I would like to share some thoughts on the current military AI behaviour.

To make AI a bit more competitive military wise, it would've been good if:

1. It was fighting with more than one army attacking, defending together.
That's what the player would often do and that's what gives player advantage even if the military rating of the factions is comparable. Almost always it is 2 players armies vs 1 of AI + (may be) militia...

2. Make AI watch military ratings of other factions (particularly one led by player) and when seeing it going over certain threshold make AI build military of it's own. Currently, AI is still not prepared to invasion for both reasons: not combining it's armies and not building enough military.

3. When seeing the direction of player's (other AI's) attack, bring the army to meet the invasion.

4. Some of the faction's military is handled by AI worse than the others. I.e. Forgotten assassins led by AI can't do anything (at all) against Mezari\Vaulters marines led by player or\and another AI in the mid game (2-nd - 3rd era, turn ~30 - 60 fast game). Mostly it happens because these assassins are poorly equipped and low in numbers (see points 1 and 2). Mezari and Vaulters are better handled by AI and more or less competitive. Allayi put up some fight (mostly due to having some numbers), but seekers in AI armies lack advanced equipment...

Thank you for the good game and reading

I would add to this how higher difficulties are handled in this game.



Pretty much, it is economy bonuses to the AI.



But that is not enough, since properly equipped, hero lead army, with trump 3-5 same sized armies that are not lead by hero or equipped properly.



I do not think current AI is capable on handling properly decked out human force, regardless of economy bonuses. Even if their economy is better enough to have more units, or even better equipped units, way too often they will be without hero assigned, making them easy pickings.



Also, due to game high reliance of having one good army group, led by hero, losing such army can prove catastrophic, with no easy way of replacing it.

Actually, I recently started a war with the Wild Walkers, besieging a city of theirs with an 8-unit army + skyfin. They sent 2 8-unit armies to respond, each led by a hero (problem was that both heroes were Allayi. For some reason the AI keeps getting Allayi generals). Interestingly, they could have attacked with one of the armies, but they waited a turn until the 2nd army was in range as well. Not to say that the AI is cleverly opportunistic about army movements (unlike say Nobunaga's Ambition, where these F-ers would leap at the sight of an ill-defended border).



And it was nice to see the AI using 3 unit types (not necessarily the most efficient combination, mind you. They had Dekarians, Eyeless, and Ceratan. They should have had a few Tenei Walkers in the mix, as opposed to 2 supports). The biggest problem by far was how woefully ill-equipped they were (still mostly on tier 3 iron, while I was rocking tier 3 adamantium). My army crushed these 2 + city garrisons without the health of any unit dropping below green.

I'll just to throw one tip:



When in battler, try killing enemy hero as soon as possible. It it is a good army leader, it will greatly decrease effectiveness of enemy army, during same battle.

player1 wrote:
I'll just to throw one tip:



When in battler, try killing enemy hero as soon as possible. It it is a good army leader, it will greatly decrease effectiveness of enemy army, during same battle.

Interesting. Never seen anything like that on my Serious games. Mostly seeing that either AI attacks with single army with hero or several hero-less armies of different (smallish) sizes... presumably those, which happened to be in the area at the time I've attacked...



Maybe in that case it is a proper preparation on AI side which is lacking then. Since I usually attack early mid game. But even in the proper mid game, i.e. turn 100 on fast, AI still lacking behind with the army equipment... And this is considering that I am going for the cheapish Titanium tier3 equipment myself... Since it is easier to field more units and replace them that way.

Auramagma wrote:
Interesting. Never seen anything like that on my Serious games. Mostly seeing that either AI attacks with single army with hero or several hero-less armies of different (smallish) sizes... presumably those, which happened to be in the area at the time I've attacked...



Maybe in that case it is a proper preparation on AI side which is lacking then. Since I usually attack early mid game. But even in the proper mid game, i.e. turn 100 on fast, AI still lacking behind with the army equipment... And this is considering that I am going for the cheapish Titanium tier3 equipment myself... Since it is easier to field more units and replace them that way.

I've seen them attack with 2 hero-lead armies on Normal. The effectiveness of the armies seems to be the biggest issue. Even very rich AI still seem reluctant to spend on upgrading units with the very best gear and currently there seems to be a bug/issue with them using Dust equipment. I have not seen any AI armies with Regen, for instance. That's a Dust accessory that can be very useful in battle and costs no resources but Dust. For most AI empires, by the mid game, they have a huge bank and are making a large amount per turn. Dust equipment would serve them very well.

Auramagma wrote:
1. It was fighting with more than one army attacking, defending together.

This is behaviour the AI is capable of, though it may not always have a second full army to spare.

Slashman wrote:
2. Make AI watch military ratings of other factions (particularly one led by player) and when seeing it going over certain threshold make AI build military of it's own. Currently, AI is still not prepared to invasion for both reasons: not combining it's armies and not building enough military.

The balance has currently swayed in favour of economic development, but we've taken note that players feel the AI is not aggressive enough. Looking into this...

Slashman wrote:
3. When seeing the direction of player's (other AI's) attack, bring the army to meet the invasion.

The AI is capable of intercepting attackers, but because it executes only at the beginning of the turn it can have trouble reacting to Blitzkrieg-style assaults (because it is effectively one turn behind). It isn't able to conceptualise "fronts" and "advances" but I'm not sure this is really necessary - or rather I'm sure there are many other things we'd be wiser to improve first!

Slashman wrote:
4. Some of the faction's military is handled by AI worse than the others. I.e. Forgotten assassins led by AI can't do anything (at all) against Mezari\Vaulters marines led by player or\and another AI in the mid game (2-nd - 3rd era, turn ~30 - 60 fast game). Mostly it happens because these assassins are poorly equipped and low in numbers (see points 1 and 2). Mezari and Vaulters are better handled by AI and more or less competitive. Allayi put up some fight (mostly due to having some numbers), but seekers in AI armies lack advanced equipment...

Auramagma wrote:
Maybe in that case it is a proper preparation on AI side which is lacking then. Since I usually attack early mid game. But even in the proper mid game, i.e. turn 100 on fast, AI still lacking behind with the army equipment... And this is considering that I am going for the cheapish Titanium tier3 equipment myself... Since it is easier to field more units and replace them that way.

Slashman wrote:
Even very rich AI still seem reluctant to spend on upgrading units with the very best gear and currently there seems to be a bug/issue with them using Dust equipment. I have not seen any AI armies with Regen, for instance. That's a Dust accessory that can be very useful in battle and costs no resources but Dust. For most AI empires, by the mid game, they have a huge bank and are making a large amount per turn. Dust equipment would serve them very well.

The issue seems to be the following: if U is the utility of the current unit design D (discounted by cost) and U' is the utility of the best potential unit design D'. For the AI to replace D with D', U' must be larger than k*U where k is a constant factor. The cost-benefit ratio of equipment doesn't follow a linear curve over the course of a 300-turn game though, so late-game the AI is hesitant to change its designs. Thing is if we just lowered k the AI might well backrupt itself changing designs all the time in the early game.

Regarding army upgrades, there are also some finesses that are even difficult for human player.



For example, let's say that you are strategic resource staved, but have some stash of them.

You can make design that uses some strategic resource in moderate amount.

Then upgrade some troops.

Then realize that you do not have resources to upgrade all of them, or build new ones, while still having some very outdated troops.

Then make new design that only uses dust, and upgrade most outdated troops, but not those that use strategic resource, if not too old.

Then create again design that uses strategic resources in moderate amount, after you obtain more resources by waiting, or even buying from marker/trade.



How would AI faction handle finesses like this?





EDIT:

That winder effect that reduces strategy resource gain by -1 is pure evil, for non-Vaulters.

wilbefast wrote:


The AI is capable of intercepting attackers, but because it executes only at the beginning of the turn it can have trouble reacting to Blitzkrieg-style assaults (because it is effectively one turn behind). It isn't able to conceptualise "fronts" and "advances" but I'm not sure this is really necessary - or rather I'm sure there are many other things we'd be wiser to improve first!

Usually, even if I attack AI city quickly, I am still besieging it first. For reasonable sized city, i.e. defence ~100-150 - it takes (in the early mid game) about 3-4 turns to get the defences down. So, AI has these 3-4 turns to launch rescue mission... Perhaps, as you said, it simply doesn't have enough troops to lift the siege effectively. Which means that:



a) it doesn't have enough field troops in general

b) those troops it has are not well equipped (as we discussed) and not sufficient to lift the siege effectively, which means that they die quickly and are not available pass siege one.



It might be an idea (if AI is not doing it already) for AI to collect defending troops from nearby city(s), put them in to army, add hero if possible, calculate whether it has enough time before the defences of the besieged city dropped say below 20-30%, reinforce the besieged city or lift the siege by attacking.



If it doesn't have enough time or too weak, then it might want to regroup in the next closest to enemy army(s) city.



Regarding Units equipment, would it help (and I am not 100% sure if it is a good idea ), if AI have had few checks to help it not to spend all of it's money on unit upgrades. I.e. it could've checked:



a) How much money\resources are required and spend only upto ~60-70% of what it currently has + say next 3 years of production... It must be willing to buy some resources on the market if it is still within budget.

b) Give it cooldown time for the upgrades - for each game speed, once it does one upgrade, I think it is OK if doesn't attempt another upgrade for X turns - might be worth calculating it based on the production rate (I mean economy strength progression).

c) Prioritise the upgrades, if two are available, then the one which benefits the specific unit type being upgraded, is attempted first.

Auramagma wrote:
It might be an idea (if AI is not doing it already) for AI to collect defending troops from nearby city(s), put them in to army, add hero if possible, calculate whether it has enough time before the defences of the besieged city dropped say below 20-30%, reinforce the besieged city or lift the siege by attacking.

We keep a minimum stock of units in garrison in the mid- to late-game, I'm not sure these are deployed even in the event of the a siege rescue mission. Another thing to check. Starting to be a lot of things now, so it's going to take us a while to get around to it.

Auramagma wrote:


a) How much money\resources are required and spend only upto ~60-70% of what it currently has + say next 3 years of production... It must be willing to buy some resources on the market if it is still within budget.

b) Give it cooldown time for the upgrades - for each game speed, once it does one upgrade, I think it is OK if doesn't attempt another upgrade for X turns - might be worth calculating it based on the production rate (I mean economy strength progression).

c) Prioritise the upgrades, if two are available, then the one which benefits the specific unit type being upgraded, is attempted first.

Pretty much already the case I don't want to go into exactly how the system works but resource production and stocks, equipment prices and provided attributes and the specific unit body are all taken into account when evaluating a unit design.

Not trying to nitpick Wilbefast, but some of those quotes in your post with my name on them aren't actually mine.



On another note, I saw better equipped AI armies in my last game on hard playing the Ardent Mages (I lost btw). I saw the AI actually equip Harmonites with Regen (unless they have that by default...I didn't think they did). So the AI is capable of using Dust equipment.



They also had minor factions mixed in well with their regular units but again failed to have researched or deployed second or third tier units.

Slashman wrote:
Not trying to nitpick Wilbefast, but some of those quotes in your post with my name on them aren't actually mine.

Oops I will edit the posts!

Slashman wrote:
On another note, I saw better equipped AI armies in my last game on hard playing the Ardent Mages (I lost btw). I saw the AI actually equip Harmonites with Regen (unless they have that by default...I didn't think they did). So the AI is capable of using Dust equipment.



They also had minor factions mixed in well with their regular units but again failed to have researched or deployed second or third tier units.

Yeah they can definitely use it, it's just as matter of how quickly they're upgrading to dust: it could be faster.

wilbefast wrote:
The issue seems to be the following: if U is the utility of the current unit design D (discounted by cost) and U' is the utility of the best potential unit design D'. For the AI to replace D with D', U' must be larger than k*U where k is a constant factor. The cost-benefit ratio of equipment doesn't follow a linear curve over the course of a 300-turn game though, so late-game the AI is hesitant to change its designs. Thing is if we just lowered k the AI might well backrupt itself changing designs all the time in the early game.

Maybe k shouldn't be a constant over the course of the game, then -- it should reduce over time, based on the game speed? Given that the cost-benefit ratio isn't a linear curve, the reduction probably shouldn't be either.

Getting back to the AI armies' composition, I'm still not happy that the AI refuses to build Tier 2 or 3 units with some of the factions, Forgotten, Ardent Mages, Drakken, Roving Clans, maybe the Allayi, Wild Walker, and Vaulters. Just now I was playing as the Cultists and the Forgotten are in my game: they refuse to build anything other than Assassins, which is quite silly. If they rounded out their armies with Predators and perhaps Mysts then they'd be much stronger than they are now. Not only that but I find that having to fight just one unit gets repetitive and makes battles more predictable.



I'm going to upload some save games that hopefully give some examples.
Cultists .zip
Forgotten .zip
Necrophages .zip