One of the thornier issues in AD&D is on when a spell resolves in the round. There are – in general – two schools of thought on the subject. The first believes that spell-casting starts at the very beginning of the round and concludes in the segment described in its casting time. Thus, Magic Missile always resolves in the first segment of the round, and will always beat a Fireball, which takes 3 segments to cast. The second believes that spell-casting starts later in the round, generally on the segment given by the opponent’s die roll, and concludes later on, perhaps even in the next round.
Although I generally subscribe to the first method, amongst the precursors to the AD&D initiative system, there are a few indications that the second system could be used.
In Chainmail, there is an optional system that describes the complexity of the spell, a value of 1 to 6. Darkness has a complexity of 1, whilst Anti-Magic Shell has a complexity of 6. When a spell is cast, a roll is made to see how it works – roll high enough and the spell takes immediate effect, but it could also be delayed until next turn or be completely non-effective. So, a spell could take place the round after it was cast.
The Eldritch Wizardry system gives modifiers of -1 to -8 to Dexterity for casting spells of above first level. Unfortunately, the entire system is somewhat spoilt by the example not making much sense when compared to the tables. Honestly, it’s particularly difficult to work out what is going on. (Is there an entire missing table? There well could be).
However Swords & Spells has a very interesting system where spells can take place immediately (1st/2nd level spells), be delayed half a turn (3rd-6th level spells), be delayed a full turn (7th-9th level spells) or a turn-and-a-half (high level spells on scrolls).
Lenard Lakofka in Dragon #34 describes a system where spell-casting begins in segment 1d4, and concludes its casting time in segments later. Meanwhile, melee attacks occur on the segment rolled on a 1d10, or on two individual d6 results for multiple attacks. Weapon speed factors add from 0 to 3 segments to the segment the attack takes place in. (1-3 = +0, 4-6 = +1, 7-9 = +2, 10+ = +3). In a later update in Dragon #45, Lenard decided that the d10 for weapons was too slow – a d6 should be used instead.
Jim Ward in Dragon #42 explains initiative as follows:
Question: After the surprise dice are rolled, what happens? How does weapon speed factor fit into this? What was the line in the DMG referring to .multiple attacks and speed factors (page 66, under Weapon Speed Factor) about?
Answer: The first part of the question is detailed under the initiative section of the DMG found on page 62, with the nonsurprised being going first. A weapon’s speed factor can partially negate the initiative gained by surprise when slower weapons are used versus high-speed weapons (consult the section on page 66 for details). There are several allowances in the rules for beings attacking more than once in any given melee round and these must be taken into consideration when using weapon speed factors (Rangers and Paladins receive multiple attacks after gaining experience and beings of larger hit dice attack the low hit-point creatures in multiples).
All of which is not particularly helpful.
By AD&D, we had explicit casting times for spells, and the first division of the round into ten segments of 6 seconds each.
2nd Edition had an optional initiative system where a d10+weapon speed or casting time indicated when the blow or spell fell in the round; the numbers were comparative only, however, and did not correspond to actual segments of the round, with that idea being dropped.
Finally, 3rd edition – much like Basic D&D – returned to the idea of spells being cast instantaneously, albeit with a chance of disruption due to Attacks of Opportunity, which could be negated with a high enough Concentration skill.