@AbdulAlhazred asserted a disjunction: GM fiat
or roll the dice.
Absolutely. We have - roll dice for each decision - as a rational exit from the dilemma that I outlined.
He also stipulated certain conditions as underpinning the disjunction: a realistically detailed world, with unconstrained variables, such that solving it as a puzzle is impossible.
We're likely to find some clear water between our views here, as for me all TTRPG worlds have a vast number (literally infinite) unconstrained variables
even if they are not realistically detailed. It might not be worth our debating that, but just accepting that we have differing views.*
If we reduce the number of variables, by reducing the realism of the fiction and stepping up the number of recognisable tropes/conventions/stereotypes, then GM decision-making won't have to obviate player decision-making. This is how Gygaxian skilled play works. It's also how The Green Knight works, although the latter system reaches for a quite different set of tropes, conventions and stereotypes and so sets up the problem-solving in a different way.
From descriptions of play and watching videos, I can see that DW is revealing in the way it manages and is successful with - roll for each decision. In terms of unconstrained variables, it
relocates them (e.g. to 6s or less) and
declares boundaries to the decision space (so that it is vast, but less vastly diverse - I recommend reading Borges Library of Babel to grok that).**
Still, one can see that whereas in an example a wand bounced down a shaft, there were a million (again, genuinely infinite) things that could have happened to that wand: the precise outcome wasn't defined and wasn't rolled for. Most choices might have amounted to different flavours of the same thing - e.g. 'I can't use the wand', or, 'I can use the wand if'. But then, there are so many ways that "if" could have gone - if I wait, if I do X, if I pay Y, if I roll and succeed at Z, etc - and the knock-ons from there.... a vast cascade.
* So I am saying that the intent of - "
reduce the number of variables, by reducing the realism of the fiction and stepping up the number of recognisable tropes/conventions/stereotypes" - rationally only changes one kind of infinity into another kind of infinity. That is true unless you reduce to a stochastic or deterministic mechanic with properly bounded inputs, functions, and outputs. The second infinity is more useful (see below), but one thing it is absolutely not doing is removing DM-fiat. Only refining and relocating it. That is worth doing,
but it is not the same as removing DM-fiat. Rather, we are making statements about better and worse kinds of fiat for given purposes... and there is nothing wrong with that! It is helpful to do.
** Making a space less vastly diverse is extremely valuable. It becomes a helpful subset of infinity (albeit also infinite). Say we have a principle - choose only numbers divisible by 7. There are an infinite number of them, but still fewer than there are numbers overall, and if what we care about is divisibility by 7, then we have done something useful. The picture is not quite so clear with TTRPG principles of course, as we see in debates as to what principles going by the same name include and exclude.
EDITED My view is more open than was captured in my words. Also, removed some text that might not have helped my explanation.