Spells: preparation vs. improvisation
Everybody has an opinion on the so-called Vancian casting system of classic D&D. Some (myself among them) find the idea of "fire-and-forget" spells to be a poor representation of the kind of fiction we want to represent. Some (again, including myself) find it too rigid, and dislike the incentives it provides toward spell selection. And then there are those on the other side of the fence, who like the in-fiction implications (still not a fan) or who see value in such a system for the purpose of promoting game elements of resource management and player skill (for which I do have an appreciation.)
Not a whole lot need be said about the in-universe implications of spell preparation. It could literally be a case of impressing patterns of magical energy (or imprisoning a minor daemon or other entity) in the mind and then discharging it, losing the pattern or knowledge until it is prepared again) but it could just as easily be reskinned as carefully preparing proper reagents, expending the reagents in the casting while retaining actual knowledge of the spell, or even making complex calculations to calibrate a spell to the current positions of the stars or what-have-you, and having to recalculate them each day, or indeed for subsequent castings on the same day.
The effects on actual play deserve a little more analysis.
On the most rigid and restrictive end of the spectrum, we have strict preparation. You can prepare x number of spells per level per day, and you can cast those spells, and only those spells, in the exact numbers in which you prepared them. If you come across a situation in which read languages would be useful, but you didn't prepare it, you're out of luck. Similarly, if you memorized knock once, and then encounter a second locked door, you're also out of luck. The intention of this approach (so far as I can tell) is to encourage careful consideration of what spells to prepare for a given adventure or adventuring day, in much the same way encumbrance forces players to consider which equipment to bring on a given expedition. In practice, however, it often results in players choosing a predictable slate of spells highly likely to be useful (such as combat spells) at the expense of niche spells that are highly useful in situations that may or may not be encountered. Conversely, it may also put pressure on DMs and adventure designers to cater to those expectations, so the player who rationally bulks up his spell selection with magic missile and sleep doesn't feel cheated.
On the other end of the spectrum, we have characters able to cast any spell they know, as often as they like, until their spell slots or points are exhausted. Obviously, this allows the player great freedom, but at the cost of the planning and resource management aspects, which are arguably important challenges to player skill. It also allows spellcasters to spam particular spells once it's clear what's most effective in a given situation, which may or may not be a problem for you, but definitely contributes to power creep.
Is there a way to maintain some elements of the strict preparation system while allowing some of the benefits of improvisational casting? Maybe.
Let's start with preparing spells as usual. A player selects a list of spells believed to be most useful for a character on a particular adventure. A spell may be prepared multiple times if desired. Prepared spells are cast with the standard casting time (usually one round, in basic D&D) and have the full effects according to the spell description.
Now... a caster can choose to cast a spell he or she has not prepared by consuming a prepared spell slot of equal or greater level, losing the prepared spell. A particular spell can only be improvised once per day in this fashion. It also loses all benefits of level scaling; that is to say, any parameters of the spell that improve with the caster's level, such as damage, range, or duration, are automatically limited to the lowest level required to cast the spell. A 10th level wizard improvising a lightning bolt would do 5d6 damage (1d6 damage per level, with 5th level being the minimum required for casting a 3rd level spell) rather than the usual 10d6, for example. If this still seems too permissive, an improvised spell might take 1d6 rounds to cast rather than the normal one round. (This second penalty might be a good idea anyway, to prevent abuse of the improvisation rule by casters who actually are the lowest level needed to cast a particular spell, including all 1st-level casters!)
This would seem to favor non-combat utility spells for improvisational casting, since additional rounds of casting time mean little outside of combat, and the limit on level-scaling is less likely to be a significant handicap.
I wish I even had a group with which to playtest this thing, but alas, I do not. Does it sound like a good idea to you, or unnecessary, ill-advised, or absolutely terrible?