A lot of this depends on what the players are trying to have their characters do.
Sure--but participation in good faith is a pretty major filter for much of what the players might try to do. Some games, good faith is compatible with anything goes, and be it on the DM's head for such a choice. Most games, there are gonna be some limits. I know you personally favor a...let's say "anything goes, but don't take enough rope to hang yourself with" kind of approach, which I personally never would. I find it much, much too likely to foster bad feelings or result in un-fun consequences. But within what I consider relatively reasonable bounds (e.g. "please don't play evil characters because I, as DM, won't be able to run a fun game for evil characters" or "it's okay to have PCs disagree, even 'fight'
briefly over what to do, but keep it minimal and figure it out without taking up too much session time"), I'm willing to hear out most things my players would want to do. They just need to be ready for there to be serious consequences if they do something stupid, reckless, or hurtful.
Thinking about it, my by-far most common answer is neither yes nor no, but "maybe". If they try something off the books that seems to make in-fiction sense, I'll quickly dream up some odds of it succeeding and then the dice get rolled.
Sure. "Maybe", IMO, usually cashes out as one of those things I mentioned above: "yes, and", "yes, but", or "no, but", just with dice as a filter. Having the dice decide whether something
succeeds is distinct from whether you say yes or no.
The big thing is whether the dice are actually offering a meaningful chance of success, or are functionally being used as "no, but with more steps". E.g. if you require that a player score three back-to-back natural 20s to succeed, you're
functionally saying a flat "no" even though there's a slim (1/8000, 0.0125%) chance that it could still happen. IMO, if you aren't going to give it more than at least a 10% chance to succeed, just say no--it will be functionally no different in most situations, and substantially more forthright than the pretense of success.
A very recent example from my game: party of 7 spent several days in two large-ish canoes making their way up a slow-flowing river through a swamp. No solid ground to camp on, and not enough room in the canoes for people to sleep. So, in a nice bit of pre-planning they had brought a number of large planks and some extra rope so as to raft the two canoes together into a platform for when they wanted to camp.
Knowing boats as I do, I-as-DM could see all kinds of ways this idea could go all kinds of wrong...and also see that it might work quite well as intended. And so, "maybe": I got them to roll on a the-higher-the-better sliding scale for the first few nights to see how well this worked out; they did OK and so after that I ruled they had it figured thus no more rolling needed.
Had they tried to bring enough wood to make a cottage, however, that would have been a flat "no" if only because there wouldn't be enough room in the canoes for both the wood and the people.
No real complaints here then. That actually sounds very similar to stuff I've done with Dungeon World. E.g. one of my players (a Bard) wanted to learn how to shapeshift for purely RP/exploratory purposes--no combat forms, they explicitly brought that restriction from the beginning and never wavered on it. Since we had a Druid, who very much had a desire to teach others as part of the character, I said sure, but it would take time and effort from both of them. Failed rolls would mean there was some kind of mental roadblock one or the other needed to overcome. Partial success meant training was fine but slow (and thus might
benefit from something concrete), full success meant a clear learning milestone. After a few months of occasionally stopping to have shapeshifting lessons, I said the player could buy the Druid shapeshifting move via the Bard multiclass options--they'd earned it by doing the work to make it happen, driving personal story forward.
Had the Bard tried to
add the whole Druid playbook on top of their Bardic skills, I'd have said no--that's not fair to the Druid player, and not warranted by the fiction, since every tradition of magic requires substantial background learning before you can practice it, even if the edges can be fuzzy and blend into other traditions.
I love this kind of stuff!
Thing is, your example with the blood spirits was both realistic and cool: the players found a way to turn the in-game reality to their favour, resulting in a cool and memorable moment.
Sure. But if realism is not of
equal priority to what is cool/fun/interesting, then there are only three options: either it is
more important than cool/fun/interesting, and thus there
must be times where cool/fun/interesting is given up in order to have more realism; or it is
less important, in which case there
must be times where the reverse happens; or there is some kind of exchange rate. Thus far, I can tell the second option is completely rejected, but I have yet to see anyone (in this thread) suggest the third. I know
@Micah Sweet has previously agreed to something at least loosely of that shape in prior threads, and I considered it a very commendable admission on his part--that is, that while realism is a
very important goal, and sacrifices to it cannot be made lightly or casually under any circumstances,
it isn't an absolute goal. There really can be times where a sufficiently minor loss of realism, plausibility, reasonableness, whatever we want to call it, is a worthwhile sacrifice for a massive gain in cool/fun/interesting consequences.
Personally, I think any cosmology which allows
any form of supernatural power has already done exactly that. Magic is just one of the most pointedly obvious examples, since it is quite literally the practice of doing certain actions which cannot plausibly have a special effect....but in a given setting, they
do have that effect. Thus realism/plausibility/reasonableness has been violated in a clear and established way
because magic is so incredibly, inexpressibly
cool.
It's why we have dragons, which emphatically should not work in a dozen different ways, showing up in damn near every fantasy setting. An open violation of the square-cube law, and aerodynamics, and evolutionary adaptation, and reproductive biology, and materials science, and thermodynamics, and probably another half-dozen things besides--but we accept it because dragons are
so freakin' COOL that it doesn't matter that they're implausible as f---, we want them anyway. The ways they're unrealistic are subtle to everyday thought, and the ways they're incredibly cool are really really obvious, so...we accept them.
I consider the properly-applied rule of cool to work the same way. If the unrealism (in the proper sense of the term,
unlike to our real Earth) of a thing is minor, or obscure, or hard for a typical person to grasp intuitively, that sets the stage for something that is
sufficiently awesome to be allowed in, even though it is technically a net negative in realism. The greater the loss of realism, the more unequivocally awesome the cool thing has to be in order to get through. An absolutely tiny drop (e.g. "the chandeliers in this place have ropes strong enough to hold the weight of a grown man in full armor" or "I can stay on the bannister of this staircase while 'riding' on a shield, even though that should be crazy unstable") may not require
that much awesomeness at all. A moderate drop in realism requires awesome so extreme it cannot be denied. A severe drop? It's unlikely
anything is sufficiently awesome. The exchange rate between realism and coolness is very much biased in realism's favor: every "cent" of realism you give up must be met with many dollars of coolness or it's just not worth it.
