AI

[Seravy]

Which instruction tells it to form stacks, and how?

The AI has 1 rally point per continent.
It makes a list of "available" units, picks the 9 strongest, and sends as many of those towards the rally point as there is room.
If the rally point is full and any large stack fails to find a target, the continent is marked "uninteresting" and a new rally point on shore is set where the troops will be loaded into ships.

Units are "available" if they weren't attacking a target, moving towards another continent, boarding a ship, etc. Units in garrisons might be available depending on the number of defenders and type of garrison (see the AI thread for details)

The stack on the rally point can go towards a target any time it is "strong enough", which will empty the rally point and the process of gathering units starts over.

But even so, why not have it rush early economic buildings to further develop its economic advantage?

Because that strategy loses to early attacks just as badly as overdoing settlers. Marketplaces don't defend cities, halberdiers do. As is, the AI is doing far too poorly early, and more than well enough in the late game. If they manage to stay alive, they have no trouble winning with their resource advantage...but staying alive is the hard part.

It was a bersker, a ghoul, a skeleton or something similar against 9 nomad pikemen. Over and over. I'll see if I still have the save.

Berserkers are worth 134 combat value. Nomad Pikemen are 20. That sounds about right to me, no bug then. I wonder why they didn't win the battle. (Note that this assumes every unit had no levels)

[Catwalk]

Berserkers have 18 health, nomad pikemen have 12. 9 of them have 108 health. The combat values sound way off in that case. However, the AI also attacked with 2 ghouls.

How is the rally point determined?

[Seravy]

Berserkers have 18 health, nomad pikemen have 12. 9 of them have 108 health. The combat values sound way off in that case. However, the AI also attacked with 2 ghouls.

How is the rally point determined?

It the average of the sum of the location of the cities the AI has on that continent if I remember right - this is for "active" and owned continents. On uninteresting it's a shore, don't remember contested and enemy continents.

Berserkers have 7 swords, while Pikemen have 4. Berserkers have an extra +1 to Hit as well. And Berserkers have thrown for an additional 3 "axes". So while health is only 50% more, and shields are the same, attack power is like 4 times as much. (and the berserker might have had a level advantage - Barracks, Warlord, or Heroism. After the first attack the pikemen got damaged, which reduced their value further, opening them up as targets for even weaker forces)

(Pikemen have +25% for their armor piercing though, but that's nothing compared to berserkers.)
These are the values used in the actual strategic combat, so it should be accurate - it's used as separate attack and defense score there though.

[Nelphine]

How fast does the AI research?

I'm playing a hard game 3.31 right now, and I'm about to finish researching my first very rare in 1409. (I've also found 1 very rare from a node.)

My opponents have yet to show any signs of having rares, and even uncommons are.. not common? Now I know it's 'only' hard, but I feel like AI doesn't research all that fast.

Thinking back to my other games, mostly on extreme, this seems to be a fairly regular trend as well. I'm consistently ahead of them on research, regardless of my retort picks.

So, is this just due to the fact that by doing early research, I'm opening myself up to early attack, and if I get attacked that early I die, so I start a new game? In other words, I'm playing a pure luck game where either I survive early game and then I'm way ahead, or I die early? (Quite possible since that's how I feel the game always goes, but I've never considered it to be an effect of research before.)

[Seravy]

How fast does the AI research?

This entirely depends on their personality, retorts, and ofc power base.

I'd say they probably spend around 25-40% of their power base on research under normal circumstances, and much less if they are in war and spend mana in combat. Overlapping traits to boost research like Theurgist and Sage Master can push this rate over 50%.

Hard might just be too easy for veteran players.

it's also worth saying that research is the only area where the AI is not getting a significant bonus - so their primary advantage in research comes through the high power income.

If a human player keeps combat mana spending to minimum and doesn't raise skill much, they can beat the AI in research by pushing their allocation much higher than what the AI usually does.

[Nelphine]

Hum. Given that usually (not in this hard game) the ai have much higher power production, due to them getting 1-3 max power nodes before I do something still seems off. I'll have to look into it more closely.

Also, this hard game, maniacal militarist life orc wizard declared war on me. Their home island had 3 cities all on the coast; they have at least one other island.

Its been 4+ years and they've never attacked me (except an endurance trireme hunted down my scout magic spirit when it was crossing the ocean).

Any ideas why?

[Seravy]

Hum. Given that usually (not in this hard game) the ai have much higher power production, due to them getting 1-3 max power nodes before I do something still seems off. I'll have to look into it more closely.

Also, this hard game, maniacal militarist life orc wizard declared war on me. Their home island had 3 cities all on the coast; they have at least one other island.

Its been 4+ years and they've never attacked me (except an endurance trireme hunted down my scout magic spirit when it was crossing the ocean).

Any ideas why?

1. They probably didn't pick your continent(s) as the main action continent. While being at war increases the priority of doing so massively, it's not a 100% chance.
2. They are at war with someone else who is capable of sinking their transports. it's a Life wizard so there won't be any flying units that can avoid using ships.

The game I played felt far too easy as well to the extend I had to check if the AI resources are not bugged but...in the end I concluded I was way too fast due to early heroes. Very rares in 1409 are not realistic unless you you can expand too easily.

In fact, 1409 is turn ~100. The AI starts research on turn 30. Assuming a fairly large power base of 200 (yes that's large for the first 100 turns - most power buildings are not built yet), and a very high research spending of 50%, that's a total of 7000 RP. Enough for all commons, all uncommons and 1-2 rares. But we assumed above average numbers. If we consider a power base of 100 on average, and a spending of 40%, we get 2800 RP which gets all commons and probably half the uncommons. Add some wars to drain mana and we are at almost no research done at all.

[Nelphine]

I have.. 11 cities (fair land). One of the 3 arcanus wizards has more than me, although that one has never been at war with anyone. The other two were at war for a while, so probably that's what happened to my opponent.

For the research then, its probably the risk/reward gamble. I go for tons of research get all my commons and almost all uncommons, then skip rares. Now I have a sky drake and I don't think anything on either plane can fight it.

[zitro1987]

Opponents do not appear to work around the effect of 'suppress magic' wasting a LOT of mana on low-cost spells.

Sort of related - the great power of the global I think deserves 100 mana upkeep.

[Seravy]

They should be using low cost spells are a reduced chance, at least where I was able to do that, but this isn't simple in a weighted random system with multiple categories.

Which spells you mean in particular?