Specifically, uncertainty in potential results. Swinginess. Random happenings because the dice get a mind of their own. That sort of thing.
When used in moderation, this is very useful, yes. A significant chunk of games which feature it make it much
too swingy. Particularly because gaming in general has moved away from having a default of the ultra-hardcore "one crit and you're just dead" (and similar "meatgrinder" lethality) gameplay.
I have played and like some "story" games, but one thing many of them lack is uncertainty. Their mechanics tend to favor participants being able to say things that become true in the fiction (even if they don't call it that).
Not sure which games you're referring to. PbtA games certainly still feature uncertainty. Even when you're working with your best bonuses (+3 bonus)
and getting some other kind of benefit (+1 forward or ongoing), snake eyes (2.78%) will still give you a legit failure and rolling 3-5 (a further 25% of cases) will get you only partial success, where you must accept a lesser, incomplete, partial, or trade-off result rather than full success. And since there are three possible results (failure, partial success, full success), the uncertainty is actually much more significant because (to make an example of arguably the most common roll, Defy Danger) the exact "worse outcome, hard bargain, or ugly choice" will always be context-specific.
It's even baked into one of the Agendas and several of the Principles, which are the things that defined DW's design and rules. The Agenda in question is, "Play to find out what happens." By definition, all of the rules are meant to serve that purpose, and they do so extremely well if you permit them to. It requires
not over-preparing,
not taking the reins,
not trying to "plot" things, but instead letting the rules do their job and focusing on responding to what the players choose to do. It's tough, and I don't always do it correctly, but that's on me, not on the rules.
As for the Principles, you have things like:
"Draw maps, leave blanks" (read: you should do
some prep work, laying out frameworks and spaces and events, but
do not prepare down to the smallest details,
do not completely fill up the world,
do not prepare consequences)
"Make a move that follows" (your GM moves should be in response to what the players do, not adhering to some personal plan; they should also always make sense within the context of the fiction, though the players may easily not always know why)
"Ask questions and use the answers" (take player input and feed that back into the process: player answers to GM questions are a
huge source of engaging development that doesn't require dice in order to create unexpected things)
"Think offscreen too" (make use of established past events and yet-to-be-seen dangers, consequences that won't come to light until later, and other forms of belated surprise)
I prefer when participants in D&D (and similar "trad" games) say what they would like to be the case, and then the dice decide how that turns out. That goes for the GM, too, btw -- the GM being subject to the same uncertainty is equally important in creating a truly surprising and novel experience.
Alright. But I assume you also would prefer to avoid things that frequently produce deeply unsatisfying results, yes? That's the problem of totally unconstrained randomness. It's usually unsatisfying. That's why mechanics to
manage--rather than
eliminate--randomness are significant. For example, and
please for the love of God don't make assumptions about what this means for my game, I usually frame situations in such a way that my Dungeon World group is unlikely to face (note all of these words, each of them is
extremely important)
random, permanent, irrevocable character death.* That is, by definition, reducing some amount of the "swing" of the dice, because I'm taking a particular option off the table.
*Random: An unforeseen result of probability; if the player wishes their character to die, then have at it. Permanent: The character will remain dead and unplayable unless and until something is done to revive them. Irrevocable: Nothing the other players can do will return the character to life (though other forces, outside their control, still might.)
Only death that is
all three of random, permanent,
and irrevocable is off the table. Random and permanent but not irrevocable? Awesome, that opens up a whole "save our friend" story. Random and irrevocable but not permanent? Ooh, how will this person come back to life if the other players aren't involved? I sure as heck don't know!
This isn't to say that no participants should have certain choices. I think players should get to design their characters without having to deal with dice, and GMs should be able to build the initial conditions of play (the "situations") with as much or as little random information as they desire. But once play starts, I say roll those bones in the open and stick by what they say, whether it's a random encounter with an ancient wyrm (don't forget to roll reaction!) or the BBEG gets one shotted by the torch bearer.
Okay. What about ways to manage things, to
occasionally take things out of the hands of the dice, but not
always do so? Because that's always been one of the issues with D&D--it puts a ton of ability to evade or invalidate the dice only in the hands of
some character archetypes, while leaving others utterly at the mercy of dice no matter what, frequently leading to really silly things. E.g., one of the greatest faults of "critical fumble" rules is the sheer idiotic rate at which swordsmen fall on their own swords while fighting, a thing which is
only a problem for characters that make attack rolls--spellcasters, who may never need to make an attack roll at all, are thus insulated from this "swinginess," which creates a pretty clear form of rules-derived unfairness.