Combat in traditional swords & sorcery adventure games is typically described as an abstraction, rather than a simulation. The significance of this is sometimes misunderstood and conflated with a perceived absence or presence of realism. For instance, a common complaint is that the one minute combat round is unrealistically long, to which the typical Gygaxian response is that the abstract approach "assumes much activity during the course of each round". However, this does almost nothing to address the real issue, which is that "one telling blow every sixty seconds" arguably lacks verisimilitude. That is to say, the one minute combat round is both abstract and unrealistic, the former having little to no bearing on the latter. The criticism must be contested on its own terms to be successfully rebutted.
For other elements of the traditional combat system, the degree of abstraction does have a direct bearing on the level of realism. Of these, hit points are perhaps the most widely identified as being unrealistic, which is a pity and possibly related to their widespread currency and representation in computer games where a "hit is a hit", but meaningless beyond being a step closer to death. In fact, the truly abstract nature of hit points allows the game master the freedom to determine the extent to which realism will be a concern. When a character loses twenty of thirty hit points to a single attack, it is up to the game master to describe the event, and also to decide if there are any effects beyond their loss. To put it simply, the value of the abstraction is in its ambiguity.
Which brings us to the subject at hand. The term "armour class" as it first appears in Dungeons & Dragons is derived from the nomenclature of Chainmail, which describes troops, weapons and armour as being divided into classes. The closest it comes to using the actual term "armour class" is with "armour classification" and "class of armour worn", on pages thirteen and forty one, respectively. There are eight classes of armour in both games, or more accurately there are four classes and four modified classes. These are: unarmoured (9), leather (7), mail (5), and plate (3); each is lowered by one if a shield is added [i.e. 9/7/5/3 becomes 8/6/4/2] and every class is five percent more likely than the last to entirely negate the chance of a hit. What this means is that each unmodified category of armour makes it ten percent more difficult to land a blow. Since fighting-men similarly advance ten percent in fighting ability every three levels until sixteenth level [i.e. a total of +40% in four equal steps between levels one and fifteen] there is an obvious dichotomy of fighting ability relative to armour class.
With the advent of Advanced Dungeons & Dragons, this elegant symmetry was abandoned to accommodate a greater number of armour classifications and a more granular progression in fighting ability. The protective value of mail and plate was increased by reducing default fighting ability by ten percent, but only by five percent for fighters and clerics, the separating of normal men from first level fighters (which carried over into subsequent versions of Basic) and further stratification of monster fighting ability by hit dice (which was partially carried over). The weapon versus armour modifiers, which were derived from Chainmail for Supplement I: Greyhawk, were also included. These last can make weapon selection more interesting, but by the same token they seem to increase the degree of simulationism in the combat system. Quite what they simulate is open to question, as they are implicitly optional and appear to have little basis in any historically authentic relationship between weapons and armour. On the other hand, the modifiers do tend to make two handed weapons a more viable choice.
One significant thing that Advanced Dungeons & Dragons carried over from Supplement I was the idea that strength should not only increase hit probability, but also the amount of damage delivered. Along with variable damage dice for weapons, this was yet another step away from the initial abstraction, and requires some interpretation to be reconciled with the assumptions of the broader system. In general, armour is overcome in two basic ways; it is either bypassed or it is penetrated. Since strength is not only an indication of the raw power with which a combatant may deliver a blow, but also contributes to its speed and precision, there is a logic to a hit and damage bonus even when facing unarmoured opponents. However, that brings up another question; if strength and weapon class directly increase the amount of damage inflicted, should not armour class also reduce the amount of damage sustained? Is it reasonable for a successful blow against an unarmoured individual to have the same potential for damage as one delivered to a character fully encased in plate armour? Every addition a step away from the initial abstraction...
What to do about this simulation creep, then? Where to draw the line? For my home campaign, I decided that I would take armour class literally and partially divorce precise forms of armour from the traditional ten categories. Instead of looking at AC6 and seeing "studded or ring armour and shield" or "scale armour", the exact armour type is left vague. Perhaps it represents a mail haubergeon, a rusty mail hauberk, or a bronze hoplite panoply. There are limits to what is plausible, but I would argue that in assigning the conventional ten categories default numerical values in accordance with their relative position, rather than as a reflection of the armour they nominally represent, a greater degree of flexibility and robusticity is achieved. For pricing the armour, I took the rough "doubling" effect evident in the original version of Dungeons & Dragons, and extended it from the three basic armour types to the nine advanced armour classes, which gives roughly similar approximations. Since, in addition to unarmoured, there are nine armoured categories, I decided to divide them into three groups of three for the purposes of movement and damage reduction. The former was a fairly straightforward assignment of 12/9/6 to the three groups, respectively. The latter required a bit more thought; the temptation was to use a 1/2/3 progression, but in playtesting I found that a 0/1/2 progression was generally preferred on account of the difficulty of inflicting damage with a dagger. Anyway, for your entertainment, I put together my variant in a pdf: Armour.