Some random advice on how to run a compelling game, from someone with nineteen years of experience. And many interesting delusions.
This is very, very easy to forget in Amber, where the PCs are usually junior members of a family full of people who can destroy entire universes before breakfast, wade through an army and kill it all before lunch, and draw a napkin doodle that exceeds the works of Leonardo Da Vinci in quality at dinner. The Elders can slap the PCs around like red-headed stepchildren, can probably outperform any given PC in most areas, and if allowed to rampage unchecked, could probably solve all the campaign's problems while the PCs catch up on their taped TV shows.
But you can't let them do that, because the Players didn't come to the game to watch your NPCs save the day. They're supposed to be the heroes, not supporting cast in their own game. This isn't to say that all NPCs must lurk mutely in the back, carrying spears and torches and fetching coffee (see rule #4), but it does mean you need to design the plot in such a way as to give some reasons why the PCs are in charge of (leading the expedition, investigating the mystery, bungee-jumping into the Abyss, etc) and why they are important to Amber's survival. Certainly, an expedition under the command of one of the Elders can be a quite fun scenario, as long as the GM makes sure the PCs are relevant to the success of what is being done, not just along for the ride.
I myself use several mechanisms to ensure the PCs get to be the heroes of my Amber campaigns.
With a sufficiently large and complex plot, one can pin the Elders down solving various problems off stage, while the PCs deal with a set of problems scaled to their capacities. Makoto may be off infiltrating House Formorian to plant bombs in their airfleet, while the PCs are trying to break into House Hendrake to steal their WAR TRAIN!
In my own games, the Elders usually have duties around Amber itself which pin down many of them, preventing them just gallivanting around Shadow solving all the campaign's problems. After all, idle hands are the Devil's workshop, and no King of Amber wants his relatives hanging out there...
Thus, the PCs are free to act as free agents and can run about in Shadow doing things and being the stars (while one can still call in the Elder Cavalry if it's really necessary)
In some cases, the Elders may simply not want to be publically seen to be entangled in something, and ask the PCs to do it in order to solve the problem without directly connecting themselves to the methods used to solve it. This is useful because it shows the Elders as involved in the plot, but leaves the PCs free to actually do the decisionmaking on the spot themselves.
Ancient prophecies make a useful plot device. If they say that Space Elvis is the one destined to recover the Pattern Maguffin, it provides a good reason why Bleys hasn't already nicked it and added it to his toy collection. This is, like any DM trick, vulnerable to over-use, as one can also straitjacket your players and plot. If, for example, your Chosen One trips on the Pattern before the Prophecy is fulfilled, how does one get around this?
In some cases, you can set up things so they pique player interest, but wouldn't logically have sent the Elders on the hunt. Or the Elders may send the players off on some mission that seems trivial, but turns out to have major consequences.
The Elders may give the PCs things to do simply because they want to provide chances for the PCs to grow in their skills and prepare them for the duties they will one day have. Some of these might be tests set up deliberately by said Elder(s), others may be problems that need solving, but which would eat up more time than they're interested in spending, but could provide a good experience for said PC(s).
This may turn into Seems Innocent Enough if the Elder doesn't know what's actually going on...
Some villains will simply be 'the guy I killed on day 35 when he tried to mug me'. But the major antagonists need to be more interesting than that. You have to give them flair and character and you have to get them to do things which will make the players care about them. Different major villains should approach problems in different ways, so every major confrontation isn't the same confrontation.
There's two aspects to this. First, find out what their characters want, and what they hate. This will let you build villains who have reason to come into conflict with them. Villains shouldn't be clashing with the PCs just because of the 'PC' stamp on the players' foreheads.
Secondly, design your villains to match the players' capacities. Make sure the villians CAN eventually be beaten. It might require a player teamup, or finding some special item, or simply for the PCs to advance their stats to a point of superiority. Make sure the PCs face foes that don't require things they don't have and likely can never get (in time anyway) to beat, or at least that they have enough warning to avoid said foes if they can't beat them.
Of course, PC action may place PCs in unwinnable scenarios, and sometimes they may motivate villains they can't beat to come after them. But if most of your PCs are Psyche people, don't throw endless waves of mind-shielded warfare gods at them, and if they're Warfare people, give them some enemies they can fight with their weapons.
This is something very difficult for GMs, as we sometimes become attached to our beloved villains. It's very tempting to cheat to enable them to always get away. While they shouldn't just fall down when the PCs arrive, don't drag things out by making it eternally impossible for the PCs to beat their foes once and for all. If you drag things out too much, they'll just get frustrated and bored.
This isn't to say your villians shouldn't fight for life with every trick they can muster (see Rule #2), but that if your PCs out-think you (and they will, sometimes), don't deny them their victory because the plot demands it (See Rule #5). Of course, if the PCs kill the Master Villian in Scene #2, there goes the campaign, but hey, they earned it (As long as it was due to deliberate PC action and not just dumb luck. No campaign climax should bring either side victory just from dumb luck, especially in a DRPG).
If the villians escape because the PCs screwed up, that's another story.
One temptation for DMs is to let the NPCs walk all over the PCs, constantly outdoing them and showing them up. But an equal temptation is to reduce them all to helpless incompetents who can't wipe their own bottoms without PC help. NPCs who are nothing but a burden will eventually drive your PCs up the wall (and/or become nothing but more marks for them to shill, in the case of some players).
While the PCs should, in general make the big decisions when in joint operations weith NPCs (They are the stars, after all), the NPCs shouldn't always defer to the PCs. This varies a lot, of course, by NPC personality. They should certainly offer advice. Of course, you have to make sure they don't just become clue machines; sometimes they may give bad advice.
A very important thing here also is reciprocity; PCs should be able to get help from the NPCs when they need it; that way when the NPCs are in trouble, they'll actually care enough to go to their aid. (Although some NPCs may, of course, be sufficiently obnoxious to end up out of luck regardless of how helpful they've been or are able to be)
A good campaign has a plot. But the Plot should never be more important than the players. This doesn't mean the PCs should necessarily have to be present for every event, or that they always have to have a chance to stop something bad which happens when they're not present. But some things it DOES mean:
Never write in your notes that the PCs cannot prevent some event X if they're present for it. As per Rule #3 and #6, if the PCs earn their victory, give them their victory. The antagonists should fight back with everything they have, but the players' actions should be able to derail the plot; they are the stars of the show (rule #1) after all.
If the PCs do fail, make them feel the consequences of failure. (Rule #7). But never set things up so they must fail for the story to go on.
There are times when PCs may find themselves forced into courses of action by circumstances or by NPC action or by their own bad choices. But you should never decide that the players must do X to make the plot work, and then set out to make them do X. The PCs should always have the choice to let Amber be burned to the ground while they go to the Shadow of Willing Bi-Sexual Women or whatever. Of course, such actions have consequences they may not like (see Rule #7), but you shouldn't straitjacket them into a plot where they must run along linearly with no choices of other options.
The one big exception to this is that sometimes the players may, by their own choices, trap thesselves in a linear situation where their choices are minimal to none. That is another story entirely; if PCs bring something on themselves, it's their own fault.
PCs don't surrender. It's just a fact of life. Even if 200000000 Angry Logrus Masters surround them, they'll either fight to the death (or unconsciousness), or maybe try to escape. So never design a nifty escape scenario and then try to catch them so you can use it. Because no matter how cool a story it might make, you probably won't be able to pull it off in a way they'll enjoy.
The one exception to this is if you start the campaign by having the PCs escape something together, which can make a good bonding experience.
Unless you're playing Paranoia, every problem the GM creates should have at least one solution within the grasp of the players, and ideally, you ought to think of at least three ways out before you spring it on them. They may not all be successful solutions (running away may be the best solution at times), but there ought to be several ways for the PCs to at least avoid abject disaster.
This applies mainly to GM-created situations. As per Rule #5B, Subclause A, if the Players deliberately put themselves into a no-win scenario primarily by their own actions, then it's their own fault. Fortunately, players rarely do this. Especially if you apply Rule #6B (see below)
An important sub-aspect of this is to make sure that if you set up difficult solutions, that there's some reasonable way for the players to get the information they need to solve the problem. Putting the solution to killing the Giant Black Dragon Eating Shadow inside a whiskey bottle at the bottom of an ocean out in shadow when no player can shadow travel is a very good example of how NOT to do this.
It may require a lot of thinking and effort, but you need to make sure that you don't create problems the players can't reasonably solve with some effort.
Sometimes, there simply is no way out of a disaster. If a PC decides he's going to try to ambush and kill Benedict by himself, there's no reason to feel any guilt when you have Benedict carve him into giblets. However, a good DM always provides some warnings before the PCs do something foolish that will put them in a totally inescapable trap. Unfortunately, sometimes you can't realize fast enough that someone is going to do something stupid before they do it.
All actions must have consequences. This has both a positive and a negative aspect.
From the positive aspect, the PCs should be able to make a difference in the game. Their actions should be able to turn the tide of whatever struggle is in progress, because if they are unable to impact the plot, they'll slowly go mad.
From the negative aspect, when the PCs screw up, it also has to have consequences. Don't fear to smack down those who do things, especially when they go against warnings.
One key of good GMing is knowing when your players have in fact come up with a cooler plot than you have. They may be deluding themselves as to what's really going on, but hey, sometimes it is profitable to change what's really going on, if it lets you steal their idea.