I look forward to my impending destruction.

It's kinda obvious but you didn't mention it so...

1. How can I get first strike in melee or ranged combat should be something the AI worries about. Currently the AI is easy to manipulate into ending his turn within your next turns strike range, even if he has higher movement then you. Allowing you first strike and a significant advantage. The AI should try to out maneuver you so he can get first strike. 

2. The AI should try and use ZoC to limit the amount of attacks you can land on a single unit. So it's hard for you to focus fire.

*disclaimer: The smack talk has begun

Ok. Your AI should be able to do the following stuff, else I will kick its ass every single time. At least, this is how I prioritize my targets and deceide on my plans. Oh and yes, I will happily divulge my secrets. Remember what Ash says: There's no such thing as overkill, only just enough kill.

Target priority. This is decision time. The following rules apply, in no particular order ( this is what I do in any tactical combat).

- Reduce number of attackers. He who slaps the most wins.

- Send casters to hell. The Deepest Pit, as Quick as Possbie if the caster can cast AoE's and summon nasties.

- Shutdown melee heroes. Shrink em. Slow em. Blind em. Curse em. If the AI knows ( and he can) that he can kill off a very, very nasty and quick melee unit in two casts, Take em out.

- Ranged units before slower moving melee units.

- Once you are dedicated to kill something, KILL IT! Currently the AI switches targets, especially against superior numbers, and attacks all of em a bit. If its dead, its not doing damage anymore. He who slaps the most wins... dead things dont slap. A unit with 1 HP leftover is still 100% more effective than a unit with 0 HP.

- He who strikes first slaps one extra time. So put your units with counterattack in the front line

Pick a plan, stick with it or switch.

The problem is that its not always "take out the highest threat first" or "kill the archers".

Casting time.

Basically there are a few types of spells. Buffs, Debuffs, Extra troops, Molten slag.Those are the options.

- Buff your own army. Battlecry and such are very usefull if started ASAP.

- Debuff the enemy : If a single cast debuffs the entire army and most of them are troops, cast it.

- Shutdown melee heroes. See target priority

- Reduce enemies to molten slag. As per reduce numbers, unless casters in which case we melt the casters first.

- If its all melee enemy, slow em to a crawl and then pick em off.

- If outnumbered and in possession of Elemental spells... cast em. Strong elementals first, keep em coming.

- Heal when needed, not always. If something does 40 damage per hit and your unit is at 34 HP, an 8 heal spell will keep it in the fight for one more round. But if you do 60 damage with Molten Slag and your unit does only 30... Heal it!  (think about the why).

Movement

- Slow em.

- Spread out your own army, even if its hang-back-and-cast/fire. AoE spells exist for a reason ( that is, to make sure you cannot be hit by them and to make sure you can nuke your opponent with them).

- Pile em up so you can AoE em (this requires two-turn planning).

- If the enemy lines up, pick a flank and run it over. Works well with slow em ( once lined up), so you can pick em apart piecemeal.

AI should be willing (And have planned) to sacrifice an army to kill off key units, and then wipe the floor with the rest in the next battle.And then laugh. And laugh. And laugh. And if its with 1 HP and you are with 0... laugh some more.

Mana usage

So lets say we generate 30 mana every turn and fight once every 4 turns, on average. To keep mana at its current level we can expend up to 120 mana. Now lets assume the enemy has 4 cities and we are off on war against the enemy. We stand to gain his cities, his production capabilities and his cash flow. We have 1000 mana, and generate 30 mana a turn. Assuming no-one else attacks us (And that;s a gamble I am willing to take, because I will have non-aggression pacts everywhere) the maximum amount of mana I would spend is... everything... becase I will be at +4 cities and thus increase the total might of my empire by +4 cities plus all the resources those lands contain. Unless we are at war with everyone.

Come and get me

p.s. Do bring out your dead. I've been at this since I was 9, I am now 32. Tabletop, PC, PnP or board, if there's something to kill, I will kill it. Make me an AI that struggles. LOTS.

 p.p.s. Common sense tactics always apply. If you point out something that's common sense and that the above doenst seem to fill, you are allowed to assume I've taken that for granted.

1)

Ranged units are handy, but not effective if you can afford to fill multiple armies. If you screen them with expensive melee units you can build and trait them to do decent damage at very low cost sans armor on any target on the field. It's very handy while the AI doesn't use spells. I'd say think of them as cheap filler units. If you can afford it, just build more of the melee units because their damage scales with stats.

2)

 AI should focus slow units on blocking enemy units and focus firing them down, while fast units are sent after vulnerable enemies. Watch American Football, steal ideas from how the offensive plays are run (line men screening the qb, running backs plowing through holes to get deep into backfield, receivers charging around and getting in the rear, etc). American football teams think more about this sort of small number locationally limited "melee" tactics stuff than you or I or anyone else ever will, so piggy back on their work.

3)

Vulnerability is irrelevant. Anyone who tells you otherwise spends their time faffing about rather than annihilating your AI. You should remove it in favor of differentiating weapons in some other way, it's a vestigial system. 

4)

Only cast aoe spells. Almost none of the single target, buff or debuff spells are worth the mana. This isn't necessarily true vs monsters, but the tacAI is only fighting the human sooooo. Only cast them at the start when enemies are clustered. If you can control strat development for AI mages, focus on initiative and aoe damage spell lines. You can significantly reduce costs through traits and gear to the point that many spells are free. 

More broadly you guys need to spreadsheet out a CBA on mana to damage based on potential unit concentration, a bunch of your spells are totally out of whack.

There are times that you absolutely want to switch targets before fully destroying a unit.  Because individual units within a group/company/etc can be killed and the damage the unit can do scales with what is left standing within the unit if your next attack is gross overkill for your current target you should switch to a healthier company that is within reach.  This will maximize the effectiveness of your damage output to destroy your enemy's damage output capability.  

For example, if your attack is capable of destroying 5 out of 9 guys in a unit, but your current target only has 1 guy left, you should attack the undamaged 9 guy stack that is also in reach because you're damage output hurts your opponent more by reducing the 9 stack to 4 than by reducing the 1 stack 0.  

"Forward, the Light Brigade!"
Was there a man dismay'd?
Not tho' the soldier knew
Some one had blunder'd.
Theirs not to make reply,
Theirs not to reason why,
Theirs but to do and die.
Into the valley of Death
Rode the six hundred.

Cannon to right of them,
Cannon to left of them,
Cannon in front of them
Volley'd and thunder'd;
Storm'd at with shot and shell,
Boldly they rode and well,
Into the jaws of Death,
Into the mouth of hell
Rode the six hundred.

Flash'd all their sabres bare,
Flash'd as they turn'd in air
Sabring the gunners there,
Charging an army, while
All the world wonder'd.
Plunged in the battery-smoke
Right thro' the line they broke;
Cossack and Russian
Reel'd from the sabre-stroke
Shatter'd and sunder'd.
Then they rode back, but not,
Not the six hundred.

One of the reasons I've been asking for a no mana in Tac Battle option, is so you would be able to use it also, I figured the best option would be calculating if mana is needed.

(maybe allow no mana to use counterspell if needed)

Exactly my thoughts. Maybe the 0.78 wil be better but so far tactics are just irrelevent beyond getting first strike and hide archers in corners. Even if you designed an AI that played with superhuman intelligence and skill, every battle would just come down to having the better spells and stats, even if the difference between the armies was very small. Tactical Battle Needs flushed out.

Vulnerability from what ive seen is so rare as to be none existant some champion armor only and dont hit elementals with magic of their own type). I can only assume it has yet to be fully implemented which makes it hard to contemplate your question. If expanded to monsters and regular troops than you should include it for target preference but don't bother with cutting/blunt/piercing if its staying the same.

This plus the AI needs to calculate who has more ranged firepower and act defensively with ZOC if it has that advatange. It should force you to go to it and use magic to slow you down if able.

AI should be smart enough to attack units on the way to more vulnerable units and to position its units where they are likely to be able to strike more vulnerable units, not just running in a straigt line. Attacking units should not only take every opportunity to strike at archers and mage but work towards that goal by focus firing on defensive units and other tricks.

I would like to add that it should be in the AI's best interest to start its tactical combat AI before tactical combat starts. What I mean is that it should prepare to go into a particular battle with a strategy in mind. There should be different strategies and possibilities, depending on the resources available to the AI - but the strategic AI should make use of its abilities to set up the most favorable tactical combat it can.

e.g. Just before engaging an enemy army, cast a spell to reduce all fire-resistance in the enemy army stack. Then engage in tactical combat and use fire elemental summons and fireballs.

1. Did we start the battle or where we attacked?

2. Am I interested in winning this fight or should I retreat?

3. Should I cast any spell or should I conserve all my mana for a more important fight or strategic spell?

4. As an AI do I know how to focus my armies attacks across turns on a single target?

Since we don't have ZoC you can't effectively block units unless the battlefield has choke points. ZoC blocking should have been added.

I wish this was true about, idk, EVERY game i've ever played..
The AI in every game always just attacks everything a little bit, causing few casualties and hence the AI recieves the same ammount of damage as in the start of the battle, while the player always focus-targets and thereby reduces the enemy damage output. End result, the one with the least damage output wins. Period.

Actually we can. Units adjacent to an enemy can only move one space or retreat. Not full ZoC, but it makes blocking an option. I am surprised you didn't notice.

Actually you can block flanking. Just put your units on every other tile in a line. Probably the lack of tactical AI and OP of heroes has obscured the limited ZoC, but it is also tiresome to read suggestions based in incomplete data. 

Placement:

I hope that tactical maps are designed in conjunction with your tactical AI. There are currently not enough maps. The ones we do have offer no tile bonuses. A few of them have a single choke point in the middle. I understand the design is to only have one round before melee starts, but there are so many ways to build the middle choke point. The AI needs to be able to handle them all. 

Ex: A block in the middle of the field has two choke points. A checkerboard of blocked swamp tiles, or swamp tiles that reduce defense and prevent movement. Three smaller blocked areas in the field, creating three choke points. 

The AI needs to be able to handle flanking and defending these types of maps. 

Couldn't agree with last two posts more, tactical battles need more terrain variance and a slight defensive terrain bonus would be an interesting mechanic.

There is some additional variability in blocked tiles/choke points in tactical battles fought in the howling chasm/storm dragon wilderness playspace. Single tiles distributed throughout the tactical map are blocked to movement, which allows a number of possibilities for anchoring lines and juking.

As for a more general reply to the OP: what Barrynor said. And then this: I'm going to use every cheap trick I can think of.

If I know the AI is prone to making a mistake, I will beat it over the head with said mistake. Your job is to make sure it learns from that beating. Don't leave gaps to exploit, either in AI tactics or in game design.

For example? Currently, armies arrive on the battlefield seemingly randomly distributed. Often I find my vulnerable casters and ranged units are dumped at the front of the army, and will have to stay there through most of the first round of initiative, because I can't switch between unit moves to rearrange. For me, that's something to plan around. (I haven't explored it much, but my suspicion is there may be some link to the order in which the units are added to the army, because my Sov is always first in the army but at the back in tactical - can your AI account for that?). But it's also an opportunity. Will I sacrifice a high-init melee champ to take down an AI caster that's stuck at the front for a moment? Well, right now, no, because the AI is dumb. But once their casters are something to fear, you betcha.

I'm the sort of person who passes multiple-choice exams by cross-referencing the questions. So, I suspect, are a bunch of other people around here. The AI can be as gifted as it wants, but if it or the game design has holes, it's gonna be eaten alive.

What's the human player's disadvantage? Well, if they're like me, they're lazy. I don't even play strategically with a spreadsheet, never mind in tactical. So beyond one or two-turn basic arithmetic, there's no reason the AI shouldn't be out-calculating me.

Most importantly, it needs to know when it's going to lose, and use that to advantage. There's got to be some way it can look at things and say "this is going to be a loss or a draw/too-close-to-call", dump as much offense as possible in one turn, and 'escape'. With an initiative advantage, repeating this can wear down even formidable foes over time. Even without an initiative advantage, it can deter a superior force from aggressing.

[Edit] One more thing that popped into my mind, that I've seen allusions to but always do myself is: match damage output to HP left as closely as possible. Avoiding the use of your 60 dmg melee hero to kill something with only a couple of HP is only the most obvious example. So keeping each unit free to choose as many targets as possible is key. This is also a reason I almost always equip my mages with ranged weapons - there will often be opportunities to kill something that has only 1-2 HP left, when the battle is at a point that doesn't require further casting. Save mana, spend an arrow.