In role play games that I've participated in both as GM and player I've noticed that how often and for what reasons dice rolls are called for can alter the entire gaming experience. For those unfamiliar, in the majority of pen and paper rpgs rolling dice is the method of determining the outcomes of attempted actions. Some games (Nobilis) move away from this mechanic and instead give a number of points to each player to spend as fuel for various character actions. I believe conflict resolution systems outside of dice rolling are interesting, but the focus of this post will be entirely on dice rolling. Personally, I've had a love/hate relationship with using dice in games, but now embrace them - not for the gamist mechanic they introduce - but for the unexpected narrative element they can be in a well GMed game.
Some GMs will call for a dice roll on any action that has a probability of failure. Most GMs are not this dice crazy. The problem with calling for dice rolls over even trivial tasks is when they DO fail they derail the game, and can make a player's roleplay feel unappreciated because no matter how good they might roleplay or describe their character's actions for trivial tasks, a failed dice roll counters this, and then an attitude of: Why bother RPing if the dice decide everything? can enter their minds. Another related problem is that if a dice roll to gain necessary knowledge or complete a necessary task critically fails is it runs the story into a deadend. In such cases where a failure or dramatic failure occurs a player is aware that a necessary path has been blocked from them because of a bad dice roll, and the options available to the GM in such a case are not great: They can ignore the dice roll or have the player reroll - in either case it implicitly suggests that the dice rolling is irrelevant, and in turn this causes us to ask: then why did a roll need to occur in the first place? Or the GM can try to impromptu introduce a new path, which often may feel like a deus ex machina, though is forgivable if the characters get reinvested in the story and ,preferably, if the ex machina is subtle, or if not that, made fun of for how obvious it is .
Example of critical failure on obtaining necessary knowledge: Private Detective Wright (character) has finally tracked down the hiding place of Julia Moon, his missing femme fatal informant. The door is locked and no one answers, but the GM needs the player to get inside to find Moon's dead body. He has Wright's player make a dice roll to break in... but crit fails. The GM had already indicated that there was no other way in. The crit failure can either be ignored, and rerolled, again (this can be disguised by saying the character can come back the next night), but then why have the roll it in the first place?, or the GM can add a window or extra door that turns out is unlocked after all - again, why was the roll necessary?
In the above, saying the character can come back the next night or adding an extra entrance to the house are not awful options. They can be improved by adding consequences: If the character comes back the next night, the body is further decomposed, or maybe someone else had time to come back between last night and the night before and has left some further clue. If the character breaks into a window that has been added (this, I believe is the worse option, because it simply adds new paths to ignore the seriousness of dice rolls), then the GM might have some new interesting consequence occur like a neighbor calls the police and they come to question Wright for the breaking-and-entry and murder. If you believe dice rolling is essential, especially in moments like the example described, but have a critical failure occur, then these are ways that could get out of having the game meet a dead end for no other reason than a bad dice roll.
Lately, when GMing, I've asked for less and less dice rolling for many tasks. In physical conflicts I'll even sometimes label some antagonists as "throw-aways" or "extras" meaning: easily defeatable, and present not to give chance of failure for the group, but simply there to provide a scenario where characters can engage in cinematic description of action based scenes. I first borrowed the idea from Exalted, and have found similar lines of thinking in games like Spirit of the Century.
From Evil Hat Productions (makers of Don't Rest Your Head, Spirit of the Century) I saw these intuitions and house rules explicitly stated in their games: don't ask for a dice roll every time a lock needs to be picked. Only ask for a dice roll when a lock needs to be picked and there's something horrible on the other side. If a lock is only present on a door because it logically would be, then just have the character be able to pick it. Failing this stops the story for a trivial challenge. They expand this idea further in SotC, and give what I've adopted as a golden rule of dice rolling: For anytime you ask for a dice roll picture what the success will be, picture the failure, are both interesting and move the story forward? If not, then do not ask for the roll.
A nice consequence of this rule is that I try harder to come up with stories that leave out the boring parts. Scenes of inconsequential work that add nothing significant to the story are cut out or briefly described by GM or player and the next scene of significance is immediately gotten to. This might give the impression that I run only games with every scene happening at a break-neck speed and the characters constantly surrounded by insurmountable danger, but this is not the case. Especially when using the World of Darkness setting a slowly built horror feeling is often hoped for. This is not always achieved (inadequacies on my part, or the group can't help but turn anything into a joke), but it is greatly relevant to the genre of game (horror) that we often play in.
A specific form of dice rolling is when it is necessary to gain new knowledge: through research, interviews, academic memory, et cetera. Such rolls have come to feel tedious to me because they are essentially knowledge dumping (I believe I stole this term from SotC). The GM has a bunch of information he's been holding back, a few successful rolls on the players part, and they gain this knowledge. If the roll/s fail then they don't get the knowledge, and the game halts. SotC utilizes one of the most interesting mechanics to do away with this problem: players are allowed to spend fate points declaring facts about the world. I love the idea of this because it makes the story creation far more interactive on the players part and allows them to share this more with the GM instead of simply being told the way the world is. An example: Two monster hunters arrive in a dark castle. One has been here before and is aware that vampires are likely lurking all over the place. He spends a fate point and his character declares: "Holy water? Don't make me laugh. Crosses? Not on your life. The sun? If it were only that simple... These immortal creatures can only be stopped by summoning Death herself to take them from this mortal world. It's not easy. It may mean our own lives, but we must do it."
Obviously such powers could be abused, but this is easily mitigated by a GM who doesn't allow purely self-serving, dull additions to be declared with fate points. There's much more to be said about fate points, and the unique system mechanic of "Aspects" that SotC utilizes, but those move away from dice rolling.
These express my general thoughts on the topic of dice rolling. There are more specific topics related to dice rolling (Making social rolls v. RPing social scenes) that I'll eventually get to. I'm not of the opinion that everyone must play their games exactly as I've outlined, but I hope some of what I said or borrowed from others is useful to you.