There are two obvious ways to handle the magic system in a Chainmail/D&D hybrid: the CM system with concessions to D&D, or the D&D system with concessions to CM. This is complicated by the fact that both systems are incomplete and ambiguous.
The mainly-CM way gives each magic-user a “spell capability” category. In 3rd edition CM, the categories are seer-magician-warlock-sorcerer-wizard. In early CM, the seer category hadn’t yet been added, and I think having “magician” as the lowest level is probably correct for D&D purposes, as seers are very likely to fuck up even low complexity spells.
Depending on your wizard’s spell capability, he has a better or worse chance on 2d6 of casting a spell of a given complexity, and of it “going off” on the round he casts it rather than the next round. There’s a nebulous rule about “the number of spells [a wizard] is able to manage” but it’s not defined further and there wasn’t any widely available contemporary clarification until someone recently dug up “The Battle of the Brown Hills,” which now sheds a little light outside of the text on the way Gygax did it. “Manage” might mean the number of spells you can memorize and can cast once each in a battle, or the number you can memorize and cast however many times you can roll successfully to do so.
In D&D, it’s explicitly stated that your magic-user could cast a memorized spell only once in a day. It’s not so stated in CM.
In CM, your wizard, regardless of his spell capability, can”manage” and attempt to cast a spell of high complexity — a magician-level wizard can attempt to cast up to a level 6 complexity spell, albeit with a very small likelihood of success. Again, this is explicitly not the case in D&D.
If adapting the CM system, I think the way to go is to put a character-level-based entry for “spell capability” on the D&D magic-user class table (levels 1-6 = magician, 7-8 = warlock, 9-10 = sorcerer, 11+ = wizard). The spell slots listed on the D&D table are the number and complexity of the spells that can be memorized, just like “normal” D&D. However, the D&D “fire and forget” rule would be discarded in favor of the interpretation of the CM rule that allows repeated casting of a memorized spell. (Although it’s invariably desirable to emulate Vance in atmospheric matters, we’re not mechanically beholden to the Mazirian magic system.)
Thus, a 1st level magic-user can potentially cast a sleep spell multiple times on a delve, but he’d have to roll 7+ each time he wanted to. This is obviously good for the magic-user on one hand and bad on the other.
(You’d need analogous spell capability entries on the cleric table, from say “adept” to “patriarch.”)
The next issue that immediately springs to mind is the matter of saving throws. I think you have to decide whether to use CM rolls for success or D&D saving throws. The former focuses on the level of the caster, the second focuses on the level of the target.
The way saving throws against spells work in D&D is, again, ambiguous, but you can infer that characters and monsters are meant to have saving throws against spells. There’s an entire saving throw category for “spells,” and the “wands” category explicitly provides that it includes “polymorph” wands, which implies that other polymorph effects are in a different category, presumably spells. Some spells explicitly mention saving throws (hold person, confusion, feeblemind, disintegrate, the finger of death).
As your character goes up in level, you have a correspondingly better chance of saving vs. a spell. This rewards advancement and gives the illusion of control over the success or failure of a roll. These are good things from a player’s perspective, and it makes sense that an experienced fighting man or wizard would be better at saving vs. spells.
Then again, it also makes sense that a high-level wizard would have a better chance of making his spells effective.
If you generally base the likelihood of success on the wizard’s level rather than on the target’s, you probably have to do that instead of allowing saving throws against most spells, as forcing/allowing both would nerf magic-users pretty hard. BUT basing the success on a roll by a non-player character wizard gives the sense that the result is somehow less in the target’s hands, even if the probabilities end up exactly the same. For some reason, the attacker rolling to hit instead of the defender rolling to dodge seems OK, but the attacker rolling to cast instead of the defender rolling to save does not. Maybe it’s just me, or it’s an ingrained thing.
(Note that the only “spells” that can be saved against in CM arefire ball and lightning bolt, and then only by heroes, super heroes, and certain fantastical creatures. However, they’re repeatable abilities in CM and didn’t become spells until D&D.)
If you pick the CM way to do things, you really have to junk saving throws vs. spells to some degree, probably almost thoroughly. I’d definitely keep saving throws in place for other effects, and for fire ball and lightning bolt and maybe a few others as well.
So, in deciding how to proceed, I think there are two practical considerations:
1. Do you want magic-users to cast a spell a limited number of times with guaranteed success, or an unlimited number of times without the guarantee?
2. Do you want the success or failure of a spell to depend on the caster’s skill or the target’s?
(One interesting avenue that the CM method opens up is an allowance for the efficacy of a wizard’s magic wand or staff, which could add a +1 to the casting roll or whatever.)
Anyway, I apologize for being all over the place, this wasn’t a planned post and I’m thinking it through as I type. I’m actually leaning towards the CM method just for the sake of trying out something different from traditional D&D, but doing so undoubtedly has implications I haven’t teased out.