Ok, here's the next question. How do you handle encounters that originate from the GMs side of the table?

For example, say my band of adventurers are camping for the night. I have a clan of rock goblins sneaking up on them. I ask the guy on guard duty to make a Descernment check...

Now, if he succeeds, he could start stating facts which could ruin the encounter I have planned? (e.g. "I see an elf wandering through the trees."). I can see handling this in one of three ways...

1) Temporarilly disallow the rule of successes when the roll is initiated by the GM (as opposed to player initiated actions). This would be similar to the way damage rolls and a few other rolls are handled in the game. Or simply require them to use the successes as bonuses instead of facts.

2) Incorporate their facts into my encounter as best possible. "The elf takes two steps into camp and falls to the ground. A half dozen goblin arrows are sticking from his back..."

3) Just go for it, and abandon the encounter.

More importantly, how does this affect actual game play. How easy/hard is it to introduce plot points from the GMs side of the table?

-Rich-

The BEST way for a GM to introduce to his own encounters is to just wait for players to give him an opening, then narrate it in with the facts. This would work like the elf example, where you just add in the encounter on top of the player-provided facts during narration.

Another way to do it: pretend the NPCs are your PCs and then make a test and narrate facts. So if goblins are sneaking up, count up their successes and, state your facts (or bonus dice) and hand it to a player to narrate. For example: Fact 1: "Three goblins jump out of the forest to attack you." Facts 2+3: held for bonus to attacks. Then the player has to narrate in these facts. "My character hears a branch snap and gets up to see three goblins rushing in from the other side of the camp. Oh no! Ralph's character is right in their way. I'll go help him!"

This is the same system that would be used if the player had failed a test, but I think it works well in this case and doesn't force you to restrict the fact system.