The concept of a saving throw both predates and appears in Chain Mail. Of course, the Fantasy Supplement contains several references with regards to spells and dragon breath (pp. 31 & 35), but its first appearance in Chain Mail in general is in the siege rules, where a “rock dropped down a ladder will kill the first climber, and the second and third men on the ladder must roll a die to see if they survive, 1-3 saving the second and 1-5 saving the third” (p. 23). This idea that the defender makes a roll to save himself, rather than the attacker rolling to determine if a kill is scored, takes the onus away from the aggressor and assigns it to the victim. In most cases the ability of the attacker is of no consequence, he is assumed to either be one-hundred percent effective, or else the probability of his action failing as a result of his own insufficiencies is accounted for in the saving throw. A notable exception to this is found in Supplement II: Blackmoor and the monk class, which can make a saving throw to negate a successful enemy attack roll with a missile weapon; these rolls are not opposed, however, and the probabilities of success are largely independent of one another.
Analogous to class based saving throws are the various other fixed probability rolls in Dungeons & Dragons, such as the chance of springing or detecting a trap, detecting and opening a secret door, and gaining surprise, for example, all of which are described together in Monsters & Treasure (p. 9). Indeed, initiative and just about any action that has a probability of failure and success could be described as a form of fixed “saving throw”, the consequences of failure being usually undesirable. However, the five class based saving throws (Death Ray or Poison, Wands, Stone, Dragon Breath, Staves and Spells) are differentiated from these in that the probability of success differs by class and increases as characters advance in experience levels. The categories were somewhat changed for Advanced Dungeons & Dragons, but are otherwise pretty much the same. It is very likely that these categories were organically developed and represent specific things that Gygax or Arneson thought of as particularly within the realms of heroic ability. Indeed, who can hear of the “death ray” and not think of Conan in Red Nails facing off against Tolkemec and his jade-hued wand?
Further elucidation, or perhaps post facto rationalisation, as to the meaning of saving throws, and indeed fighting ability, hit points and other improving characteristics, is provided in the Dungeon Master’s Guide, where it is explained in no uncertain terms that “the accumulation of hit points and the ever-greater abilities and better saving throws of characters represents the aid supplied by supernatural forces” (pp. 111-112), which is also related to the idea that “whether or not the character actively professes some deity, he or she will have on alignment and serve one or more deities of this general alignment indirectly and unbeknownst to the character.” (p. 25). That is not to say that saving throws are entirely unaffected by natural ability, since dexterity and wisdom certainly affect saving throws, and they can be entirely based on an attribute score, as with a system shock roll, which is specifically denoted as a saving throw in the Player’s Handbook (p. 12). Indeed, both the spells dig (p. 76) and phantasmal killer (p. 98) have saving throws that are simply attribute tests, where the player is required to roll under the relevant attribute on a specified number and size of dice.
Testing an attribute became a standard method for task resolution from around 1985 and was included as an optional rule in second edition for making saving throws. The recategorisation of the five specific instances to the three broader defences of fortitude, reflexes, and willpower in D20/3e represents a further move away from sheer chance or divine sponsorship towards emphasis on the importance of physical ability, neatly tied into the core mechanism of that iteration of Dungeons & Dragons. Direct magical protections can significantly improve saving throws in any edition of the game, but it is also interesting that regardless of the measures taken, class based saving throws always succeed on a 20 and always fail on a 1, which was an idea later borrowed for the attack roll as well, though interestingly not for skill checks. Personally, I am not inclined towards the “there is always a five percent chance” model, though the attraction is understandable there is a not inconsiderable case to be made for “assured results” as an alternative. Certainly, some of the most potent or powerful magical spells and effects do not allow saving throws and are rendered considerably more fearsome thereby; sleep, slow, and energy level drain are of this order.
As with many things in Dungeons & Dragons, saving throws are abstractions and lend themselves to interpretation and rationalisation after the fact. If a giant spider scores a hit on a fighter, but does not slay him by hit point damage and he makes a successful saving throw versus poison, then that can represent anything from a spider bite that failed to inject its venom to the character successfully fending off the attacks of the monster. The Dungeon Master’s Guide devotes a good page or so to the subject (pp. 80-81), with a lengthy preamble concerning the game function of saving throws and alternative ways to rationalise them by class and level. A section of interest concerning the potential modifiers follows, giving the example of a character standing in a pool of water as potentially more susceptible to lightning attacks and less vulnerable to fire attacks. In general, the game master is encouraged to adjust saving throws however he sees fit in accordance with what he feels is balanced, but urged not to remove all chance of failure or any chance of success, though such is permitted. Anecdotally, it would seem that such adjustments are little used, but perhaps this is something that should be more widely encouraged.
When it comes down to it, a saving throw is much like every other abstract and randomised element of the game, a roll of the dice to decide an uncertain outcome by means of probability. There is no reason that the game master could not simply assign the probability in every case on an individual basis, the charts are there as a guideline and aid to reduce his workload, as well as promote a sense of consistency and fairness to the other participants. It is somehow more acceptable for a base probability to be established by level, class, attributes and equipment, subsequently altered by the game master to account for circumstances, than it is for him to assign it out of hand, a sense of lesser arbitrariness, perhaps. Whilst there are aesthetic reasons to murmur against the concept of a single saving throw, modified by the above factors, there is little by way of practical effect. Even within the seemingly esoteric saving throw tables of Advanced Dungeons & Dragons there are quite simple mathematical patterns to be discerned and it is questionable as to whether there is much more to them than that. There is certainly something attractive to me about the idea of an unspecified generic saving throw as a baseline concept.