Wonderful article. Let me throw in my two cents.
First off: I really appreciate the way that you describe the role of the GM. The analogy I use is Monopoly, or volleyball, or almost any other game that humans play.
In any other game, unless we're playing with children far too young to understand the idea of deferring fun for the sake of retaining a fun game or with people with disabilities where the goal of the "game" is non-conventional, we don't think it's not "fun" that we lost the game of football, or that we landed on Broadway. We don't say, "Well, the basketball game was 98-2, but everyone wins anyways" or, "No, it wouldn't be fun if you landed on Broadway, you land on Go instead".
The challenge comes from having a chance to lose, and surpassing that chance. It comes from having to make hard decisions and take risks. Research indicates that we as human beings are hard-wired to find a long shot exciting. When we avoid landing on Broadway three consecutive times around the board, we build up suspense and excitement, just like playing poker or roulette.
Now, I'm not saying every game has to be competitive. Most role-playing games are mechanically cooperative, even if competition still emerges (e.g. "Who hurt the dragon the most?") In fact, I tend to find cooperative games to be fun and interesting. But the point is that fudging the dice rolls, or playing differently because someone's having a bad day, or whatever else, screws the game. You're not playing volleyball or Monopoly or Settlers of Catan anymore. It's not fair, it's not exciting.
This comes up in the context of different skill levels in roleplaying games too. Sometimes, newer players will feel like they're having less fun because they don't know how to design their character appropriately, because they contribute less. But the game is still fair, from a Rawlsian-perspective: We all had the same shot to make the same decisions. (I'm obviously abstracting out some things, like game balance issues where only a few builds become viable, or similar game balance issues where one person gets to be a limelight hogging class and everyone else has to play support so only one person really CAN take that slot, or when people are shoehorned into roles, or whatever). Tiger Woods may make me look like a chump in golf, but it's not unfair that we played. Now, golf does have a handicap, and one could imagine similar handicaps among players with different skill levels, but nonetheless, nothing is unfair about me vs. Tiger. If I was as good as him, nothing would impede me from winning.
However, I will note that, while the GM isn't responsible for GUARANTEEING that everyone has to have fun, they are responsible for making it THE MOST LIKELY for people to have fun. So if Rob has a bad day, the GM SHOULD throw in a little more comedy, or even run a different campaign if everyone else is game. Taking into account what your players feel like is essential. But a) you have to be having a good time too and b) you are only responsible for your best effort.
Also, I will note that the argument of dice leading to agency is actually fairly absurd. Free Will and Omniscience may be opposed, but so are Free Will and Chance. If my decisions are 100% random, there IS no agency. Early on in D&D, there is almost NOTHING one can do besides hope one rolls well. That's one reason why so many people choose Wizards, Druids and Clerics, balance be damned: Many of their mechanics (Magic Missile hitting automatically, for example) get rid of chance entirely and make it about TACTICS.
Uncertainty and risk are good things, but even if one chooses to represent them with dice (which isn't always necessary), AGENCY means having different LEVELS of risk and reward.
Power Attack, for example, turns what is a pure luck activity (roll a D20, hope it exceeds AC, roll damage) into something with actual agency and choice. A player could choose to go -1, -2, -3, -4, -5, etc. to her attack roll to get +1, +2, +3, +4, +5, etc. damage. Now the player has to determine what the opponent's likely AC is, then figure out what probability they are willing to accept of missing for what probability of dealing sufficient damage they wish to embrace.
Simply put: A game where one's tactical decisions are moot and the higher dice roll always wins is just as opposed to agency as a perfectly railroaded game is.
And one doesn't need dice, or actions with ANY chance of failure, to give uncertainty and risk a go.
Let's say I'm a wizard in D&D. I can decide to cast Magic Missile for guaranteed damage. Or I can go for Fireball or Lightning Bolt or summoning a monster, at greater risk that they don't hit but (hopefully) more payoff. The low-to-no risk option, where even rolling a 1 wouldn't prevent some degree of success, acts like betting on evens in roulette: A very safe bet with low reward.
Or let's say that we're playing a superheroes game. I have flight and forcefield projection, my opponent has flame powers and invisibility. He turns invisible. I reason that he can't create flame without revealing his location, but he creates a backdraft so I don't know where it's coming from. I use my forcefield powers to survive the backdraft and hide where he doesn't think I'll be, then ambush him when he shows up. Etc. etc.
Your example of the player not being able to hit the dragon, for example, happens even in a game with dice. "Even if you roll a 20, you cannot hit him". Unless you have a game where any critical is a chance for damage and therefore death, and NO mechanic can stop this (likely some very grim and gritty war game), this MUST happen to prevent low level characters from beating high level ones. If a level one character has a chance, however slim, of killing an epic level character, what's the point of being epic level, or even level two?
But if the player had to think, "All right, it's a dragon. Well, its weight might be too big to lift with my telekinesis, so I better try a grenade", or what not, choosing among strategic options with limited informations, then dice are actually moot.
Further, dice can actually make these various options moot. I play a game where, in general, even the stupidest, most suicidal action can succeed on a critical, and the most well-thought out actions can fail ENTIRELY (not just do less damage) on a critical failure. After several years of this, the mentality this has created is simple. People think, "Why bother trying a strategic set of actions to guarantee success? I roll a 1 and it doesn't work. Why bother avoiding stupid things and thinking carefully? Just need to roll a 20 and it'll succeed". It has made the game farcical in some respects.
I love using dice for things that I really do want to be random, but dice need to be heavily contained for games to be predictable and to have real agency, e.g. choices.
But what Alex is talking about in this article is the four functions of a GM. And of those, I'd agree that storytelling is last because as a GM, your players can (and will!) help you with story. The first three - judging, world builder, adversary are uniquely within the court of the GM and therefore logically rank higher as a function for a GM.
Thank you! I could not have said it better myself.
Or, even more simply:
It's not that story is unimportant, it's that story is SO important that everyone should be involved.
When the GM railroads, s/he isn't serving the end of the story, s/he's serving the end of THEIR story. The players are ancillary. It's storytime. That's fine, but it's not roleplaying. Roleplaying is virtually defined by the ability to make SOME decisions about what your character does. And those "decisions" are meaningless if they don't change responses. It's like talking to the TV screen: You can say whatever you want, but no one would confuse that for freedom or interactivity.
Also, the discussion about Tolkien is so out of left field for me. The idea that failure being possible means the game is grim and gritty is stupid. Yes, you can create a game where there's no such thing as a "total party kill". A TOON game, for example. A game with automatic resurrection by the gods, or cloning, or Fate Points, or a reversal of the Hero Point system (e.g. a player can buy out of a wipe but then they must choose an element of the next encounter that will get harder or something), or whatever else works for you. You can also make a game where everyone dying and having to make all new characters can HAPPEN, but is remotely unlikely.
None of that has ANYTHING to do with agency. Whatever difficulty level is out there, unless it is so hard as to be impossible to meaningfully complete no matter one's decision, agency triggers when players able to make meaningful choices about what they do in response to the difficulty. The likelihood of a total party wipe is an element of game difficulty, not agency.
If players run straight at machineguns with no body armor and hope to luck out and dodge the bullets in a gritty war game, they should die. Stupid decisions should have consequences. Otherwise, why would a player ever STOP doing stupid things?
And it's important to clarify that players CAN back themselves into a corner where they are "railroaded" to death... by their own actions. Let's say you play a game where people are set in the modern world, for example. One of your players shoots a soldier in the head on a military base. That character WILL die, or be captured and tried for murder. This is not "railroading". The GM didn't make the character do that. They just enforced consequences.
