AI

[Nelphine]

What would happen if we changed the required overland strength required for an AI stack to consider attacking a target based on difficulty? For instance let an easy AI attack with something that only has 25% the overland strength of its target, then add 25% for the next two difficulties. Then for master and lunatic, we could reduce it a little bit again since AI at those levgls usually have very strong spells so don't need as strong units. (Probably 70% and 60%?)

So something like: required overland strength to consider attacking a target: 25/50/75/75/70/60

[Seravy]

Even with the current 100% plus the ranged modifier, battles where I can wipe out all enemy units on turn 1 is frequent and I don't even need Guardian for it. Reducing the number will just make the game easier. (it was 75% for a while, it got raised to 100 a few months ago.)

Attacking is higher priority than stackbuilding - if the amount needed is less, the AI will build and send weaker stacks.

[Nelphine]

Right, which is why I think lower difficulty ais should be doing it. To do weaker attacks for new players.

Have you been doing much master +? I definitely think lunatic needs to attack more often.

Maybe 50/75/100/100/90/80?

[Seravy]

Are you sure you're not basing this on the excessive military power of berserkers? are your non-berserker garrisons and stacks not getting attacked often enough?

And no, so far I finished only one game in the new difficulty system which took 28 hours of raw playing time. I'm still in the early game for the next one and probably won't complete it this week.

It's true that putting multiple Behemoths into my cities (or towers) was a fairly effective way to avoid attacks, but each unit used for that means less I can use to attack the AI, ultimately it's helping them. Every once in a while a stack with 2-3 colossus or wyrm showed up and attacked anyway, and committing more then 3 behemoths to defend a single city is clearly not a viable tactic.
Stacks of 9 Magicians or beetles certainly scared away weaker stacks but anything that had any chance of winning did attack it and I lost at least 5-6 cities during the game.

It might sound weird but not attacking hurts the player more unless the attacking force is so overwhelming that player can't defend the city at all. The maintenance cost of those scary garrisons takes away a significant portion of the profit from the city, and the enemy units scattered everywhere makes it very hard to build an effective offensive stack because the unit get attacked on the way. On top of that, they eat my lairs and steal my nodes. And if you want to clear them out, you can't rely on your city defense effects (Heavenly Light!) and give them the first turn which hurts twice as much as a defensive battle against the same thing.

(I agree that making the AI weaker for normal and easy is a good idea, but nor sure about Master and especially not for Lunatic where the AI summons doomstacks at an even faster rate)

[Nelphine]

Yes, my other garrisons don't get attacked often enough. But this isn't just about that - its about taking lairs (which, since the AI has high casting skill, they can beat even with lower army strength) and attacking each other.

If the AI only attacks another AI when they have equal or greater strength, there will be a lot of stalemate wars arnuz was mentioning, because all AI follow the same rules for stackbuilding and garrison building. I was under the impression that we were still at the 75% requirement (my bad), which is (to me) one of the big things that allows AI to make headway against each other. There will be lopsided battles. Without them, its simply whichever AI has better casting skill wins, even though they both attack triggered by the same criteria.

[Seravy]

All right, I'll add this to the things to watch out for. I can't decide on it until after I played much more games so don't expect a quick change.

[Nelphine]

That's totally fine. If there's any specific data I can gather for you, let me know. Otherwise, a watch item is fine.

[Seravy]

Nothing in particular. If you see a situation where you believe the AI should have attacked but didn't, and could have at a less than 100% rate, post the save.

[Nelphine]

Hard to do for AI vs AI, but I'll try to watch for it.

[Arnuz]

Another option might be to get AIs to decide also considering their spells and books, rather than only strategic value. With strategic value only it'd seem that the AIs with the best buffs would attack more easily, but a good old lightning will do wonders against even lionhearted holy armored klackon halberdiers...