It's important to remember relatively few combats are actually with the goal of killing people - the killing or violence is usually a means or a thing in order to get another goal. The three easiest are: Get something (theft, robbery, occupy a location), Defend territory (scare folks off), Ego (thrash them and teach them a lesson, humiliate them, show that you're in charge).

Dying is usually the least interesting result, because it means the problems (for that character, at least) are over - the best results are are "How will losing this combat make your life more difficult in the future?"

I've written my Grande lista de estacas de combate which covers a lot of this.

Simple Answers

In recent years, many games have used this simple and consistent approach to fixing the problems you describe.

• If you don't want characters to die, take death off the table mechanically! I mean, that's just basic logic, really: outcomes you conhecer you categorically don't want aren't worth including in the set of possible consequences.

• Emphasize goals, not battles of pure attrition. Characters are fighting their enemies for a reason, right? Focus combat scenes around those reasons. When the opposition knows their cause is lost, have them melt away and abandon the field. When the PCs' aims have been frustrated, let them know and transition to a different kind of scene. Bankuei's Grande lista de estacas de combate is about as good a place to start as any.

  As a bonus, shifting to goal-oriented combat allows for a game structure where the battles are more than just a foregone conclusion. The PCs can lose muito without ruining the whole game. This means you can actually present fights that are more tactically challenging for the players (if they enjoy that kind of thing).

• The best way to avoid "Oh, whatever, the GM will save us!" is to trust each other, communicate clearly, and follow a consistent set of GM principles. Don't worry about the players knowing what the principles are. That's good! It means they can play more assertively since they know they can trust you to "back them up" a bit.

  I mean, if you can't trust the other players to portray their characters with good faith, honesty, and fidelity to the dramatic situation, why are you playing a character-driven game with them in the first place?

Fixing D&D/Pathfinder (Partially)

Unfortunately, the simple answers aren't necessarily the fácil answers, in this case. You're working with a huge, mechanically complicated game that doesn't support any of this stuff out of the box.

A particularly thorough fix is simply more trouble than its worth. If the "simple answers" truly speak to you, look into a system designed with these ideas from the ground up.

Here are some "quick fixes" you can use with D&D3.x/Pathfinder, though:

• Switch to goal-based combat. This'll create much more fertile ground for character-driven play in combat (which, let's be honest, takes up a lot of a D&D/PF game's session time). This'll also reduce the likelihood of character death by shifting the focus away from battles of attrition.

  In my experience with D&D, a cool giant "set piece" battle is much more fun than the four CR-appropriate encounters the book suggests. Goal-based combat lets you actually play that "set piece" as an interesting tactical and dramatic situation, with even more potential to interact with the NPCs and scenery in new and novel ways.

  If a fight has no larger purpose, it's filler. Learn to recognize filler and cut it.

• Tweak the death rules to match your goals. As written, there's a very narrow window between "down" and "dying." The bleeding rules generally encourage other characters to get themselves into trouble for fairly little gain, too. Some good options:

  • Create a more expansive "knocked out" state, so that it takes a lot of damage to truly matar a PC. For example, anything between 0 hp and -(20+level) hp is "knocked out" — you're out of the battle but won't die unless more bad stuff happens to you. To keep some of the sting in being knocked out, you can make it harder to revive a downed character, or impose a condição (try fatigued/exhausted or temporary energy drain).

  • As above, but don't use negative hit point bands at all. Instead, just have states: anything that could kill you just renders you "knocked out," and then there's a rather involved mechanic for actually morte. If you fail your save vs. Disintegrate, you're in that knocked out state, not actually dead yet.

  • Replace death with other kinds of lasting bad stuff. When the rules normally say you'd die, instead you get some kind of debilitating injury. You can get pretty creative with these, and also make them thematically appropriate to the kinds of crazy magical adventures the PCs are having.

  • Keep the death rules but give characters a limited supply of "fate-dodging" points. Lots of games have a mechanic you could borrow.

  Esta resposta has some additional information about how to hack D&D to be less lethal (especially less randomly lethal) without making it bloodless altogether.

• Tweak the reward systems to match your goals. A great way to get extra focus on combat goals is to reinforce them with rewards. Hand out XP for following your motivations or general campaign goals (here's an example; here's another) or do away with XP altogether and just have the PCs level together at whatever rate feels appropriate to you all.

• Trust your players to work with you. I think this bears repeating. A lot of advice for D&D/Pathfinder assumes that the GM is the boss and the players are just there for the ride. Also it pretty much assumes that players are terrible and you'll need to trick and cajole them to get a good game. Seriously: just don't bother with that noise! Work together to make the game you want. It's like ten times easier.

"Bad Decisions"

This is a bit of a side issue, but I think it's important to point this out.

Think long and hard about what "bad decisions" are, and whether you truly want to "punish" players for making them.

For starters, how important is challenge-based play in your game? Are you actually looking to test the players' tactical skills thoroughly?

On top of that, is that critical hit truly a "bad decision?" How 'bout failing a save? Was the player actively tempting fate or just, you know, facing danger like a hero is supposed to? Heck, maybe you quer the characters to tempt fate a bit. I mean, by most sane standards, indo em aventuras is a terrible decision. But it's presumably one that you want the protagonists to keep making consistently.

In character-driven dramatic play, characters making bad decisions is usually... good. They reveal character and push the story in cool unforeseen directions. Bad decisions should have serious fictional consequences — without that, the story is toothless and the "decision" isn't a real one — but they're not something to "punish" at the table.

Defeat without death can seem like a non-punishment, but there are ways around that.

Save the characters by having them knocked unconscious and/or captured, but then the villain's evil plot proceeds without them to stop it. By the time they get out, the village they were protecting has been burned to the ground, or the marriage ceremony between the princess and the demon is complete (alas, now they are too late to save the poor demon), or the Dark Lord was able to retrieve his magic hat and slay the King. Their reputation as heroes (assuming that's what they aspire to be) starts to fall into tatters. Too many failures like that and they may well end up known throughout the kingdom as undesirables and a source of bad luck.

Find ways like that to give them alternative ways to feel they've lost. It doesn't always have to be death - or at least, not their own.

As Bankuei points out, many fights are not necessarily to the death. If a character is knocked out, the enemy may very well not finish them off preferring to loot them or capture them or the enemy may even just return to their business with the threat removed. That all depends on the type of the enemy and why they are fighting in the first place. One specific example is that you could replace a total party kill with the enemy taking them hostage. This was actually rather common in history with significant people. They would then be released, but perhaps only after owing a rather large debt to a local noble that wants them under his thumb.

But there are other options available for when the enemy really would kill them (or when the enemy itself doesn't get that choice):

If you are willing to risk them figuring out that you will prevent deaths, then you have a lot more options. For instance, in a literal Deus Ex, a deity may step in and heal the character to full at the moment of death, but exact a price for doing so. At the simplest, the deity may demand a large sacrafice that will cost the character in a material sense. More complex and interesting, the deity might either hand off a significant quest or place restrictions on the character for the rest of their life.

If you are willing to flat out tell them you are stepping in, you could declare that reaching the point of death will not kill them, but will give them a long term injury that will take substantial time and resources to overcome (the nature needs to depend on the level of the characters and what they have access to.)

When I GM I frequently declare player death to be literally impossible, but let the players know that a price will be exacted if the character otherwise would have died. Then depending on the needs of the story I tend to use variations on one of those techniques.

You can take ideas form both games and movies. For example, secretly give all your characters a certain amount of free deaths (2 or 3). after that, every time they die, their stats drop. Or even every death causes their stats to drop, usually a pretty heavy penalty. Then, in order to regain their previous power, they must go to a holy place, and have their original power restored.

Alternatively, every time a character dies, they need to make a will check to come back, starting with a simple check, then to a moderate check, and finally to a hard check. This way, they can come back, but only a certain amount of times. This can be explained by a decay of their soul, or something similar.

The best technique I can suggest is simply to steal shamelessly. A character might notice that you stole an idea or two, but that rarely gets in the way of the fun.

In our Pathfinder campaign, it's possible to avoid death by spending two Hero Points. Then instead of dying you get stabilized with -1 hit point, able to be brought back with Cure light wounds spell, &c.

As you can have at most 3 Hero Points at once, and they are awarded at end of campaign or story arc, they are not expended lightly.

On the other hand, one Hero Point is also given for writing an enjoyable passage about the gaming session in our wiki-based campaign's chronicle. Since establishing this rule, the quality and frequency of the stories has risen dramatically :).

Sometimes alternatives aren't the best answer. But face it, in most fantasy driven games there are myriad ways to suspend death. In most cases, death is about as permanent as ice-water. There are spells that cheat death, creatures that can grant life, and ultimately nosey deities that may intervene.

I subscribe to the Chris Perkins philosophy about death - ask the player. It may be that the player is dissatisfied with the character and is ready to move on to something new, or just wants to try something new. In that case, let the character die, and let the player move on. In other cases, the player may have invested a huge amount of time on the background and preparing a personality and is super attached to the character. If that's the case, then figure out some way that the character can come back, but always do this with some sort of caveat, or story enhancement.

Some people from the city recognize the down character and bring him/her back to the city where the priest raises them. After which the city leans on the character as a protector, or points out that the character owes them. A deity raises the character only to have them provide information about their own party, or demands the character take on another person/creature/entity as their charge, and promise no harm will come that charge. (If you can then make the charge something they would normally not protect, say a succubus looking to change her ways. It makes it even more interesting)

The point being that non-permanent death can lead to an even better character, and enhance the story of the game. But always, always lean towards what's going to make the game more fun for the player.