3ed - NPC's are given straight 10's unless they are Elite, in which case they use the Elite array. An NPC flat out cannot have a base stat higher than 15 (before level and racial adjustments of course).
Or they can be created like PCs...
4ed - NPC's, other than very specific individuals - are not given stats whatsoever. They have whatever skill bonus the DM feels is appropriate and that's it.
Possibilities for NPCs in 4e included statting as monsters, as Companion Characters, as charater-class-templated Elite monsters, and, of course, as PCs. All of those options include stats. The stats are just arbitrarily assigned by the DM in most of them.
5ed - NPC's, other than specific individuals where stats can be rolled but aren't necessary
Stat's don't have to be rolled, I thought, was the phrasing?
In a previous edition that was published over a decade ago, which has nothing to do with 5E.
But what a book published nearly 40 years ago has to do with 5E is still beyond me, other than a little bit of historical trivia.
5e's all about evoking the classic game, and during the playtest, Mike Mearls seemed to put a lot of contemplation - and polling - into exactly what that was, including going all the way back to the early days.
Maybe. Could've been DNPC (to go all Hero System).
Could've been an NPC someone started playing later...
Was she generated using the same number of points as Xena?
Depends on the system and how it weights abilities... Not in D&D, presumably, but D&D generally couldn't do either of them at all well, anyway.
What about Hercules and Iolus?
Same thing, really. Iolus is a sidekick, D&D generally couldn't do either character that well.
It's relevant to our debate. The reason this debate took this turn was that you asserted that point-buy lets you have the concept you want, and I refuted it by pointing out that the huge majority of possible arrays of 3-18 in six scores was not possible.
Rather, you refuted a straw man, that point-buy would let you build any array of six stats with a range of 3-18 each. Which is very different from playing the concept you want, which, presumably, includes a class & race, and includes playing it in a party, with other character-concepts that might have a bearing on what your numbers actually mean.
Of the two specific methods given in 5e, the variant point-buy method gives the player more freedom to play the concept he wants when he sits down, it can be used to build a fairly large number of different arrays, letting the player pick and arrange the one that best supports the desired concept. The default method gives the player exactly one array, if he takes the standard array instead of rolling, he has 6 different stats to arrange as he likes, giving him essentially 5 meaningful choices in distributing them. If he rolls, again, he gets exactly one array, he just has no control over what that array may be, but, whatever it is, it gives him at most 5 meaningful choices in arranging it - fewer if there are any duplicate numbers.
Rolling might not be a perfectly realistic way to generate either PCs or NPCs, but it is exponentially more realistic than point-buy or array or "every person has exactly 10 in all six ability scores".
Neither PC method in 5e is remotely realistic, both tending to give far too-capable results for a general population, and the fact they including a player arranging the stats prettymuch renders them nonsensical for the purpose (yes, the DM can do it, but a general population isn't 'designed' like that). Presumably, some variation on a system would be used for the general population, if, indeed, anything like an individual system were used, at all, which is, really a bit of a stretch - on one's actually going to do that, I don't think, it seems prohibitive, though I suppose you could fairly easily write a program to psuedo-randomly or methodically create an NPC population in detail.
So the question is realism seems, to me, to become an almost trivially distinction. Clearly, random should have the edge in terms of realism, or, at least, realistic-verisimilitude, but, with the default 5e method, the option of taking the array and the ability to arrange stats blows any sort of process-sim/associated-mechanics/v-tude 'realism' out of the water. It really takes random-in-order, and random generation of other non-character-influenced statistics (like the social class you're born into), before you get a big jump in realism delivered.
There is something, though, to rolling 4d6k3 & arrange, when the traditional assumption is you're coming out of a 3d6 in order population. V-tude, perhaps, or internal consistency. It feels like a model of a self- and circumstance-/merit- selected 'elite' - adventurers.
How stats for the general population are generated (if they even are) has no effect on how stats are generated player characters. As proven by all the different methods available for generating stats for PC's - from the many different rolling options, to point buy, to array.
Sure, but there's a clear relationship between 4d6k3/arrange and 3d6-in-order. It may not mean anything, but it doesn't need to in order to provide 'feel.'
