Yeah, having variable failure states is the interesting part of "fail forward" design. There's no reason that couldn't be combined with a design that respects player agency, it's just the opposite of the rules-light design those games have usually embraced. You'd want to spell out the effects that can happen on failure preemptively, and probably even give players some ability to choose their preferred failure state or mitigate failure states they particularly don't like. You need more detailed skill descriptions, with more outcomes specified preemptively. There's even space for some DM improvisation there for outcomes outside of the usual failure results, as long as it's called out before the check and exists as a known possible outcome or is something players can opt into.K wrote:I think anything where the results are based on DM improvisation is doomed to feel like there is no agency. See Bearworld.
That being said, "Yes, and..." is the rule of improv, and RPGs are a form of improv. That is why old-school DnD had so many tables. Tables were the original procedural generators, and rolling on those is a lot more fun than "you fail, game stops."
The other thing that really should be considered more clearly is the role time plays as a resource. Take 10/20 or equivalent rules mean most of the binary "checks to see if the adventure continues" situations actually amount to a time penalty. That's only interesting if time is a meaningfully limited resource, which mostly comes down to scenario design right now. More of that could be offloaded to resource systems (fatigue, ability durations, etc.).
There's also the problem attitude that rolling dice is in and of itself interesting, which prompts GMs to call for rolls for situation that don't actually merit them. I think your take 10/20 rules need to be less a thing a player calls for and more a result of the game state. You should have to declare that there is a time pressure, or a stressful situation or whatever and otherwise have the default result be the highest possible player outcome.
Checks for "does a character know something?" are the biggest binary "does the adventure continue" check, and need to be rooted out or otherwise accounted for. You can do it in scenario design, by just telling GMs never to lock progression behind those checks, but it's always a risk to offload that work on to GM shoulders. At the very least, you'd want to give players the ability to spend a meta currency for success on those checks.
The key is the player point of interaction. If you know the possible outcomes of trying to pick lock vs. casting knock vs. bashing the door down, you can make a decision about which results have the greatest upside and which risks are least bad for you. Once you've decided on a risk to take, having a randomized result for how that risk plays out makes the outcome interesting, and leads to emergent situations that are fun for the GM and players to work through.The Adventurer's Almanac wrote:This sort of sounds like PCs going from random table to random table, though. Maybe buttressed by an encounter with a statblock. I really enjoy random tables, but this sounds excessive. Maybe I'm misunderstanding.