Lizard said:
It's determined, somewhat in advance, by how the DM statted him up. That's where storytelling comes in -- when the DM creates the obstacles and decides how challening they are.
I think we can agree that it's an issue of how the DM stats up the challenges, how the players decide to tackle those challenges, and how the dice roll -- but another important element is how the game is designed.
And a big, if invisible, element of how the game is designed is how ablative hit points work. If a guard is at all vulnerable to a one-shot kill, then he's almost guaranteed to die to two shots. For the initial attack to have a 50-50 chance of killing him, it must average as much damage as he has hit points.
This is not true with a save-or-die effect, like the assassin's death attack. If a death attack fails to kill the guard, it's not because it did "just" 90 percent of his hit points in damage. He's not reduced to a tiny fraction of his hit points.
Lizard said:
If he's equal to the rogue, then the DM made a decision that one guard was tough enough to challenge the whole party, and letting the rogue one-shot him, except by extraordinary luck, would not lead to the story the DM was setting up.
If a 5th-level Rogue meets a 5th-level Fighter in open combat, the Rogue is supposed to lose -- not 100 percent of the time, but much more than 50 percent. It's not a fair fight.
If a 5th-level Rogue successfully ambushes a 5th-level Fighter, the Rogue is supposed to win -- not 100 percent of the time, but much more than 50 percent. It's not a fair fight.
How exactly we want to implement that is up for debate, but I'd like to see a system where the Rogue has a decent chance of taking out the Fighter with one good sneak attack, and if that fails, the Fighter instead has the advantage, and they both know it.