I'm posting this in the hopes that it gets seen by the powers that be for some consideration in improving the Tactical AI. We have come a long way, but there are still some things that need addressed.
For the sake of comparison, I generally have 2 Champions on mounts (Melee and Ranged with Heal/Haste/Slow), 2 Archers, 1-2 Spearmen ("Hammers"), and 2 Defenders ("Anvils"). There is occasionally a 9th unit, but its rarely necessary.
AI Targeting
Our opponents are doing a better job of identifying threats (Ranged>Melee for example), but there are some finer points that would add a great deal to the experience. Not the least of which is a tiny bit of randomness. I have a champion on a wolf with a full set of chain mail, equipped with the Heal spell and a bow for ranged support. Because she has less HPS than my units of archers, she is viewed as "primary target" by all the enemy archers. She is harder to hit, and does a significant amount less damage per turn than the sets of 7 archers with longbows, but is almost universally the target of absolutely every ranged unit and spell of my opponents. Not only does this make for a predictable fight, but it makes for an easy one, since She can be the ranged damage magnet, and pop a 20 hps heal to almost totally undo all damage done to her during a turn. This is poor AI behavior for several reasons.
The order of priority is currently correct in targeting units with ranged capacity first. However, it appears to ignore the capacity of that unit to mitigate or ignore damage, as well as the fact that a single full unit of 7 archers is more effective than two units of 7 at half strength because of their ability to exceed defense ratings
Ranged AI Recommendation vs. Ranged:
When I go into combat, I absolutely target archers first, but put one attack into each ranged unit to reduce their hitting power. Target the largest squad size very first. After every unit has taken a hit, the units of mine which remain the largest attack the largest, and those with reduced numbers attack smaller units to finish them off. Once the enemy units have been eliminated or reduced to 1-2 models remaining, I begin to focus on the melee units, as the ranged units become nearly inconsequential.
Targeting Melee units with Ranged units:
The same basic principle applies here as with the R v. R tactic. Use archer units that are still over half strength to attack the largest enemy melee units, to reduce their overall effectiveness. Prioritize Fast (Movement), then Soft (Armor, not hps), then Hard (armor, not hps), with tiebreakers going to the units closest to the ranged units first, then friendly melee units.
Role of Melee units:
"Anvils" - High Armor, High HPS, and since they are already expensive may as well toss on a good weapon. These units get center and move directly to intercept as many enemy as possible to tie them down. I try to protect these units, as they are costly and slow to build.
"Hammers" - The proverbial "Glass Cannons". Large as possible, Leather armor, biggest spears possible. They are cheap, and its totally fine if they die. They engage selectively once the anvils have tied units down. Target to reduce number of units on field first (weakest significant foe, above 10%), then dogpile the remainder.
In short; the AI needs to do a better job of identifying marginalized units (those below 50% health), and ineffective units (Those reduced to 1-2 models) and plan accordingly, because the combat usefulness of units by design reduces as they take damage.
Randomness:
It's all well and good to have a great system in place for picking apart the enemy, but if its rigid, it can be exploited (see the bit about my champion above). There really ought to be a "top 3 threats" calculation that is done when choosing a target at a 60%/30%/10% rate (or some such) so that I can't look at my units and know EXACTLY what is going to be targeted. I have not noticed a big issue with melee targeting (may have to do with reduced significance of targets), but with ranged, it is glaring.
