Normally, players will encounter a villain at about the time they’re ready to fight him, so he’ll be at about their level range, probably a little bit higher just to make the fight interesting. That works a lot of the time, but sometimes you might want to introduce a villain earlier in the game, to plant a seed early on that both invests the characters in their relationship with him, and expresses just how scary this guy is. One easy way to do that is to have the players fight him well before they are ready for it.
This can seem unfair, because it is. Assuming the level disparity is high enough, then the players have no real chance of winning, and as the GM you need to go into the fight planning for the outcomes that will likely come of this. The three main possibilities are TPK (total party kill), flight, and heroic sacrifice.
If your players are the kind to take it down to the wire, or you introduce the villain in a particularly bad situation, then he may take down the entire party. A TPK is something a lot of GMs will flinch away from, and often with good reason, but the trick is to be prepared for it in such a way that it moves the game forward, not drags it to a halt. Remember that in most non-horror systems, death is rarely death. More often it means "knocked out and at your enemy's mercy" and while that may mean death in some situations, it can also mean something more interesting.
Captivity is the most straightforward way to handle the outcome. The characters wake up imprisoned somewhere and must free themselves before moving on to things like revenge or thwarting. Think about how captivity can really shift the focus of the campaign. The characters may end up far from home, isolated from allies and resources, and basically finding all of their assumptions about the game so far turned on their ear. That may sound drastic, but if the reveal of the villain was sufficiently dramatic, then the change in situation should be equally dramatic. The good news is that the choice of captivity usually reveals something new about the villain, such as who his allies are.
Disdain is also popular. Leaving the characters beaten but alive is trickier. It’s a classic trope of video games to have the villain leave heroes alive as a sign of contempt or to send a message, but players may take it as a sign of stupidity. Unless you feel up for the acting challenge of really selling contempt, don’t risk it. Similarly, if the villain has a more complex message than “I am here!” then the message thing might work.
What works best of all is if you have a nuanced villain who has a reason for leaving them alive. Maybe it’s mysterious (to be revealed later), or maybe he’s trying to persuade the characters to join him. Perhaps he considers himself a good guy forced to do bad things, and he won’t unnecessarily kill. If leaving the characters beaten but alive reveals something about the villain that makes him more interesting then it’s a valid option.
The last option is to have a last minute rescue. This is the most dangerous and problematic option because if it’s done wrong it really profoundly sucks. Having a heroic NPC show up to drive the villain away is incredibly lame, and is a flag to the players that the GM is more interested in showcasing his cool NPCs than he is in helping the players. So don’t do it, ok?
One twist on this chestnut is a little more tolerable, and that is to have the heroic NPC die in the process. The idea is that the player’s love of the heroic NPC will turn towards bitter hatred towards the villain and guilt over his death that motivates them to action. It can work, but only very, very rarely. The problem is that this is incredibly overdone, to the point of cliche, and players first thought on meeting any old warrior-sage is to wonder what sort of heroic sacrifice is going to kill him. Worse, when the GM chooses the character to sacrifice, he often chooses one that is well loved by him, not necessarily by the players, and that just means watching GM Kabuki.
There are other ways to pull off a rescue that don’t leave your players looking like crap. The first is a distraction: a smoke bomb, an illusion or some other trick that draws the villains attention long enough for everyone to flee. This doesn’t work so well if most of the party is already at 0 health, so this needs to be well timed. This is straightforward, and it can be a good way to introduce a new and (seemingly) trustworthy NPC, but it is a little bit overdone. If you’re looking for a hook for an NPC, then consider it, but don’t invent one just to rescue the characters – it will feel flat.
One particularly dirty trick is to have the rescuer be another villain, especially one that the players really dislike and who would enjoy having the characters owe him. As as bonus, consider how suspicious they look getting help from him. Neither villain is likely to be committed enough to risk their life fighting one another, so the characters are left with two fresh problems, which seems ideal to me.
If your players are smart enough to flee in the face of an overwhelming enemy, then more power to them. That kind of thinking is less common than you might expect.
Fleeing raises some of the same questions and issues that a TPK does, only in this case the main question is “Why is the villain not pursuing to finish them off?” These might end up being the exact same reasons he wouldn’t kill them (contempt, rescue and such) but it might be something as practical as, say, a huge iron door slowing him down enough to buy time.
Flight calls for it's own sort of tension and tempo - when characters flee it is usually because the players feel fairly certain that the outcome of the fight would not go their way. This is a great basis for a chase scene, and if it seem slike a good opportunity for one, then totally go for that. There's lots of good advice on how to do so, and it's worth using. However, that is not always going to be the case.
There's a bit in one of Stephen King's Dark Tower books which is basically th emost boring chase ever. The heroes are in a dark tunnel, there's something huge and bad after them, but it's held off by the light, so they steadily go through their supply of little sterno lanterns, lighting them and running, one at a time. On paper it seems like it should be very tense: huge menace, dwindling resource, etc.. In reality it's incredibly flat because it pushes beyond suspension of disbelief - there is no other way this scene could possibly play out than the way it does (which is, of course, they get away in the nick of time).
A lot of chase scenes end up feeling that way in games where the GM is expecting only a chase scene. Just as it's a bad idea to call for the players to roll dice if you can't manage failure, it's a bad idea to start a chase if you have no strategy for it going south. Now, I'm not arguing against a little sleight of hand to tilt things if needed - that's a whole other thing - but rather if you only have one possible outcome, you're running a cutscene, and those suck. Seriously.
All of which comes back to the point: If there is no natural way for you to manage the chase, then let the escape succeed and move onto the next thing. Simple as that.
Realistically . though, I doubt many of you are kind enough to let the players off that easily, so if you feel you must have a chase, then consider minions. The big bad does not need to pursue the players himself, he can just send some mooks. What makes that interesting, even if the players aren't afraid of the mooks, is that fighting the mooks takes time, and the big bad might follow along behind them. That possibility is goign to be far more a fruitful source of tension than any actual fight.
There is a good chance that one of your players is dreaming of this opportunity. They want to be the one who shouts "Go, NOW!" and takes up a stance to face the big bad himself with the expectation that he's going to die in this fight, but he's going to buy his friends time.
A lot of GMs worry about this scenario. This splits the party, it unduly focuses on one character, and it leads to a PC death - all of these are by the book bad things. A lot of GMs are willing to cheat a little bit to keep this from happening. They'll let other players knock out the first one, they'll make the big bad just do something entirely unfair and knock the guy aside and take him out of the fight rather than kill him, or just use soem movement power to go around him.
Don't do that.
Seriously. This is the thing the player wants and the best thing you can do is push them and push them HARD. The only thing worse than this sacrifice being meaningless is if it's too easy. You need to trust the system and your own design of the bad guy and make it a serious throw down. Keep track of every round that he survives and give the rest of the party a token for it, and let those tokens be their budget for the subsequent escape.
Once the player loses, you can apply some of the same thinking to his fate that you do in the TPK, but don't go out on a limb to keep the character alive. If it's _really_ appropriate that he not be killed, then fine, roll with that, but be willing to end it. The player knew what he was in for when he started down this road, and by giving him closure, you're giving him a gift. This character's death will become one of those stories he tells when he talks about the best game ever.
1 - "Levels" are used as a convenient shorthand here, but the basic model applies just as easily to point based systems or any other system where such power discrepancies exist.
2 - If you really feel you want to try this approach, then do three things.
First, make sure the players are invested, not just that they’re supposed to care. Listen to them talk, pay attention to who they enjoy scenes with, note which NPCs they seek out when they have questions. Those are the NPCs you want to pull from, even if they’re not the ones you have lovingly described and statted up. Specifically, it should not be any kind of old soldier, any kind of young, novice soldier (or prince), or anyone at all who is actually a dragon in disguise.
Second, give them an exit strategy. The rescuer may be willing to risk his life, but making him actively suicidal is a little too on the nose. If he dies in a suicide attack, that’s uncomfortable, but if he gets killed because his plan almost works, then that can be really powerful.
Last, give it consequences. The rescuers purpose is not to die, it is to change the situation, usually for the worse. One good trick for this is to make the NPCs means of fighting something that can be taken away. If someone busted out the artifact sword to face down the villain, then make sure that when the dust settles, the villain now has the sword. One way or another, make sure that something of this sacrifice remains in the game so it doesn’t feel like a cheap throwaway.
3 - All this assumes this is being done heroically, not habitually. If a player is demonstrably being cavalier with the life of his characters rather than making a sacrifice, then feel free to one-shot him and move on, or otherwise bypass his personal drama.