The last time Rigale tried magic, I put 2 weeks into updating the 2008 masterpiece "Curtain of Fire" as wrapped in a Warband wrapper. Originally CoF was a non-Warband, really old mod from Mirithei (spelling?), that was for the very first Mount & Blade.
It was notable chiefly for having the source included, and became the godfather of all the mods that used magic after this, as will be shown. Thus it influenced Wheel of Time, Phantasy Calradia, Empires III, and others.
Actually looking these up I see they've moved around a bit:
https://forums.taleworlds.com/index.php/topic,111788.0.html Zelig's Wheel of Time mod
http://forums.taleworlds.com/index.php?board=333.0 Phantasy Calradia <-- I personally consider technically best
https://forums.taleworlds.com/index.php/topic,147665.0.html The Gathering: Wheel of Time with source
In my opinion the Gathering Wheel of Time is the predecessor to Phantasy Calradia in terms of the heads up display,
concept of "virtual weapon", and most of all the presentation to select next spell used (weave selection == spellbook).
Recently this code was made open source, clearing a murky issue of rights to use.
------
So with this out of the way, the Curtain of Fire source as merged into Rigale was never polished to be useable "as is".
Instead the idea was to prove things could be blown up etc but it wasn't done "right". It was mostly 2 weeks to show the concept worked and then was ported to Phosphor to use on top of Perisno, mainly because Perisno was "colorful" and Rigale was still "native" and a little "blah" without the fancy troops and armors.
weaknesses that need attention:
a) area effect spells had a bug where the thrower could become a target and powerful spells go off point blank instead of at a distance, making them unstable and dangerous. This could actually be a feature for Warsword Conquest skaven, but its annoying.
Instead, Warsword Conquest 2016 showed grenades (naptha bombs) with corrected area effect code to not explode at the thrower.
b) summoning often gave ID errors for the summoned troop. In addition the balance is off; it would be better to instead use a throwing weapon for the TIMING of when to summon but to not actually place the summoned troop at the target. Instead the troop summoned should be given initial AI direction to advance to thrown target or if it hit an enemy to set that enemy as the AI target.
The "weapon" used for throwing has to do enough damage that friendly AI will actually use the weapon. Finally, it would be better to select a random list of summoned troops, with a minus on the roll for low level summoner and a plus on the roll for high level summoners, allowing stronger results or weaker results.
c) It would be easier to make the first pass of magic for NPCs only to avoid having to implement a heads up display for mana and worst of all the presentation to select spells. Instead an NPC can start with a pool of mana that gets depleted and once depleted their ranged weapon allowing magic is out of ammunition, so that AI stops trying to use it. The spell selected should likewise be table based, with different costs per spell but a minus on the table by low levels and again when mana is < 1/3rd, with a plus on the roll at > 2/3 and again with high levels. this allows cheap spells mostly used by apprentices and NPCs low on mana, with stronger effects spell earlier in the fight and more often from high level casters.
d) A third class of buff/debuff casters or healers/anti-healers can then be introduced with effects lasting until end of the current battle, after which they are removed. Using a table driven dice roll allows some variation in which spell is used and again the most effective ones tend to be made early.
After finishing these layers, the presentation to select next spell and a display of mana remaining using the existing troop ratio bar from Rubik are all that are needed to allow the player to also use magic. This can be further limited by assuming schools of magic or religions as filtering some spells from the list of available spells. This list should be populated once only for the player at start of a battle and if several lists exist then each NPC could have a memory slot storing which list that NPC uses, for building classes of casters.
