Quoting the parts I'm responding directly to:

Quote Originally Posted by Eldariel View Post
*snip*There is first a more fundamental framework we need to work on: namely, the one Ludic refers to. What are we optimising towards? And what are our assumptions for the DM? My default for DM is "neutral DM who neither panders for the party nor builds against them" and default optimisation goal is "campaign arc efficiency"; ergo, a party that can just divine the big bad behind the story, teleport to them, imprison them forever in a ring and then proceed on to the next campaign is better than the party that has to slog through the encounters to get to the oracle, then fight through the BBEG's lieutenants and underlings, then get the McGuffin and then disable them via. McGuffin.

*snip*

Though I think you also want a decent degree of ability to defeat the enemy in cases where that is the campaign arc. If you must kill Zariel in hell, you must kill Zariel in hell; no amount of encounter bypassing gets you around that fact (though it can give you advantage in the act itself). I don't think a single module for instance is solvable via. Stealth without extremely loose stealth DMing (so not with a neutral DM). Many, however, are quite solvable via. teleportation and planar shifting combined with the ability to use a couple of Glyphs of Warding to nova an encounter dead with couple of hundred gold pieces (that you can trivially generate).

*snip*

But it's true that the framework needs at least four components before we can get to the party optimisation framework itself:
1) Combat-as-War or Combat-as-Sports?
2) Pandering, neutral or adversary DMing?
3) Encounter-level or campaign-level optimisation?
3b) Maximal difficulty or maximal endurance?

Are there any aspects I'm missing here?

*snip*

perhaps yoyo healing specifically is common enough a house rule consideration that it warrants mentioning in rankings).
I have some problems with "campaign arc efficiency" as a metric for "better" (as opposed to, for example, maximum difficulty), and it's difficult to articulate why because the issues are all related. Let me make some observations:

(1) If a campaign has a fixed arc with a specific BBEG that can be efficiently solved, and then you have to start a new campaign, then wasn't that "campaign" really just a single adventure? A campaign is what happens when you re-use the same PCs across multiple adventures.

(2) Sandboxes don't necessarily have fixed BBEGs in the first place. You could just have a bunch of factions. Narrative power in this context could mean something like "player ability to take over a faction," or "ability to leverage one's faction to achieve one's character goals."

(3) 5E... doesn't work all that well for Combat As War play, especially by RAW. I say that as someone who loves CAW, but finds that it requires the DM to innovate enormously in the empty spaces of RAW to give players even a moderate amount of predictability (and therefore agency). Super-simple example: 5E (DMG) RAW is that you can typically spot an enemy coming from a mile or more away if there are no obstructions, but of course there's no RAW on how common obstructions will or should be, and common practice (as opposed to RAW) seems to be to just make everything tiny enough to fit on a battlegrid with 5' squares. If the goal is "neutral DM", then you can't really have Combat As War as a goal too without making a whole bunch of other axes about DMing style.

Maybe we should talk more about this part, but I'll give one other example of something I think doesn't fit well into 5E: surprise attacks on PCs during downtime. In principle there's nothing stopping an enemy of the PCs from collecting the bodies of various foes the PCs have defeated (a few beholders, a couple dragons, a Mind Flayer Arcanist or six) and Wishing them back to life over the course of a few weeks, then one night after midnight Teleporting everyone to wherever the PCs are, throwing up an Antimagic Field, and having the monsters eat all of the PCs who are present. (If PCs aren't glued to each other during downtime this scenario becomes "eat one of the PCs.") From the perspective of gameworld logic this could happen, but in practice we don't do this, and I think one reason is that it doesn't fit into the 5E idiom. 5E makes implicit promises with its approach to saving throws, easy resurrection, etc. that your PCs will have a fighting chance, and having the DM arbitrarily decide one day to kill you isn't fun. Having the DM arbitrarily decide not to kill you is just as arbitrary, but more fun, so we do it. If 5E were a game where things like divination and scry-and-die were meaningfully integrated into gameplay so that this became a game of play and counterplay, or politics and intrigue, then maybe even dying in this way could become fun, but it's a game without rules for stuff like this beyond handwaving: "the DM will decide what makes sense," so in practice we just don't do this stuff, unless the DM innovates heavily beyond the borders of RAW.

(4) Solved games cease to be fun. I don't play Tic Tac Toe any more except with little kids who don't know how to play it properly. Or rather, I only play Tic Tac Toe variants which increase the difficulty to the point where it's no longer solved for me: three-person Tic Tac Toe on a 5x5 or 4x4x4 grid, for example, is still interesting.

So anyway, one reason I don't really love the idea of optimizing for campaign-arc efficiency is that it sort of implies that your goal is to make your own adventures boring. In a sandbox this isn't as much of a problem assuming you can just seek out greater challenges, but that sort of ties into CAW play which 5E isn't good at (maybe we should talk more about this part?) without innovating beyond RAW, and we've said RAW is a goal.

With that in mind I'll list additional axes that you'd have to consider if you did want to do this analysis:

(1) How does the DM approach spell component price elasticity and availability? (I.e. can I kill a red dragon and reliably turn into 100,000s of gp into 100s of gems with which to Planar Bind to kill more red dragons, or are there only so many gems available?)

(2) Is it a sandbox with PC-driven goals, or is there a DM-created plot which players are expected to participate in?

(3) What kinds of hirelings are available, and how does the DM handle morale and loyalty?

(4) Does the DM have a good way to run mass battles, or do they just get handwaved?

(5) What's the availability of poisons? (Can they be magically created via spells like Polymorph + milking the Polymorphed creature? How about Creation for toxic plants?)

(6) Is there some kind of game structure for players to engage with when it comes to investigations and divinations, or is it just "I cast Clairvoyance and the DM makes something up about what I overhear"? If a player intends to spend a week casting Clairvoyance and Arcane Eye over and over to learn everything possible about a given locale, do they need to have 500 ad hoc conversations with the DM while the other players sit around being bored?