Maybe the to hit chance floor should be 10% rather than 3% for everyone... That would go someway towards counter balancing the dodge stack strategy?

I am not sure that that's a good idea.

The issues, I see, for dodge stacking include:

• The AI has some severe disadvantages, for building high accuracy armies.
• The AI sovereigns do not use magic properly.
• Lucky makes dodge stacking easier.

If the AI issues are addressed, dodge could flip from being strategically good to being strategically bad.  But I think that the goal should be to have it be strategically valid, but not always valid.

Anyways, currently the two weaknesses of dodge stacking are magic (but armor is also vulnerable to magic) and accuracy (which has a minor use outside of dodge stacking -- if you can get +25 accuracy you'll have a 33% damage increase against non-dodge troops and the benefit increases against moderate dodge troops and there's a threshold effect against high dodge troops), and the AI is just awful about deploying accuracy and magic.  So, dodge stacking is currently overpowered.  

Lucky makes this situation worse but ironically also includes accuracy.

But if you have troops that can't be hit because of dodge, I do not think that it should automatically be the case that everything can hit them in a prolonged battle.  I think it should be OK to be strong against some opponents.  For dodge to be a viable strategy, I think that there should be some tradeoff in picking its counters.  And, perhaps ironically, lucky is one of its counters.

Is there a reason you entirely ignored my commentary on accuracy and magic?

If the AI begins taking way of the warrior, and way of the mage, I would expect that we would see accuracy contributions from some champions, and perhaps deeper spell use from other champions.

Meanwhile, some accuracy is available in troop traits, so we should expect them to show up in rushed troops if the AI notices or anticipates serious problems from dodgers.  But, yes, the AI would need to be able to assess army strengths and weaknesses.

That said, I have seen the AI ramping up to get 40 mana per turn from a single town with a couple essences and shards.  This was on insane, and was perhaps an overspecialization on refined arcana but it's not like building mana is outside the realm of possibility for the AI.

Also, there are many kinds of spells, and some can be effective enough without damage bonuses.

Anyways, the AI needs to be fixed -- if all that happens is the dodge numbers get changed the AI will still be crippled in other respects (like: magic).  I suppose we could cripple magic, also, to keep from having to fix the AI?

You do not need 40 magic per turn to be able to use magic against dodge troops.  40 mana per army might even be sufficient, though the details can vary depending on the army and on the map and a few other things and 100-200 mana per army is a plausible range for a empire with only minor specializations in magic.  Anyways, I'll go off and play a default game using resoln's magic and a warlock so I can do a better job of mentally translating into default turns the things which I have observed.

That said, I agree that the AI should be able to design units to counter threats, and it ought to be able to anticipate threats based on scouting reports.

Edit: I've played about 80-ish turns playing Oracle Ceresa (until the game crashed) on 1.12.  I am on a medium map with everything set to default/balanced/normal, and I am averaging about 1 death shard every 10 turns and I have three uses of wither available to me on a 3 turn cooldown (so I can cast it every turn, or I can use three instances in one turn in an emergency, and then wait a few before I can use it again).  I am not yet ready to take on strong monsters or high level champions but if things go like I expect, I doubt they will pose me any significant problems.  Anyways, right now, I could fight an army unit with 7 members each doing 10 damage (so 70 damage in total) and take no damage from members of that stack if nothing has a stronger attack -- it does not matter what I am fighting with, I can reduce that army's damage capability to zero.  That said, in this version, elemental damage (for example, the fire component of the melee damage from fire shrills - but that is new and I have not seen any armies with elemental damage, yet) does not get reduced by wither, so I am not completely invulnerable.  Also note that I have not harvested any crystal yet, nor have I done any military research yet, and I am not sure if I will bother to do so.