territory, even if it sees a smaller stack. They'll reinforce slowly and have no where good to retreat to should they need if they run into enemy territory. Not to mention that "pathetic 10 regiment" can (and in my case, usually is) a lure to simply have 3-4 more stacks pounce into the fight a day after the enemy arrives and eats terrain penalties.

Two biggest improvements for army AI would be simply get out of scorched provinces, and avoid doom stacks to begin with. Saw an 80 regiment Great Britain stack in my Austria game. Looks scary, but simply avoiding it and letting it chase things around while I ran in behind to take stuff caused pretty severe war exhaustion. The bigger issue for the AI though is on the strategic level. This comes to three main issues which have already been stated really.

A: It likes to leave it's home lightly defended while it conquers worthless desert lands (Castille being the biggest offender, but far from the only one).
B: War exhaustion, between doomstacking and war taxes, the AI is a master of ramping up obscene war exhaustion for even the smallest of wars. Combined with A, their nation tends to get smacked around by rebels.
C: Just give up already! The AI really seems to enjoy fighting to the very last man, even if all you want is one province. I've seen the AI refuse to cede even a single province at 30-40% warscore and held out until they hit bankruptcy on top of their maxed war exhaustion. This is especially noteworthy in cases where the AI feels you can't conquer them, but has no way to re-secure much of their lost land (Ottomans split by a blockaded straight, colonial powers, a Great Britain with many mainland holdings). Even outside of those though, I've seen some nations such as France hold out from giving a single province until 80% war score or more in some more extreme cases. This creates a massive dogpile effect and pretty much destroys the nation for the rest of the game.

 

[unmerged(199227)]

Although it should certainly get the heck out of scorched land, I'd try to avoid having it simply run deeper into enemy territory, even if it sees a smaller stack.

Perhaps Land battles should be fought with attrition being supplied from front lines? I've never been terribly fond of the Mongol-type land tactics the AI is prone to.

 

[Trin Tragula]

When a province is under siege, you should be able to buy another 1k garrison strength at the cost of population (limited to larger cities of course) Possibly repair breaches that way as well. No reason at ALL Constantinople should fall to 2000 infantry with it's 200k population, only 1% of which is defending :wacko:

Barring this it'd be nice to experiment with city_population increasing local defensiveness. You're right that a big city should be harder to seige...

This was the case in the real world at that time, and it still is a problem today

No it wasn't. Not to this extent. Atleast not in this era. You might've had a point during the middle ages when there was still crusades.
Castille will dow and take provinces in the middle east before trying to absorb Navarra. The ottoman Empire will prefer invading Italy over Egypt. China will prefer to expand into siberia rather than subdue korea or south east asia. Etc, etc. That's just wierd.
Obviously a religious enemy should carry some weight. But given the choice the AI will fight a religious enemy a lot more than it's actual rivals that hold land that might actually be valuable to them.
This isn't due to religious CB's either as I tried removing them and this had no visible effect at all on who the nations chose to fight when given plenty of both kind of targets.
Also outside of europe it makes even less sense. In my indian example muslims and hindus frequently allied against common enemies in real life and there was never such a thing as a religious coalition during the entire timeframe of the game.
It'd actually make more sense to have members of the same religious groups war more than against heathens. Shiites and Sunnis fought many bitter wars after 1501 as did lutherans, calvinists and catholics, and so on.

 

[Malurous]

Kapt Torbjorn and The_Tom already touched this, but I think that the single most important thing to improve is making the AI break their doomstacks when they aren't needed. It's ridiculous that you can siege half the territory of a major, including the capital (therefore allowing peace gains) while they sit on one province with a larger army than you can possibly build. Leave a siege detachment and go for the next target.

But it also goes the other way - when the AI needs a lot of troops, they should gather them together. No sense in sending a long line of 1-5 regiment armies to die against my 15 regiments when each of those armies could wait on the border (preferably in different provinces when safe) and then come in with 30 k. So really, what is needed is a better recognition of what army size is required for which task. That would make fighting a major power A LOT more difficult - especially if war taxes are reworked so that the AI WE doesn't skyrocket like it tends to.

Also, the peace acceptance and negotiations need to be reworked. Less weight has to be put on occupying the capital, and more logic used to remove the situation where you need to absolutely destroy an AI nation when all you're looking for is one province (which also often leads to wars between AI nations that result in nothing), but also the situation where you can grab provinces off a much stronger opponent who would have no trouble crushing you if it just took the time. I can think of a couple of rules of thumb that the AI should ALWAYS check in these situations (leading to less total destruction but more wars where something happens):

1) Has their enemy already crushed most of the AI's armies without any serious losses (these are in the war history so they are probably checkable)? If so, try to make peace and accept fair peace offers.

2) Does the enemy have so many sieges running that rebuilding their army - maybe relative to force limit - is unrealistic in any sensible time, taking the slower recruitment speed from WE into account? If so, strive to make peace and accept fair peace offers. This has the added bonus of reducing the 1 k pop-ups.

3) Does the AI have a much larger force than its enemies, and can reach them? If so, don't accept peace offers that demand provinces. Coupled with army size recognition, this would render the "spread out for sieges before the AI armies walk over" strategy obsolete.

All of these should of course also take alliances to account, to some extent.

 

[HNT]

The AI tends to use mega stacks of 60 k troops or so and use it e.g. for a siege lasting a year. This gives loads of meaningless attrition.

 

[anomanderus]

This was the case in the real world at that time, and it still is a problem today

So basically you're saying that the American ai should have gone for canada, not Iraq?

 

[anomanderus]

Kapt Torbjorn and The_Tom already touched this, but I think that the single most important thing to improve is making the AI break their doomstacks when they aren't needed. It's ridiculous that you can siege half the territory of a major, including the capital (therefore allowing peace gains) while they sit on one province with a larger army than you can possibly build. Leave a siege detachment and go for the next target.

But it also goes the other way - when the AI needs a lot of troops, they should gather them together. No sense in sending a long line of 1-5 regiment armies to die against my 15 regiments when each of those armies could wait on the border (preferably in different provinces when safe) and then come in with 30 k. So really, what is needed is a better recognition of what army size is required for which task. That would make fighting a major power A LOT more difficult - especially if war taxes are reworked so that the AI WE doesn't skyrocket like it tends to.

Also, the peace acceptance and negotiations need to be reworked. Less weight has to be put on occupying the capital, and more logic used to remove the situation where you need to absolutely destroy an AI nation when all you're looking for is one province (which also often leads to wars between AI nations that result in nothing), but also the situation where you can grab provinces off a much stronger opponent who would have no trouble crushing you if it just took the time. I can think of a couple of rules of thumb that the AI should ALWAYS check in these situations (leading to less total destruction but more wars where something happens):

1) Has their enemy already crushed most of the AI's armies without any serious losses (these are in the war history so they are probably checkable)? If so, try to make peace and accept fair peace offers.

2) Does the enemy have so many sieges running that rebuilding their army - maybe relative to force limit - is unrealistic in any sensible time, taking the slower recruitment speed from WE into account? If so, strive to make peace and accept fair peace offers. This has the added bonus of reducing the 1 k pop-ups.

3) Does the AI have a much larger force than its enemies, and can reach them? If so, don't accept peace offers that demand provinces. Coupled with army size recognition, this would render the "spread out for sieges before the AI armies walk over" strategy obsolete.

All of these should of course also take alliances to account, to some extent.

I feel like the last part about the ai recognizing that it has a bigger army than it's enemy and can reach them will lead to ai ming and ottomans being morons when I have their provinces like macau. Since they don't recognize straits as being water, but an actual land connection. So it would result in wars with those ais dragging on forever and ever.

 

[Malurous]

I feel like the last part about the ai recognizing that it has a bigger army than it's enemy and can reach them will lead to ai ming and ottomans being morons when I have their provinces like macau. Since they don't recognize straits as being water, but an actual land connection. So it would result in wars with those ais dragging on forever and ever.

Well, it shouldn't end up in such situations in the first place. If it stops running around with doomstacks, it won't end up in "Macau" situations anyway. Of course, it would be even better if it knew straits for what they are - both in the sense of avoiding those situations and in the sense of knowing that it can't reach you if they're blocked.

Besides, that's a rare and specific situation where, without any changes, the AI is screwed anyway. So it's hardly a reason to not make improvements elsewhere even if it wasn't fixed.

 

[anomanderus]

Well, it shouldn't end up in such situations in the first place. If it stops running around with doomstacks, it won't end up in "Macau" situations anyway. Of course, it would be even better if it knew straits for what they are - both in the sense of avoiding those situations and in the sense of knowing that it can't reach you if they're blocked.

Besides, that's a rare and specific situation where, without any changes, the AI is screwed anyway. So it's hardly a reason to not make improvements elsewhere even if it wasn't fixed.

Except that ming and ottomans aren't some little unimportant countries, for the early game both of them are major powers.

 

[Malurous]

Except that ming and ottomans aren't some little unimportant countries, for the early game both of them are major powers.

I'm not claiming otherwise. However, my original suggestion has nothing to do with the straits situation (that I agree should be fixed) - the AI is screwed as it stands now.

 

[unmerged(199227)]

my original suggestion has nothing to do with the straits situation.

Would that be the Chunnel conspiracy?

No walking to France - that's :rofl:

If you can't manage boats then never pick England