Different people have different preferences regarding character deaths in pen & paper roleplaying games. My previous campaign had about 1 death per year, and that was fine by me. I don't want players to think that their characters can't die, but I do want character deaths to be rare and special. A good part of the motivation for roleplaying games in general comes from character progression, and death puts a damper on that progress and thus on the motivation.

Now 4th edition Dungeons & Dragons basic concept was one of extreme balance. Two characters of the same level have the same number of daily, encounter, and at will powers, regardless of class. While there is some necessary difference between ranged and melee classes, a ranger shooting people with arrows has powers that are very much comparable with a wizard shooting people with spells. Between the balanced classes and powers, and the splitting of combat into more rounds, 4th edition Dungeons & Dragons fight were relatively predictable. Good tools in the DMG helped me to repeatedly design challenging fights in which players were afraid for their lives, but ultimately succeeded in a satisfying win.

5th edition is lacking those tools. Or rather the DMG is trying to provide tools, but they don't really work well for different cases. That is not because the developers didn't put effort in trying to build the encounter design tools, but rather due to the different combat math of 5th edition: 5th edition was built on the idea that extreme balance is boring, and we should have the choice between classes that deal a more constant stream of damage for a long time and classes that are essentially glass cannons, able to deal massive damage for a short time before running out of steam. And because the goal was to make combat shorter, in each round a higher percentage of the total hit points of player characters and monsters needs to be dealt. If a fight is designed for 3 turns, the players need to deal damage equal to a third of the total hit points of all enemies each turn. With these high stakes each turn, a single fumbled attack or a critical hit can determine whether the players win or lose the fight. Unpredictability makes for difficult encounter design.

The tools that 5th edition Dungeons & Dragons has to build encounters are built around a concept of "challenge rating" of the monsters. Which is more or less equivalent to the 4E concept of having an experience point budget for monsters in an encounter. There is a table in the 5E DMG on page 274 in which for each challenge rating you can find both defensive stats (AC, hit points) and offensive stats (to hit, damage). The overall challenge rating is the average between defensive CR and offensive CR, so your monsters can be built more aggressively or more defensive. As long as the enemies are simple monsters with standard melee or ranged attacks, that works reasonably well. There is a lot of disagreement what the correct monster challenge rating should be for a party of n characters of level X, but that has more to do with different preferences of deadliness, as discussed above.

But when I tried to use that challenge rating tool for the villain of the adventure I am planning, an evil sorcerer, I quickly discovered the limitations of that approach. In a world where a player sorcerer or wizard has very low defenses and a few very powerful spells, an enemy sorcerer with high hitpoints and constant damage every turn wouldn't really look believable. The table doesn't even tell you for a given challenge rating what level that spellcaster enemy should be. The monster manual, which has some examples of NPC enemies, suggests that caster level is about 1.5 x challenge rating. But even if I decide on a caster level, let's say making my evil sorcerer level 5 to give a good challenge to my group of players of level 3, I'm still far from having done my work.

The NPC enemies in the Monster Manual all have far higher hit points than a player character of the same level would have. A player sorcerer level 5 would have around 30 hit points. That would be rather low for a boss encounter main villain, because he would risk being killed on the first turn by a level 2 spell or some lucky rolls from the players. On the other hand the suggested hit points for a monster of that challenge rating are over 100, which seems too high for a sorcerer. I think the good compromise is somewhere in the middle of that.

Much more difficult to assess is the damage output of the villain. A level 5 sorcerer has access to level 3 spells, but there are huge differences in the power of different spells of the same level. A fireball, dealing 8d6 damage in a large area, could potentially we a one-shot total party kill for a group of level 3 characters if you consider that a level 3 fighter has on average 28 hit points, which is just the average damage of that fireball. Other level 3 spells for a sorcerer (except lightning bolt) are comparatively harmless and would just constitute minor inconveniences to the player characters. So the caster level and hit points aren't enough to determine the challenge rating of a NPC villain, it depends very much on the exact spells chosen for him.

What I will basically have to do is to determine the attack spells that the villain is likely to cast in three rounds of combat, estimate how many players might get caught in these spells, determine the average damage per turn from that, and then look up to what sort of offensive challenge rating that corresponds. Not an easy exercise, and there is an obvious danger that the real combat will be very much different from that estimate.