Well, I design the world mostly as we play in it, the cause of any event can easily be retconned to suit our need.
But the most important part is this : First, I ask my players, as a group, what kind of story do they want to tell with their characters and what do they whant to happen to them. One wanted to have a story where he twart the arranged marriage of his one true love so that she may be with him, another player wanted to play a noble struggling to find a way to finance his gambling addiction while not giving shame on his family, another player wanted to play a rich merchant, a member of the gnostic sect hunted down by the inquisition who lost friends and family because of it and is bitter of the world, and the last one wanted to play a spy.
Then, we decide on the plot : The kingship in the kingdom is matrilineal, Foreign princes become king of the kingdom by marrying the princess, who is the cousin and true love of character 1. However, Character 1 is rather naive and dim-witted, and the other Characters will want to have him become king of the kingdom instead of the foreign prince, hoping to control what an incompetent king character 1 would become.
Then, we agreed to have the player character cooperate until character one finally become king, and then they will antagonize each other, until most characters get killed. We agreed to let the story unfold like a classical tragedy, with the focus on the villains, their rise to power, and their downfall because of the heroes.
We played a system that the other players did not know how to play. So, I asked of the player, before I revealed how character creation works, to find about 5 things that their characters would like to gain from the coup d'état, and that would be the focus of the story from their perspective.
This is when player 2 had his character become more complex than simply a gambler. He decided he wanted his character loyal to his friend the prince, but in the same time, secretely in love with the princess : by making the princess and the prince closer to get together, the princess also became closer to him. We decided we would later on have this character torn between his lust and his loyalty, which the player actually playing it gleefully accepted.
So know, we have a plot, and we have flesh-out and complex characters. Next, we need the stats for these characters. We know what the characters will do, so it does not matter what they can or can't do. We assume they will drive the plot we agreed on. Stats, however, will dictate How they will drive the plot, and that choice is up to the players.
Then, we have Npcs. When a Npc is important, I always try to give out their names first, let the players imagine what this character is like.
Then, when a Pc and a Npc met, sometimes, I will, more referee than game master, tell the players : your character meet with Lord Falsevor. Tell, me, what does the guy look like, and what have you heard on this guy?
The player will then tell me some details and juicy gossip about the Npc, and then, I will ask :
What is it you don't know about this character? what is it that you don't know about him, but will be relevant to the plot, and how?
The trick is to be more a referee than a storyteller. The players are the storytellers. It is not up to me to initially dismiss an idea because it is "not original" or because it makes the Pc "overpowered". That is up to the other players.
This is probably why they don't like fight scenes : when a fight erupts, I retake control of the game. They lose that control and a lot of freedom. Suddenly, their stats become limits instead of being inspiration. And, more importantly, the more we play the fight by the rules, the more likely the plot we agreed on will become impossible as a result, and this is not an issue we want to happen, as a group. That basically means that we know in advance how the fight will go, and that makes it not interesting.
