As @Manbearcat states, I don't think you can separate things in TTRPG (there may be a few degenerate cases) into 'game' vs 'narrative', except in pure classic DC play, and even there there tend to be episodes of ambiguity where its not clear which is which (IE when PCs are interacting with each other, or attempting to RP a parley with a monster or something, or even edge case adjudication of things like evasion).Sorry about the delay, I've been traveling.
Why should stakes be set at the level of an action declaration? The game is in stringing together a series of actions that will produce the outcome you want, and the whole creative expression of play, the thing I'm calling "ludic agency" comes in having different possible sets to pick from. Any game that can be resolved in one action is necessarily going to be very low-agency, unless you're playing on a repeat loop and moving the actual goal to something else. Rock-Paper-Scissors is an almost no agency game (you can make a choice, but the choice is meaningless) but it has the potential if iterated between two players repeatedly to be slightly higher agency as you start getting into patterns, strategies or at least yomi.
This is the precise differentiation I'm talking about. Ludic agency is your ability to affect success/failure by adopting different strategies. In order for it to exist more than one strategy must potentially yield success (the situation should be a board state, not a puzzle), and strategies must be able to be evaluated; some must be more likely to succeed (or to produce different, more desirable board states for future strategies to be tried) than others.
You're describing narrative agency here, in that the player has control over narrative outcome in some circumstances. Some narrative agency is intrinsic to TTRPGs in that players can generally set their own victory conditions, a thing that is rarely true in other forms of gaming outside of maybe some simulation focused wargames or open-world immersive sim video games. You're arguing for greater narrative agency, that players should be able to set victory conditions (and seemingly important to your position, influence the nature of failure states as well) inside a wider field of outcomes and more often. I'm saying that is not the same thing as ludic agency, and indeed, can be harmful to ludic agency as it grows by shrinking the possible space for gameplay. A state where all strategies are equally (or roughly equally) viable has just as little ludic agency as a state where one or none are.
I don't necessarily know we're in disagreement here, though I am somewhat unclear on what "narrative trajectory" means here. Given a board state and a goal, trying to best achieve that goal (and given the nature of TTRPGs, possible other, assumed subsequent goals) is the play loop, certainly.
I don't think I have sufficient understanding of your point to dispute this. I think what you're calling narrative trajectory is the same thing I'm discussing when players go about declaring goals? I agree that assessment is necessary for gameplay: you need a metric to be able to produce a strategy, outside of pure exploration play, which I think is a pretty unstable state that generally devolves into more conventional play quickly. That is, using a mechanic for its own sake without pursuing a goal, which tends to quickly become adding a requirement to use that mechanic to some other goal.
You're describing a few different things here, none of which I would consider narrative. Adding a meta-game (a tournament) to a game changes the results necessarily. Strategies have to span multiple, repeated plays, which moves the game away from optimization to game theory. To be completely honest with you, I find poker deeply frustrating for the same reason I find fighting games frustrating: the proposed level of mechanical skill necessary to play the "actual game" is huge, and the activity isn't particularly interesting without being played at that level. That, and I think most gambling games rely on emotional valence of loss/risk tied to money to portray themselves as more interesting than they are. That's a pretty divisive issue in gaming in general though; Shut Up & Sit Down in a recent podcast mentioned struggling with coverage of mahjong, precisely because any stance on gambling would alienate about half their audience.
My tastes aside though, assuming players have the appropriate ability to calculate odds and the relatively value of their board position compared to the field of possible board positions, what you're talking about is fundamentally adopting the correct meta-game strategy or, assuming that problem is knowable and solved, yomi. I don't think it's necessary to yield a narrative component to either of those: when faced with yomi, you're trying to find the quickest way to break it into a state that does yield to conventional strategy for as long as possible, or if you can't find a break that way, to continue it as long as possible until such a break can be achieved.
In any RPG where narrative is truly important at all it will mix with more traditional game state considerations, like resources and tactical positioning, or any of the situations I mentioned above, plus other similar ones. In fact trad play evolved from classic play precisely as the clearly defined game state and play environment broke down into a more nuanced set of mixed game/narrative situations such that you can no longer say what is game state and what is related to story or plot anymore. In early trad play this often manifested as players picking goals for their characters which aligned with game-mechanical considerations (IE picking a favored enemy which was likely to maximize the beneficial use of that mechanism, or to focus on any situation that might bring undead into play as an opponent when the party is well-supplied with clerics and paladins). Over time more games, and now to a degree even D&D, have graduated to where literal narrative player goals can translate directly into strategies, such as character build, etc. These kinds of things expect at least some degree of agency to say 'yeah we can pick a path that leads to undead' or whatever. You can definitely call this 'ludic', but it is ALSO narrative! I can decide for purely aesthetic reasons, based on my conception of what should constitute the trajectory of play, that I want to take on undead, or that my character is divided in mind about the integrity of his patron, etc. These things can then be given shape within the mechanics of the game, in at least some systems. The two can inform each other. These types of games demand a lack of adherence to GM provided narrative, it just won't really work. This is the difference between narrativist/SN play and trad (even fairly sophisticated trad) play.