A couple of people asked in comments yesterday what I mean by unfairness. I started to reply, but it ran long enough to turn into today's post. Now, while it would be easy to turn to fiction for examples, the simple truth is that most any good example of unfair from fiction would be a spoiler, so I must tread carefully.
At its heart, unfairness hinges on expectations, in this case player expectations. They did X, so they deserve Y. They Killed the dragon so they deserve the reward. They broke into the vault, so they deserve the treasure. When they don't get it, it's easy to be pissed or to feel the GM pulled a bait and switch. The players had a reasonable expectation of outcome (both mechanically and within the fiction) and the GM is explicitly defying that expectation, usually through simple expedience of the GM narrating the world (which someone will insist on calling fiat because, hey, what's a good discussion without fighting words?)
A very bad example of unfairness would be the player's rescuing the king from assassins, but he then dies falling down the stairs. That's kind of random and capricious, and it makes a useful example because of the reasons it doesn't work. It definitely violates the player's expectations of outcome (they saved the king, he should damn well stay saved), so why is it a bad example? It's because the reversal is too neutral. It's bad, sure, but it's not BAD. In contrast, consider the example of the heroes saving the king only to have him believe that THEY were the assassins, and call for their heads. That's unfair, but it's the right kind of unfair.
Good unfairness can be found throughout darker fiction (Martin, Morgan and Abercrombie spring to mind). Heroes are reviled and villains exalted. No good deed goes unpunished. You know the drill.
Dramatically speaking, it's all about the emotional charge. Mckee and Snyder both talk about this, but I'll sum up: in fiction, a good scene starts at one emotional state (positive or negative, + or -) and changes state over the course of the scene (or beat, depending). Sometimes those go to double positive (++) or double negative (--) for great victory or terribly defeat, but the general idea is pretty easy to grasp. In almost every interestingly unfair situation, the players are expecting a big payout (++) and the GM instead hands them a ticking time bomb (--). That's a huge emotional jump, going from the high of the expectation to the abrupt low. This is why the king falling down the stairs is kind of lame. It's bad (maybe - at worse) but it's got no real punch for the players. The emotional level doesn't make as big a jump, so it's just kind of annoying.
But here's the rub - the power of the event is all about that unexpected reversal. The bigger the gap, the more powerful the moment, and that's what demands unfairness. Specifically, it must be unexpected, and under any kind of measure of fairness, that big a jump would simply not be possible. It requires disempowerment, opacity and surprise, all of which are INSANELY abusable things. But they do the job.
Now, I want to note that unfairness is not necessary for the _events_ to occur. A fair table populated by players with a good sense of drama are fully capable of inviting outcomes on themselves every bit as brutal as the dramatically unfair GM is going to do, perhaps even moreso. But I am saying that unfairness (or more aptly, the surprise and dramatic shift which only unfairness can allow) is the only way to deliver the real gut punch.
Obviously, this is only one sort of payout. I don't expect every table to prioritize it the way that I do, nor would I want them to. Games offer a huge array of emotional rewards, and it's well worth going towards those you value most. But it's an important one for me, and I consider it a tricky one to do well, so I figure it was worth some air time.