The prototypical RPG mechanism is the contest and the first example given is arm wrestling.
If my character has a strength of 13 and yours has a 14, yours should win, right? Your character is objectively stronger than mine, so obviously when we arm wrestle, yours wins.
Except that’s not how it works. If there are 5000 RPGs with stats and dice, I would be shocked if The Internet can name 2 where that’s how it works. The rest all say “Whoa whoa whoa, sure the character with a higher strength has an advantage, and should win more on average, but there are so many other factors; and anything can happen; and uncertainty is more fun; and if the outcome is predetermined, why bother!”
None of these arguments are invalid, which is a big part of why this design choice hasn’t been questioned much. But what are the arguments against?
How often are these two characters going to arm wrestle? If the answer is usually once, having them each reveal their score and naming the higher the victor is a fast and intuitive resolution in the game, and serves the story well because the audience learns which of these characters was stronger.
If there are factors affecting the outcome of this match, what are they? If I’m finding an extra reserve of strength because your father beat my father in arm wrestling, the audience wants to know that. If my partner is distracting you, we want to see how. If you’re fatigued, or de-motivated, or whatever, we want to feel that. Better than “roll d20 +STR” and abstract those factors, would be “choose motivations and event cards face down and simultaneously reveal to see whose final strength is higher.”
D20 + STR, btw, is a joke. Even in the theoretical probabilistic mean, your character only has a 5% advantage over my character: In 40 million contests, yours will win 21 and mine will win 19. Maybe it’s not obvious that’s not enough of a difference: Suppose you have an 18 strength, the maximum possible, and I have an 11, the average: You’ll win 31 and I’ll win 9. I don’t know about you, but I will never ever win a wrestling match against Arnold Schwarzenegger.
And that’s assuming every point of strength matters. There are too many games with calculated secondary stats where only every 2 or 5 points of strength actually affect contests. Collapsing modifiers is a good way to differentiate characters when the randomizer is throttled (Apocalypse World’s 2d6+stat, for instance) but actively bad in a d20 system (or percentile, heaven forbid).
What about criticals? Should a 1 mean auto-fail and a max roll mean auto-success, even in the worst matchups? Yeah, sure. I agree that never having 100% certainty in a contest is better for suspense, drama, story, and fun. But you know what’s more fun than my feeble character going against your masterful character and winning because my die landed on the 20 face? My feeble character scrabbling and scheming in story-enriching ways that I as a player orchestrated.
Our cause and effect reversed. Rather than roll a die representing abstract factors to see how a thing goes, make a thing go and notice whether that was a really good (like a 20), really bad (like a 1) or something in between. Maybe you make that thing go by deciding what factors are in play, or maybe its determined by the game state (including your characters stats/attributes), or maybe you just choose.
What about multiple contests in sequence? AKA combat. It’s not hard to argue that even if your character is a slightly better sword fighter, even if your character is ultimately going to win the fight (barring external or extraordinary factors), the audience still wants to see an arc with multiple blows traded, with risks taken, costs paid, and surprises launched. And I don’t disagree.
But in the same way I think we’ve got our cause and effect reversed in individual contests, I think there are better ways to approach combat than a series of random events. Particularly since 3-6 rolls does not begin to approach the probabilistic norm we like to imagine it does. I’ve alluded to one such system that I’m sure would be fun in combat above, so here’s another:
Suppose that each character is a bundle of feelings. (They are, if your story’s any good, but suppose your game system mechanizes that.) My character is a monk: focused, optimistic, dedicated, obligated, and trained. Your character is swashbuckler: intuitive, amused, vain, unpredictable, and talented. Each round, we simultaneously choose which feeling our character is feeling and reveal them. The resolution is us as players figuring out what’s happening in this moment that’s causing our characters to feel this way now, and how that interaction plays out where we are.
If I played optimistic and you played unpredictable, maybe I see you’ve forgotten to tie you shoes and go to trip you, but you swing from a chandelier, bowling me over. Of if you play unpredictable and I play optimistic, you hop on a table for no good reason and I see it’ll be easy to defeat such a sloppy opponent, tipping you off. Hmm, looks like order matters, or else some kind of priority; If the goal is to have a back-and-forth, it might literally be: this turn figure out how our cards describe me pulling, next turn, you.
Point is: Question precedent.