AiCIM - Ai Casting Improvement Mod (pre-Alpha), current version is v.0.08

AiCIM - AI Casting Improvement Mod

Some Q&A about the mod.

Q. When will work on the mod begin?
A. Probably when all my current games end.

Q. When will the mod be completed?
A. Probably never, as like with most mods, it will probably always be a work in progress.

Q. What is the aim of the mod?
A. It's an ambitious one, as the aim is to try and get the casting AI to choose better spells more often when a mage runs out of script. As right now I think most people would agree that when the casting AI gets to choose a spell, it mostly always chooses one the player wouldn't have (as the player would have preferred to choose a spell that actually helped win the current battle).

Another aim (as of V.0.01) is that I want mages with default scripts to be effective in battle, and during testing I aim to have default scripted mages being able to successfully defend against both armies and SC's with only limited assistance from the player. (with the assistance likely being placement only)

Q. How will the above Aim be achieved?
A. The aim of the mod was the easy part, how to achieve that aim is where things get a bit tricky. Modding the casting AI itself is obviously out of the question, so the only modding options available (that I know of) are to change the various parameters of spells so that the AI will more likely choose to cast certain spells over certain other spells.

Q. How can you know what spell the casting AI is likely to pick?
A. This will be the laborious part of the mod. As knowing which spell the casting AI is likely to cast in any particular scenario is only really going to be achieved through a vast amount of testing and player feedback from practice (or even better save games files with battles to analyse).

From my time playing and testing the game I have built up a fair idea of how the casting AI decides what to cast. Although I will really need to sharpen my debug-log reading skills if I am to fully understand how the casting AI works. (but I do know friendly fire often plays a huge part in the equation. Especially when it comes to fighting thugs/SC's)

Q. What are you hoping to ultimately achieve with this mod, and how do you want it to affect the way current MP games play?
A. To be honest, I'm not 100% sure yet. My immediate short-term goal will be to stop the casting AI choosing useless spells in the heat of battle (another single longdead skeleton for 30+ fatigue anyone?).

Right now, I'm not that sure what the long term goal of this mod is, as I want to see how the short-term goal goes first. But I think it would be interesting to see if the way MP games are currently played is changed if, for example, a few N1 Indy mages could successfully defend a province against a thug by reliably casting Tangle Vines or Vine Arrow once their scripts run out (and not just spamming protection on PD troops that will die regardless like they often tend to do now).

Q. Will this mod be compatible with my favourite mods?
A. Lets slow it down a bit shall we My modding knowledge hasn't even reached the diaper stage yet, so I have no idea how compatible this mod will be with other mods on the market. In theory I want it to be fully compatible with all other popular mods, but I have no idea if that will be the reality, as all I'm looking for to begin with is a mod that functions with the vanilla game. (until my mod knowledge reaches the stage where I know about mod conflicts / compatibility etc.)


"Thank You" for your time in reading this, and look forward to your feedback (although if you are a known douche-bag, and/or known for talking shit in general about the game, then I certainly don't want your feedback)

----------------------------------------------------------

Change Log to date.

V.0.01

- Made D1 mages more effective in battles against SC's.
- Made N1 mages more effective in battles against SC's.


- Removed Animate Dead/Skelly (moved to Level 9 for now)
- Removed Decay (moved to Level 9 for now)
- Ghost Grip now produces just 1 effect, precision 100, and range 10 (lower range is to compensate for precision boost)
- Upped Protection to N2.
- Tangle Vines now precision 100, range 10 (lower range is to compensate for precision boost)


V.0.04

- (Hopefully) Did the donkey work to put all non-battlefield spells back to vanilla research levels, as that was the main drawback of using the debug mod as a base.

V.0.05

- (Hopefully) Managed to create a new summon Fire Elemental spell that can be used in scripts to summon Fire Elementals in battle, but which the casting AI won't cast on it's own. So I hope this will stop the AI wasting gems on summon spells (in theory, this nextspell trick can be used to make every spell in the game that costs gems to only be cast if it is scripted.)

V.0.08

- Changed all "Summon Lesser Elemental" and "Summon Elemental" spells to script only

- Removed all "Living" spells until I find a good way to handle them.

- Changed "Communion" spells to script only. (I hope, as this is one change I was very worried about getting to work properly)

- Changed "Luck" to Script only, but need to test this more (find better solution), as the mainspell effect I've used wears off, so there is a problem of the mage targeting units that already have the luck tag (since it only checks for the mainspell effect when targeting, and indeed will cast it on units that already have the luck tag from items)

- Continued removing, and moving around spells generally, all in the name of progress

- As a test, I have changed the "Fire Resistance" spell to "Fire Immunity", and wrote new description with an all new layout (will likely look into setting out a standardised spell description format for all spells that will give the player more info with regards what spells actually do)


V.0.11

(to be added once I have time to work it out)

Link to AiCIM v.0.11 - http://www.megaupload.com/?d=LE9ZSF0B


------------------------------------------------------------

Current ideas being considered for V.1.0. And My "To-Do" list

- Remove the animate dead/skeleton spells (Done)
- Compensate MA C'tis for loss of Animate spells
- Make Tangle Vines precision 100, and increase range substantially reduce range. (as it seems, if my log reading is correct, that the casting AI won't cast it currently against SC's due to friendly fire once the SC is surrounded by friendly troops) (Done)
- Find way to stop casting AI spamming Protection (maybe increase it to N2) (Done)
- Find way to get Protection back to N1 without casting AI spamming it. (maybe have it as script only)
- Do something about the casting AI's love of the (currently) pretty useless Decay spell. (maybe up Decay to D2) (Done)
- Think of something to do with Decay spell.
- Test the 'next spell' trick to see if it can be used effectively for summoning spells. (Done)
- If 'next spell' trick works, make all summon spells scripts only. (Half Done)
- If 'next spell' trick works, think about making all gem usage spells script only
- Try and find a way to modify existing summon spells so that they work like the new nextspell, as this will both avoid having to create a new spell for every current spell, and avoid using up spell slots. (Done, I hope)
- Compile a catagorised list of all battle spells in the game.
- Compile a catagorised list of all (vanilla) national recruitable mages in the game to see what magic paths are the most common, and to help judge how changes might affect certain nations.

- Might consider 2 versions of the Ghost Grip spell. One with percision 100 with 1 effect for anti SC use (currently used in V.0.01). The other being the vanilla version for use against several targets. ie. Against armies (which got modded in V.0.01 to the precision 100 version)

At some point in the near future I am hoping to take on the daunting task of trying to get the casting AI to more often pick the spell you want it to in battle once it goes off script, rather than the rubbish the casting AI often chooses now. I'm not sure if this is possible or not yet. I think it is from what I know of the game and the casting AI mechanics, but I could be wrong. But I won't know if the casting AI can be improved until someone tries to do it

I'm probably several weeks or months away from starting the mod right now, as I want to get all my current games out of the way first. But what I'm looking for with this post, and by creating this thread, is to try and get some feedback from players regarding what (likely useless) spells they are fed-up of the casting AI selecting when they go off-script, and what spell they wanted to see cast instead. Or indeed any suggestion regarding what spell changes that you think would help improve the casting AI. (please no stupid suggestions such as making Soul Slay S1, no resists and 5 fatigue)


I'm not looking to add new spells (for now), only mod the current ones. So please no suggestion about adding a new anti-SC spell that deals this or that damage / spell effect. I'm also only really concerned (for now) about spells that don't cost gems. So any posts like "can you please stop my mage using up all his gems on Fire Elementals" likely won't be of much help to me. (as I don't think it's possible to stop mages using gems if they have them for spells and/or fatigue reduction)


In a perfect world I want the casting AI to think more like a human player would when it comes to selecting a spell to cast. I know it's likely impossible to achieve this, but any improvement that can be managed will likely be for the betterment of the game IMO.


Edit - I'll also be looking into creating a sister mod to this one for helping with scripting. Such as making new versions of certain buff spells to help ensure the units targeted for the buffs in battle are the ones you wanted buffed. (although I will likely keep the old buff spells as well, as I'm sure someone out there really did want to have stronger archers with armour piecing weapons, rather than having the same effect on melee troops.)

This post has been edited by Calahan on Dec 2 2010, 12:25 PM

Hmm.

You can do quite a lot by chaining the spell you want to a spell that allows you to better control what the AI does via nextspell.

Burnsaber has come up with quite a few ideas along those lines. The AI decides on the first spell, ignoring all subsequent stuff.

Actually (and possibly not 100% aligned with this mod, which seems to be mostly combat AI...or are you trying for both?) - does anyone know how the AI proritizes summons? I've been trying to create some pre-built thugs for the AI to summon, but it's frustrating when the AI instead decides to Summon Animals or something...

Calahan - if this is threadjacking too much, just say so and I'll start a new one.

Thanks for the feedback guys. Always good to hear the thoughts of elite modders.

That's interesting, and I do seem to remember reading some of Burns's posts about this at some point in the past on this. And I'd imagine it's possible to do a fair bit of casting AI tricking using this technique. (I'll likely go hunting for those Burns posts at some point when I actually get around to working on this mod for real).

Can't say I know much about how the casting AI chooses which summons to blow gems on, but can't recall it ever casting anything other than summon elemental or undead horde, so maybe there's some hard coding involved designed to help the playing AI out?!? Although that's pretty much a total guess right now. But a bit of light modding and testing will probably be able to find out if the casting AI will cast any type of summon spell. (just mod it so that they are the only spells it can cast, and see if it does).

Along similar lines, I did wonder what would be the effect of changing the summon elemental spells to 99 fatigue and no gems. The summons themselves would need to be toned down a bit to stop spam abuse, but maybe some X1 mages being able to summon 2-3 fairly weak elemental monsters per battle for no gems would be an interesting change for early-mid game battles (again, no idea if it's a good or bad change, just interesting at this point)


I don't mind a bit of light thread-jacking now and again but I think this issue of battlefield summons is on-topic for this thread, since adding new spells might be something I'll look into once 1) I know what I'm doing, and 2) I know if what I'm doing is actually making the game better or worse. (hoping for the former obviously). So I'm happy for the conversation on battlefiled summons to stay here if you like.

And I'm basically looking to mod any spell that can be cast in battle that will improve the casting AI. Be they combat, buff, summon, whatever. If the casting AI considers it, then it's likely to figure into the equation somewhere.

Calahan, you're my hero. If you manage to even so much as make a moderate improvement here it would be IMHO an astounding achievement.
Thanks you for championing this cause!

Who said dom is dead???- just look at all the great mods in progress.

You could also look into making some of the useless spells better, like decay wouldn't be so bad if it hit reliably and perhaps had some additional effect like a light poisoning.

Removing animate dead/skele is a great thing all by itself. C'tis ain't really hurt I think, there are other options for the assassinations and they still gains a lot from the removal in normal battles.

Protection and other spells you don't want the AI to cast could perhaps be achieved with the nextspell trick, by replacing it with a spell that does nothing (so the AI won't cast it) but has the protection effect nextspelled. Thinking about it, couldn't that work for the animate dead spells too? Can you make spells without any primary effect?

Or a primary effect that's so minor it doesn't matter. I doubt many will begrudge you an AoE 1 fatigue damage effect.

That's interesting, and I do seem to remember reading some of Burns's posts about this at some point in the past on this. And I'd imagine it's possible to do a fair bit of casting AI tricking using this technique. (I'll likely go hunting for those Burns posts at some point when I actually get around to working on this mod for real).

If you want to see some AI tricking, I've done spells on the matter in CPCS (follow my sig link) and other mods. I usually use it to make a spell that the AI simply refuses to cast without scripting. The trick is rather simple. Make a really weak spell that has simply neglible effect (like mr easily negates haste buff on caster) and chain the effect you want into that weak spell.

The casting AI thinks that the spell is really, really lame and simply refuses to cast it because it has "better" options in even low research levels. The dowside to this trick is that if the targets already have the buffs that the weak spell is supposed to give, the AI won't cast even if ordered (just as it won't cast haste on troops that already have haste). This can be circumvented by fine-tuning the buffs that the weak spell "gives", like making it give haste+trick fate

This trick could easily be used on some current spells to make them less attractive to casting AI (animate skeleton/dead, blink, the berserk buffs).

Oh, right! One thing could be especially good for this is the spell effect number (17) for the Man spell song "Song of Bravery". It increases temporarily the morale of the affected unit, it needs testing but could be a good "neglible" effect for AI tricking purposes. (since mr-easily negates buff could get through and prevent further castings of a summon spell, for example).

I'm somewhat convinced (note: not tested) that the only thing the casting cares about it BF-summons is fatigue and the number of summons. I don't think it has any way of measuring how strong the summons are. This could be tested by putting a D4W4 mage with water gems on the back and see if it prefers to cast raise skeletons or Living Water (with both spells researched or set to research 0, of course).

It really sounds like it would be way more trouble than worth. Summon minor elementals didn't cost gems in Dom:PPP and the AI very rarely bothered to cast anything else, even if dudes were in range of lighting bolts and the like. So you would have to really make a AI-tricking mainspell for each of them.

Many thanks for your input Burnsaber. Your post has really got me thinking that I do have some hope of tricking the AI into the illusion that it has become smarter


That next spell trick could certainly remove a lot of annoying spells from the equation. After Sombre mentioned it, I did start thinking how it would work, since how would the casting AI target enemies with spells that had no primary effect? But your post shows I was thinking about the problem from the wrong angle, as this trick can instead (in theory) be used to remove all the 'friendly' spells from the casting AI's list of options. (If like you say a negligible spell effect can be found that the AI will always reliably cast on it's own troops.). This could have a huge impact if it works like I hope it will (and I will check out your CPCS mod the first chance I get, since I shamefully haven't looked at it before now )


And if I can kindly pick your brain Burns, I have just thought of two questions on this next spell trick.

1 - Can the linked spell be a summon spell? So for example, could the current Summon Element spells be replaced with a 100 fatigue + 1gem "Song of Bravery type effect" which the casting AI would target on it's own troops (or even target the caster), and then the linked spell would summon the elemental. As in theory, this could be a great way to stop the AI wasting gems on unwanted summons.
(Edit - After re-reading your post again, you seem to reference that the linked spell can be a summon spell. So I think that answers this question unless I'm missing something)

2 - Are you limited to just one linked spell, or could you in theory have as many spells linked in a chain as you like? My thinking here is if the first spell removed an effect caused by the second spell (with the third spell in the chain being the actual spell), then the casting AI would always have a valid target, thereby getting around the "won't cast a spell on a unit that already has the effect" problem. (although must confess right now I can't actually think of a spell effect that fits the bill on this one)


Thanks again for your thoughts Burns. They are certainly very much appreciated.


Edit -

It really sounds like it would be way more trouble than worth. Summon minor elementals didn't cost gems in Dom:PPP and the AI very rarely bothered to cast anything else, even if dudes were in range of lighting bolts and the like. So you would have to really make a AI-tricking mainspell for each of them.

Yeah, after I posted I put some more thought into the actual effect of gem-free summons, and realised the result would probably just be the AI would spam it to infinity at all times, and above all other spells. Since the casting AI really does love summoning more chaff onto the ballpark. So the effect would probably be to go against the entire purpose of this mod (nice start Calahan :headslap:)

My main idea was to avoid the casting AI wasting gems when they go off script, but if that can be achieved with next spell trick (as I briefly outlined in Q1 above), then there's no need to have gem free summons.


Yum yum, lots of food for thought

This post has been edited by Calahan on Dec 3 2010, 12:14 PM

All elemental summons should be in the 'don't cast unless scripted to' category.

In fact a good starting point would be to make a list of all those spells that you never want to cast unless scripted to (usually because they are spells you'd always want in the first 5 rounds of combat, have potentially negative effects, are very situational and expensive etc).

Agreed

Also agreed. My current thoughts for when I really get started on this mod is to list and catagorise every possible battle spell into groups along the general lines of

"AI should never cast, and is for scripts only"
"generally always a useful spell in most situations, happy for AI to always cast it"
"only want AI to it cast when fighting/targetting SC's"
"only want AI to it cast when fighting/targetting armies"
"only want AI to cast it when the targets are immune to it's desired first choice spell(s)"


I'd like to think the first one can now be achieved with he next spell trick. As for the others ones, I have no idea how realistic they are as objectives (but know that the reality will probably be that only very minor improvements, if any, can be made)

Cool, glad to hear that all that time I used on spell modding for CPCS was not wasted .

There are a lot of neglible hostile effects that you can use for AI-tricking hostile spells (like Decay). Damage 0 fear damage that is not set to AP or AN or just 0 stun damage, for example. But I think better option would be to make decay D2 and boost it somehow (increased AoE? no save?).

I have to admit though that buffing spells are a bit problematic with the AI-trickery, because you can then end up in a situation where the AI casts a buffing spell on units that already have the buff. For example, if you made the N "berserk" spells have non-berserk causing mainspell, the AI could cast it on already berserked troops.

Another downside to having AI-tricking mainspell is that the in-game spell description will lie about the true AoE of the spell, so you have to write OOC stuff in the fluff description like:

"THIS SPELL HAS REAL AOE of 5, UNLIKE IN THE DESCRIPTION ABOVE"

Another thing is that effects that increase with caster level (like AoE: 5+) do not work on #nextspells. This sucks for "Living X" elemental spells..

Oh, and one word of warning. Do not use "only on caster" buffs as AI tricking mainspells, because if communion master casts it on a communion, the spell will be copied on all slaves too (which would be very abusable with say a elemental summoning spell ). I actually used this to mu advantage in my Alugra EA nation mod, which has a spell like this that has self buff tricking main spells, clears 100 fatigue and blinks the unit. So if it is casted on a communion, the master and all slaves are blinked and recover 100 fatigue (which is intentional since the nation is supposed to be good on thug communions). It's rather crazy, but fun . However, you probably shouldn't create a spell like that on accident.

On hindsight, CPCS does not have lots of AI trickery. Many of the spells just have tons of nextspells that make the spells look less valuable in the eyes of the AI so that they do not mess up regular out-of-script behaviour. I'll mention some examples from the mod though, just give some more food for though.

Curse of the Salamander + Static Bolt = Contain a hostile effect as mainspell to force some "buffs" on the enemy troops (heat aura on non-fire immune units and charge body on non-lighting immune units respectively)

Spring Winds = Healing mainspell tied to a fatigue restoring next spell (so that it cannot be used to restore fatigue from mages, since the AI will cast it only on injured troops)

That's the extend of it, me thinks.

Yeah, totally possible.

You can link as many spells as you like. However, I don't think that there is any way to remove any spell effect. I mentioned the "Song of Bravery" effect because it seems to be only temporary spell effect in the game. It gives a unit +5 morale and the bonus is reduced by -2 every combat turn. I don't know how the AI targets that effect (like would AI cast it on a unit that already has the +5 mor bonus).

BtW, were you thinking about touching the Pan/ Man/ TNN spell songs? They are rather useless at the moment (except the fatigue reducing spell) and I think they could use a rather hefty boosts. I'd kind of like to see them made into 2 turn cloud effects centered on the caster (the sing lingers on for a while)

Can't really think of any way to do this. When elemental damage routes are blocked the AI usually goes for buffs. I don't think that AI even figures out to use "Stellar Cascades" or other "auto-hit" things against SC's. Perhaps it does not think of stun damage as real damage?

High numbers of effects that are not AP/AN (ergo: target chaff)? I'm not sure, but low precision might lower the estimated damage calculations that the AI makes, which might make a spell undesirable against small amounts of troops.

Tough one. I think you would have to start unraveling this per magic path. The preferred spell of death is summon skellies/dead, but it's pretty darn near impossible to make a spell that AI would rather cast than animate skellies. I think that fire path with "fire damage"<"Blindness" when facing fire immune units is working okay for now. Parhaps try to make similar structure to other paths?

Air mages love to spam confusion on SC's, which is... stupid. Make a inslge target irresisitble confusion (like CBM rage)?

This is some truly awesome feedback Burns, which is really starting to spike my enthusiasm levels for getting this mod started. You have my deepest thanks as always for all the help and guidance.

Cool, glad to hear that all that time I used on spell modding for CPCS was not wasted .

I'm very sure there are a load of people out there who truly appreciate all the work you do on the modding side of this game, and all the little modding secrets you find that open up a whole range of possibilities. But most of these people are just too shy and quiet to articulate their thanks. "Learn to articulate more praise where due you articulate-lacking bastards."

There are a lot of negligible hostile effects that you can use for AI-tricking hostile spells (like Decay). Damage 0 fear damage that is not set to AP or AN or just 0 stun damage, for example. But I think better option would be to make decay D2 and boost it somehow (increased AoE? no save?).

Thanks for the info regarding ways of getting the AI to target enemies for using next spell options. Although not sure where such negligible effects will rank in the casting AI's priority list to get them to pick a "next spell" spell over a regular one (but trial and error testing can likely provide some answers)

I have to admit though that buffing spells are a bit problematic with the AI-trickery, because you can then end up in a situation where the AI casts a buffing spell on units that already have the buff. For example, if you made the N "berserk" spells have non-berserk causing mainspell, the AI could cast it on already berserked troops.

Another downside to having AI-tricking mainspell is that the in-game spell description will lie about the true AoE of the spell, so you have to write OOC stuff in the fluff description like:

"THIS SPELL HAS REAL AOE of 5, UNLIKE IN THE DESCRIPTION ABOVE"

Another thing is that effects that increase with caster level (like AoE: 5+) do not work on #nextspells. This sucks for "Living X" elemental spells..

Oh, and one word of warning. Do not use "only on caster" buffs as AI tricking mainspells, because if communion master casts it on a communion, the spell will be copied on all slaves too (which would be very abusable with say a elemental summoning spell ). I actually used this to mu advantage in my Alugra EA nation mod, which has a spell like this that has self buff tricking main spells, clears 100 fatigue and blinks the unit. So if it is casted on a communion, the master and all slaves are blinked and recover 100 fatigue (which is intentional since the nation is supposed to be good on thug communions). It's rather crazy, but fun . However, you probably shouldn't create a spell like that on accident.

Hhhhmmm, good call on the buffing problems, as the casting AI will only look to see if the target units have the primary spell effect, and not the secondary. Maybe a permanent buff will be needed for the first spell to avoid re-targeting (Hhhhmmm, what's the least helpful permanent buff? Haste most likely as you mentioned)

I don't mind re-writing as many spell descriptions as needed. If it helps the casting AI, then it will be worth the time and effort.

Good point on the caster buffs and communion abuse as well.

While I'm here, one thing I've just become curious about is can the next spell trick be linked in any way to spells that get cancelled on hit? (such as Twist Fate, Ritual of Returning, Mossbody). As it'd be cool to have a new spell that created a Guardian, which got auto-summoned to defend the mage should he receive damage. (but my understanding is that this is likely outside the scope of current modding options, as you can only mod spells to apply a certain spell effect, but the effect themselves can't be modded)

On hindsight, CPCS does not have lots of AI trickery. Many of the spells just have tons of nextspells that make the spells look less valuable in the eyes of the AI so that they do not mess up regular out-of-script behaviour. I'll mention some examples from the mod though, just give some more food for though.

Curse of the Salamander + Static Bolt = Contain a hostile effect as mainspell to force some "buffs" on the enemy troops (heat aura on non-fire immune units and charge body on non-lighting immune units respectively)

Spring Winds = Healing mainspell tied to a fatigue restoring next spell (so that it cannot be used to restore fatigue from mages, since the AI will cast it only on injured troops)

That's the extend of it, me thinks.

Thanks for that, as I'll know exactly where to look now for examples when I get started.

Yeah, totally possible.

Awesome

You can link as many spells as you like. However, I don't think that there is any way to remove any spell effect. I mentioned the "Song of Bravery" effect because it seems to be only temporary spell effect in the game. It gives a unit +5 morale and the bonus is reduced by -2 every combat turn. I don't know how the AI targets that effect (like would AI cast it on a unit that already has the +5 mor bonus).

BtW, were you thinking about touching the Pan/ Man/ TNN spell songs? They are rather useless at the moment (except the fatigue reducing spell) and I think they could use a rather hefty boosts. I'd kind of like to see them made into 2 turn cloud effects centered on the caster (the sing lingers on for a while)

Yeah, there probably aren't many effects that can fluctuate up and down in battle that can be controlled by spells. HP's, Morale, Elemental resists, Fatigue, is about it I think (?). Strength (?) (there's a strength reducing weapon. Can it be a spell effect as well?).

Can you damage your own units with spells? As I was thinking that if the second spell damaged the target for 1HP, and the first spell healed the target, then the spell would always have a valid target. (although regen could be a problem, and of course how the unit got damaged in the first place in order to be a valid target for the first casting of the spell). But all this is likely irrelevant due to the problems of the AI re-targeting the same units all the time, as mentioned above (unless the first spell had permanent buff)


To begin with I'm only looking at changing spells in such a way as to get the casting AI to use them when they should. For most of the Alpha testing I'll likely be getting rid of a huge range of spells to streamline testing (see below)

Long-Term I'd probably look into helping/nerfing certain spells where need be. As in my mind I have plans to do a range of mods on all aspects of the game, which can be combined together. Like CBM used to be. Whether or not my mind will match both reality and my modding skills is another matter entirely though. But I'm looking to start small and see where it takes me

Can't really think of any way to do this. When elemental damage routes are blocked the AI usually goes for buffs. I don't think that AI even figures out to use "Stellar Cascades" or other "auto-hit" things against SC's. Perhaps it does not think of stun damage as real damage?

I'm hoping the unit buffs can be taken care of with the next spell trick (so making most of them script only in effect). And that each mage will only have so many self buffs it can cast. After which it won't have any choice other than to cast a 'useful' spell. Or at least that's the theory I'll be trying to work on

What is interesting though is that the Indy AI mages are often better than your mages if left to default. As the Indy AI has a limited spell book, so if you attack an Indy prov with a thug which contains (for example) a N2 mage, then once the thug is surrounded, the mage will walk up the battlefield and start spamming sleep. Whereas a player's N2 mage in the same situation would either cast self buffs, or walk into range of friendly troops and spam Protection.

So in many ways I'm hoping to get the player's mages to act more like Indy AI mages do (but with a full spell book rather than a very limited one). It's also no coincidence that some mages that were useful in battle suddenly become useless once you research a certain spell. I had a MP game 18 months ago as Yomi where my lovely Incinerate / Banefire / Drain Life DO's/Hannya mages suddenly become very unreliable just because I hit Ench 3. Skelly Hordes FTW!

High numbers of effects that are not AP/AN (ergo: target chaff)? I'm not sure, but low precision might lower the estimated damage calculations that the AI makes, which might make a spell undesirable against small amounts of troops.

I have a feeling that the casting AI would improve dramatically if all spells were 100 precision. But then that would remove an entire mage stat from the game. Is there an actual limit on the precision boost a spell can have? As I'm thinking maybe there's a sweet spot somewhere at which point the casting AI will target enemies even if they're surrounded by your own troops, as they are so accurate that friendly fire concerns are negated (but not accurate enough to be a guaranteed hit.)

But I could be talking shit here, as until I really dive into the debug-log, I can't be sure that there isn't two separate mechanics for 100 precision spells, and regular. (with 100 precision ones designed to ignore friendly fire issues). So unless a spell is 100 precision, there will always be targeting concerns for the casting AI, even at 99.

Tough one. I think you would have to start unravelling this per magic path. The preferred spell of death is summon skellies/dead, but it's pretty darn near impossible to make a spell that AI would rather cast than animate skellies. I think that fire path with "fire damage"<"Blindness" when facing fire immune units is working okay for now. Perhaps try to make similar structure to other paths?

Air mages love to spam confusion on SC's, which is... stupid. Make a single target irresistible confusion (like CBM rage)?

My current Alpha-testing plan is to completely dismantle the spell book until each pure X2 path (or maybe even X1) has just 3 spells available. One self buff. One anti-SC. One anti-army. Then once I have those three spells being cast reliably in the desired scenarios, I'll add more spells one at a time to see how that affects the casting AI.

At least that's my current plan (but may well change completely by the time I actually get started)


Edit - God damn that turned into a stupidly long post with all the quoting. I've just decided that quoting is for suckers.