There is a fundamental divide about what player agency means here. I think for Pemerton it means being able to shape the story, whereas I suspect for you (like me) it has more to do with freedom to operate freely in the setting (but not to shape or control things typically held under the GMs prevue: for example what threat lies in yonder cave). You guys can debate the meanings of the term agency all day long, but I think in the end it boils down to you have different preferences and something that is seen as a moral good in gaming (player agency) is being vied for to win a discussion about play style. These kinds of arguments are generally why I am wary of internet forum gaming discussions, or at least wary of the rhetoric we tend to encounter on them.
I honestly don't think there is a real conflict in these definitions, though. Freedom to operate freely in the setting is a manner of agency. Narrative authority by players beyond declaring what their characters do is another manner. A further manner.
I'd say that allowing a player to declare actions for their character is almost the baseline level of agency, and if it is absent then likely something has gone very wrong.
Games that allow players to more definitively shape the fiction than just that baseline of declaring actions allow more agency. I don't even really see how this is up for debate.
It seems that because the conversation largely assumes agency is a good thing, that any reduction of it is a reduction of good, and no one wants to admit that their game is less good......so they insist that their game has all the good. All the agency.
But it's just not the case. My 5E game allows less player agency than my Blades in the Dark game. That's not a bad thing. The game is designed that way, after all.
Loss of Agency is taking the players ability to change things away from them. Fudging a die roll because I've overtuned the encounter or undertuned the encounter doesn't prevent them from doing anything they decide to do. Now if I'm fudging rolls to force them to do something I want them to do or railroad the encounter to a predetermined end, or control how the fight ends, sure that's taking agency. Fudging a roll in combat to fix DM Human error's doesn't take away agency anymore than deciding if they fight a monster they can't possibly beat or just throwing an orc at them.
By that logic if I screw up and throw a monster at them that they can't possibly damage in a situation where they can't run away, dropping some of the monsters special abilities that the party doesn't know about so the fight can possibly be won would be taking away their agency. I'd argue it would be giving it back. If DM isn't allowed to adjust as the game goes on he's irrelevant and shouldn't be there. Might as well play a video game.
Most of these discussions seem to assume the DM never makes a mistake and any change in combat or too a die roll is taking control away from the players.
I can understand this, and I've done it myself, although more so in the past than I would today. Because it absolutely is reducing the players' agency. If the fight is too easy....then the fight is easy. Let the results stand. I get the idea of trying to preserve the status of a threat that's been built up, but the game isn't about preserving my ideas of what it is about.
If the encounter is too hard.....then maybe they need to approach things differently than trying to fight? Maybe they need to negotiate or run away?
Typically, the GM would only do these things if they had an expected outcome, which could even be something like "after a difficult battle, the PCs emerge victorious!"
Well, I'm clearly weird (which I knew).
Let me try this formulation: "Success" is "getting what you want, and not what you don't." "Complicated success" by adding something the player/character didn't want, turns success into failure.
I have been turning things over in my head, thinking about this, and I have further come to realize that a resolution mechanic that A) had greater odds of uncomplicated success and B) allowed the player/character to choose to accept a complication-esque consequence to turn failure into success would bother me a good deal less. I'm sure a game exists with such mechanics, I just can't bring any to mind at the moment.
I don't think that your definition of success can ever really exist. I mean, in D&D the PCs want to win the fight with the evil necromancer and his undead minions. Ultimately, they do win.....but it took some resources, and no one emerged unscathed. Would you say that they failed to win the fight?
I feel you are applying the concept of success far too broadly. I mean, my PC wants to jump the chasm. That's his goal, and that's what the roll is for. In the case of BitD and most PbtA games, the roll is also folding in several other rolls (the kind typically made by the GM in D&D and similar games) into that roll.
So, to kind of compare it to combat in D&D, the player makes a roll to attack the orc. Then the orc and everyone else involved in combat would get a turn of some kind. The equivalent roll in BitD/PbtA games would encompass the orc's response, and potentially his allies' responses as well, depending on what had been established already. Each individual roll is doing more than what a roll in D&D does.
one that tries to mix failure and success
I don't think that's an accurate way to look at it. Complication is not a failure.
I realize I may be beating a dead horse here, and I don't mean to.....I just think that your take on this is skewed by a bit of flawed reasoning.