Getting back to the main topic I do not think Wizards should look back at previous versions either with an eye to avoid or bring back particular sorts of mechanics. There is such a vast number of new players who have no experience with previous versions of the game. They should listen to what those players want, come up with designs, and test them. Us old heads are not particularly relevant (whether we like 4e or not).
I mean...this reads a bit like saying we must, without exception, totally abandon
all ideas from the past 20 years of car engineering and design in order to make cars that appeal to the young, hip car buyer. Just because something is old does not mean it has no value. Trying to make an absolute clean break would be as bad as trying to change nothing at all--worse, most likely, since it is old hands who predominantly fill the most important role, as Dungeon Masters.
There's stuff to learn from asking tough questions and getting good data about 0e. The same may be said if literally every edition. From the answers and data, you then establish clear design goals, and critically examine how to accomplish those goals. Once you have your system, actually bloody TEST it--not just by saying how it feels, but by collecting data and seeing if it does, on average and in general, do what it's supposed to do.
And if it just so happens that the data leads you toward a solution that resembles a prior edition,
so be it.
I don't think that WotC has been catering to old fans at all: not since the D&D Next Playtest when they were courting the Pathfinder and OSR crowds. Since then though? It seems mostly oriented towards the Critical Role crowd and the newcomers.
The Banneret seemed pretty transparently an attempt to reach out to both 4e and OSR fans...but only
maybe successful with the latter. The focus on popular classic adventures of yesteryear also looks rather a lot like still putting pre-WOTC fans at least reasonably high on the priority list. What things would you say they've done to OSR fans that look anything like the snubbing 4e fans got?
I believe the problem is properly defining the spectrum a group want characters to fall in, and the features they value in selection. For example, I decided I wanted ability scores to net to +2, with nothing worse than -3 and nothing better than +3 at 1st-level. Additionally, I wanted to avoid overshadowing by having the ability scores for all characters sum to the same total. It was fairly straightforward to design a 12-card deck to draw from without replacement, drawing and summing two cards for each ability in the order drawn. I could instead have listed the complete set of arrays and have players roll for one at random for their character. Both give effective, easily generated, truly-random characters. (I suspect what you mean by truly-random isn't to do with random, but to do with yielding both arrays the group enjoys playing, and arrays they don't. That's a choice about the spectrum arrays will fall in, not really the generation methods.)
When others (this isn't my view, I VASTLY prefer point-buy) tell me what they want, many of them make it very explicit that they want true, genuine, uncurated randomness. They want a distribution which favors neither good nor bad, and which they have no way to know in advance whether it will produce excellent, mediocre, poor, or uneven results. That they must be able to be genuinely
surprised by these results, and cannot, even in principle, meaningfully predict how things will end up, even with partial data. That is a pretty reasonable gloss of "true randomness," e.g., the values generated must be wholly independent from one another, each randomly generated without bias, and drawn from identical populations of possibilities. IOW, no "if you have X bad stat, you automatically get (N-X) as a good stat," and no "drawing cards without replacement," as in your example, since that means you can with high accuracy predict future values solely on the basis of the first few current values. Such fans expressly want it to be the case that the game itself is designed to support BOTH "I rolled 9, 7, 5, 8, 9, 8" AND "I rolled 18, 15, 12, 14, 17, 14," at the same time and table, no wrinkles, no hard feelings, no wildly divergent experiences. And that may be an impossible request, particularly given that many
other players (such as myself) want a well-balanced experience where everyone gets an equal opportunity to excel and big numbers correspond to sizable benefits (such that one must generally focus and think about how best to use the benefits one has, rather than simply being more or less equally effective at all tasks.)
If one defines the problem as - can I have characters, some of which have far lower modifiers than we want to play and some of which will be mechanically ideal - then any 'solution' is going to produce that. If one instead defines the problem as - can I have characters falling fairly across the power spectrum that we want to use - then good random solutions are available.
Okay. How do we then square the fact that there are (quite a few, apparently) people who want the spread to be "I literally have no idea whether the result will suck or be amazing but overall it will average low to weak benefits" with the fact that there are people who don't want it to vary
at all because such variance is unfair? That is, there seem to be dramatic disagreements about whether there should be allowance for variation
at all, or whether variation should be hard-required and dramatic.
The problem is having a clear enough definition for what = good in this context.
Unless we have a literal actual split in the fanbase as to whether characters should even BE expected to be "good" at character creation. Which, well, people tell me is the case. People claim to want to not know for sure whether their character will be good at
anything at all.
Again, I don't believe it is the 'randomness' that is the heart of the problem you describe. The PHB methods have two shortfalls. Foremost the PHB doesn't tell groups why they should choose one method over another. It should start with motives, not methods. Second it offers only a very limited set of methods, when the 'tech' is available to offer more.
I can only go by what people explicitly say, and people have told me many, many times that methods like yours are insufficiently random--that they just look/feel like (effectively) drawing an array out of a hat, not truly making a character that is unexpected. The high, even extreme emphasis is not just simplicity, though simplicity is in there, but rather that being fed an
expected character, even one that is not absolutely foreknown, ruins the experience. I believe the phrase used in a thread either this year or last year was that such characters are "born lucky" in such players' eyes, and playing someone "born lucky" just feels like a foregone conclusion of success.
Whereas to again to compare to me, someone who deeply values balance and equal opportunity, I consider essentially all forms of rolled stats to be "ability roulette" and rather hate them a lot, ESPECIALLY when they theoretically produce better average stats. I feel I am going to he punished no matter what, either I accept "weak" PB stats or I accept that my awful luck will give me
technically, theoretically viable but crappy results, worse than if I'd just settled for PB. Or, if you prefer, rolling stats
at all makes me feel "born unlucky." And even if I get great stats (which does, rarely, happen) I'll feel terribly guilty if even one person has demonstrably worse stats than I do.