if all fail forward is is setting the stakes, or setting them a particular way, I'm not really sure I grasp what it is. But by manbearcat's example, there is clearly an implied apparent stake to the failed roll (you fall down the ravine). That is the threat everyone discerns going into the roll. What seems to be happening is the actual stake (the side step) is that falling down the ravine was never really a potential outcome because what is really on the table is losing your divining rod. To me that reads like, the GM is altering the stakes to suit the drama of the situation and keep things going forward, when a more standard reading of a failed roll would be falling. True, he may have set those stakes in advance, but it is still a bit of a sidestep because he is circumventing the obvious outcome of the failed roll for a more dramatically appropriate one
I've started with this passage in my response to some of the recent posts, because I think it is key.
In a "fail forward" game, the table understands that the result of a failed climb check need not be falling. So it is not the case that "the threat everyone discerns going into the roll" is that the PC will fall; it may well be that the threat everyone perceives is that the PC will lose his/her diving rod.
In Burning Wheel, the official rules state that the GM must clearly set the stakes in advance of the player rolling the dice. In the GM's advice, Luke Crane notes that in his own game he often doesn't do this, and establishes failure consequences only if required to by a failed roll. He goes on to say that this is OK at his table, because his players trust him and he has a good rapport with them and they tend to have a shared sense of what is really at issue in the fiction. But he reiterates that the official rule is a good rule.
At my table, I tend to play more like Luke Crane does, than in accordance with what he says. Which is to say, I often leave the consequences of failure unstated but implicit in the shared sense of what is going on in the ingame situation. Like Luke's players, my players trust me and we have a good shared rapport based on many, many years of RPGing together. When my players are contemplating some course of action for their PCs, and putting together suites of abilities to build dice pools (in BW) or get bonuses (in 4e) they will often speculate about the evil consequences I might inflict on them for a failed roll. They don't generally assume that it will simply be falling down the ravine, if the Climb check is failed.
Failing a climb roll involves the person climbing failing to climb. It can be no progress or a fall. Dropping a divining rod is failing to hold on to it, not failing to climb. Falling into the ravine and dropping the rods are full effects for failure, but they are full effects of failure for two completely different things.
In my opinion, mechanics should almost always test what they are intended to test. Climbing should test you climbing. Failure should involve that test. Failing to climb or falling.
Well, this relates to the issue I posted about not too far upthread (maybe a page or two). Different games have different rules.
In AD&D, for instance, as in Moldvay Basic, there are clear rules for what a failed climb check by a thief PC amounts to: namely, falling.
But in 4e, there is no pre-defined consequence for a failed Athletics check in the context of a skill challenge.
And in Burning Wheel, the rules expressly provide that, when a check is failed, the GM is able to ascertain the failure by reference either to intent, or to task, and is encouraged to place the emphasis on intent.
In Burning Wheel, and even moreso in "free descriptor" type games like Marvel Heroic RP, HeroWars/Quest, and the like, having a good Climb score doesn't just mean that your PC is a good climber. It means that, when you declare actions for your PC that involve climbing, you are more likely to get what you want. The flipside is that, when you fail such an action, you don't get what you want and the GM instead narrates you failing to get what you want. This may or may not involve failing to climb, depending how important the climbing, per se, was to what you wanted. However exactly the failure is narrated by the GM, the mechanics are testing what they were intended to test: namely, by declaring an action involving climbing, you are testing the chances of the PC getting what s/he wants with a higher likelihood than if the action involved (say) fighting (on the assumption that your PC is a better climber than warrior).
In [MENTION=6696971]Manbearcat[/MENTION]'s Mt Pudding example, climbing is just a means to the end of getting pudding, and so narrating a loss of the divining rod - which makes the prospects of getting pudding very bleak - is more significant than simply narrating a failure to climb (= a fall down the ravine).
Would you say my other points conflict with Fail Forward in any way: 2) A sense of the world being separate form the players they are exploring and 3B) the players desire to be in a story
On 3B: if the players really are indifferent to issues of pacing, dramatic tension/momentum, etc, then "fail forward" seems less likely to be useful as a technique. If the players delight in Gygaxian dungeon crawling, with all its puzzle solving and mapping and beating the wandering monster clock, then I'm not sure "fail forward" has anything to offer at all.
On 2: as I've posted upthread, I think there is no particular connection between "fail forward" and shared narration of backstory between players and GM. However, "fail forward" clearly implies that the GM will establish stakes and narrate consequences having regard to the dramatic and thematic concerns of the players (as evinced by their play of their PCs). Thus, if the player chooses to have his/her PC scale Mt Pudding in quest of pudding, narration of failures will likely be framed in some sort of relation to that goal. (Eg as losing the pudding diving rod.)
If the players truly want the GM to narrate the gameworld, and consequences, without regard to these sorts of dramatic concerns, then "fail forward" would seem in appropriate.
My personal view is that
without any regard at all is actually a very stringent condition which I think few RPG games satisfy - look, for instance, at Gygax's city encounter tables in his DMG and I think you will see they are constructed with a very high degree of regard for likely dramatic concerns (eg that PCs should encounter interesting adversaries). But there are matters of degree here, and "fail forward" is a technique in which the GM makes it blatant that the gameworld is being narrated having regard to dramatic concerns.
For instance - to go back to one of my actual play examples - when I narrate the mace as being carried down the stream out of the cave and to the base of the keep, where the servants doing the laundry find it, no one at the table is under any illusion that I rolled for that result on the "maces dropped into cavern streams" table. They know that I narrated this because (i) it gave effect to the failure of the PC who tried to fish the mace out of the stream, and (ii) it increased the stakes for the other two PCs who were dealing with the servants, because one of them was the PC who wanted the mace and the other was the PC who had promised to help him get the mace.
If players don't like having that sort of knowledge about how the GM decided to introduce content into the gameworld, then "fail forward" won't work for them.
You mention the players setting the stakes. Can you elaborate on this? That may be another major point of divergence. One of my big gripes with skills like Diplomacy in 3E was that players sometimes used them to set the stakes or direct the outcomes (i.e. "I use Diplomacy to get to the princess to marry me" where the player is framing the consequences of a successful roll rather than allowing the GM to do so....the wording can lead the GM to believe that a successful roll must result in the princess saying yes to marriage, even if the character in question simply wouldn't' or couldn't do that). It took me a while to figure out why this bothered me, but eventually that seemed like the cause.
In the games that use "fail forward" that I've been referencing - eg BW, some approaches to 4e, HW/Q, MHRP - the GM has ultimate authority when it comes to framing a check.
But the players contribute to setting the stakes in various ways. Eg in choice of skill to roll - Climbing vs Navigation when climbing Mt Pudding will likely involve different fictional contexts for failure, which results in different outcomes (diving rod down the ravine vs . . . ? [MENTION=6696971]Manbearcat[/MENTION] is better than me at wilderness stuff). In BW there are other aspects to building dice pools eg players can augment by bringing in related skills, or by helping one another, but this changes the fiction and so changes the stakes. And by stating an intent for the roll ("I want to get that pudding at the top of Mt Pudding!) the player also helps establish the stakes.
With your princess example, I think there are at least two issues: (i) the overall fictional context doesn't make it sufficiently clear whether or not this is a valid action declaration (eg is it too much like "I flap my arms and fly to the moon", which for most contexts will not be a permissible action declaration), especially because D&D tends not to like a single check governing months or years of activity, which is a more typical time it takes to woo someone for marriage than a single roguish wink; (ii) D&D doesn't really provide much context or support for adjudicating the consequences of a successful marriage proposal to a princess. (There are some obvious exceptions: eg OA, or 4e as part of the paragon tier. I'm sure name level AD&D could handle it too.)
The latter tends to make GMs worry that marrying the princess will somehow break the game. In games with more robust mechanics for handling social and political elements, marrying the princess is more likely to be just another ingame event like acquiring a Heward's Handy Haversack, which can be accommodated and built on without breaking the game.
