For some reason you seem to be perfectly fine with the dice saying no, yet have an issue with the GM saying no to the exact same thing. In either case, the player action declaration was nullified, so their agency was eroded just the same, right?
I am arguing that the GM pre-establishing these facts do not infringe on player agency, at least not any more than your "roll for it" method would and probably less.
This claim, asserted as baldly as this, seems wrong.
If I toss a coin to see who has to do the washing up, and I lose, that's fair. I took my chances and lost! If we are wondering who is going to wash up, and you just decide that it will be me, that seems pretty different!
The use of lotteries, in all sorts of contexts that extend well beyond playing games, is a pretty well-known way of disclaiming decision-making. Losing a lottery is quite different from having someone decide that I will get the short straw.
What has been suggested that the interaction of the player choices and established myth is consulted for outcomes.
Sure.
And sometimes this is purely "mechanical" (not in the game-mechanics sense, but in the sense that there is an algorithm-like procedure that determines a unique outcome). For instance, the player declares "I look through the archway" and the GM consults their notes to see what it is that one will see beyond the archway.
Other times it's not, though. The Alarm spell example was chosen by me because it's not (outside of some pretty narrow contexts, like dungeon rooms with single doorways).
Some sort of GM decisions might indeed erode player agency, whits others could empower it. Like I have been talking about information that is at least potentially knowable to, and actionable by the players. "There exist this specific assassin sect with an ability gained via a defined method that allows them to walk through magical defences without trigging them" is that. "The dice say no" really isn't.
<snip>
It is just that if the GM has coherent reasons for their no, then those can become leverageable by the players, thus allowing an avenue of agency, in a way that cannot happen with mere randomness.
How secretive the sect is would be determined by the GM. Whether the PCs already know about it would be determined by the intersection of the established setting background and the PCs personal background, perhaps taking account some mechanical thing such as skill choice. Even a roll might be involved, which should make you happy! But the thing is, this is an established thing about the setting. Depending on the choices the PCs make on their adventures they might have already stumbled upon it.
<snip>
The GM saying no for valid reasons is no more infringement on your agency than the dice saying no is. Ultimately the situation is the same. Like if there are six boxes, one of them containing the MacGuffin, and the GM has predetermined it is in box four and you declare your PC looks in box five, and the GM tells the MacGuffin is not there, then you think the GM infringed upon your agency? But if instead of predetermining the location, the GM lets you roll d6 every time you open the box and on six you find the MacGuffin and you roll two when you open the box, your agency was on infringed? Like what?
The end result is the same. It is just that in the first scenario it is at least possible to foreshadow the location of the Macguffin.
<snip>
What could not happen in such paradigm however, is the PCs deciding to visit the city of Nyx seven sessions ago, and whilst pursuing other matters stuble on the secrets of the assassin cult, because that information didn't exist then.
The play you seem to be describing here is what I have sometimes called "puzzle-solving" play; and have sometimes called "playing to learn what is in the GM's notes". The clearest example of advocacy and advice for this sort of play that I know of is found in classic D&D rules - eg in Moldvay Basic, Moldvay notes (p B4) that eventually the players' map will come to more-or-less duplicate the GM's map.
Now I think I can come up with an example of play that nearly everyone would regard as unfair or unreasonable: the GM builds a dungeon (call it D1); there is a trapped treasure that can only be obtained by speaking a magic word; the magic word is found in a second dungeon (D2) that the GM has (fully or partially) built; in the first session, the GM tells the players that their PCs are at the entrance to D1; the players then have their PCs enter D1, and are promptly killed by the trap on the treasure.
This is unfair, because although
in the fiction, the secret word was knowable - all one has to do is go to D2 and learn it! -
in the play, the players had no chance to learn it. The GM framed their PCs into a loss.
How close is the secret assassin's guild to the example I've just given? To what degree is it
really true that the players could have had their PCs learn about this secret guild? Play time is finite; what the GM makes salient - via exposition and framing - is hugely significant in what choices the players then make.
There is a huge range of possibilities here. But for information about the setting to be amenable to being leveraged by the players, there needs to be far more done than simply having the GM write it down in their notes.
But what practical difference does it make in this randomised fiction after model? Like it does not matter if you take specific steps against specific foes in your camp, because we just randomise it anyway. Then bad results just means someone else attacks you.
I am not sure how such information would really even matter execpt as flavour in this fiction after system. Like sure because this time the camp event roll said you were ambushed despite your warding spell the post hoc justification migh be that the ambusher was an Assassin of Nyx, who can bypass such magics. But what do you do with this information is such a system? Next time the dice roll low and you get ambushed, there will be some other post hoc justfications which again is just mostly flavour.
I don't what resolution system you have in mind here.
The whole existence of assassins of Nyx is only invented seven sessions later to explain the random result of the warding spell not protecting the PCs from an attack.
I don't know what RPG you have in mind here. What you've reminded me of is
this, from John Harper:
I've seen people struggle with hard moves in the moment. Like, when the dice miss, the MC stares at it like, "Crap! Now I have to invent something! Better make it dangerous and cool! Uh... some ninja... drop out of the ceiling... with poison knives! Grah!"
Don't do that. Instead, when it's time for a hard move, look back at the setup move(s) you made. What was threatened? What was about to happen, before the PC took action? Follow through on that. Bring the effects on screen. Bring the consequences to fruition.
You seem to be envisaging the sort of bad GMing that Harper criticises, rather than GMing that actually conforms to principles like
make a move that follows.
