NPC Tactics: Magical Specialists

Volo’s Guide to Monsters includes stat blocks for 11 different magic-using specialists: wizards from eight different schools and warlocks of three different patrons. The wizards are all at least level 7; the warlocks, even higher. There are also a level 9 war priest, a level 10 blackguard (antipaladin) and a level 18 archdruid. Every one of these spellcasters has a different repertoire of spells. To come up with individual tactics for each of them would take me the next two weeks.

Rather than tackle each one separately, then, I’m going to share some rules of thumb for developing tactics for a spellcasting NPC.

Begin with ability scores

Most spellcasters will want to keep their distance from the action. I say “most” because I’m all about busting stereotypes—having a beefy, brawny guy also be the best dang transmuter in the duchy is exactly the sort of thing I’d do. But more typically, spellcasters—especially wizards—will have low to average Strength and below-average to above-average Constitution, relying on a higher Dexterity to avoid taking damage and their spells to deal it out. Few of their spells require them to get up-close and personal (that’s more of a cleric/paladin thing), so they’ll prefer to keep their distance. What distance is optimal? Somewhere between 30 feet and the range of their shortest-range spell.

Calculating spell damage

Spell damage depends on two things: number of targets and whether it’s a spell attack or a spell that requires a saving throw. “Expected damage” is the term I use for a calculation of the average amount of damage done times the probability of dealing it. Based on the simplistic premise that the caster’s spell attack modifier and the target’s armor class will cancel each other out (I don’t use this premise myself, but you can use it as a shortcut), the probability of hitting a target with a ranged spell attack is roughly 50–50, so the expected damage is half the average damage on a hit. (For example, scorching ray, cast at 2nd level, deals 2d6 fire damage per ray, or an average of 7 hp. The spell’s expected damage is therefore 3.5 hp per ray.) Since spells that require saving throws usually do half damage on a success and full damage on a failure, multiply their average damage by three-fourths to calculate expected damage.

To calculate the damage of an area-effect spell, use the Targets in Area of Effect table on page 249 of the Dungeon Master’s Guide to figure out how many targets the spell should hit, then multiply the expected damage for one target by that number of targets. Assume that the spellcaster won’t cast that spell unless he or she can catch at least that many targets in its area of effect.

Spell attack modifier and spell save DC are chained together, so you won’t ever get a caster with an unusually high spell attack mod and an unusually low save DC, nor vice versa. Because of this, you’re not going to end up with casters who either specialize in spell attacks or avoid them. You will see an overall preference for save-dependent spells among all casters, simply because such spells usually do some damage even when the targets make their saves, thus increasing these spells’ expected damage by about 50 percent.

Generally speaking, a spellcaster will bring out his or her biggest guns first. But there are exceptions. A particularly cocky spellcaster might prefer to expend as little effort as possible at first, ramping up the damage only when he or she realizes that the opposition aren’t pushovers.

Time’s not on your side

Fifth-edition Dungeons and Dragons assumes that a typical combat encounter lasts three rounds, maybe four or five. Spellcasters often have more spells than time to cast them. Therefore, it’s really important to identify their most efficacious spells before the encounter begins. As of this writing, the EN World forums host a series of useful guides on the various character classes which, among other things, offer assessments of spells color-coded by efficacy (here are links to the wizard, warlock, sorcerer, paladin and bard guides—unfortunately, the cleric and druid guides don’t include spell lists). Any spell in your NPC caster’s list that’s coded red or purple, just cross it off. Any spell that’s cast as a ritual, or that has neither a damaging effect, a defensive effect nor a movement effect (e.g., most divination spells), cross it off.

On the other hand, spells that enhance a character’s action economy are worth their weight in gold. These are the spells that are cast as either a bonus action or a reaction, which means they can be combined with other actions (though leveled spells cast as bonus actions can’t be combined with other leveled spells, only with cantrips). Paladins have a lot of these: compelled duel, divine favor, shield of faith, magic weapon and all the smite spells. Clerics have healing word and mass healing word, sanctuary, shield of faith, spiritual weapon (this one is so good, you have to assume that any NPC who has it casts it before doing literally anything else) and divine word. Druids have healing word, flame blade, grasping vine and the shillelagh cantrip. A bard may have feather fall or healing word. Wizards and sorcerers have expeditious retreat, feather fall, shield, misty step and counterspell, and wizards additionally have magic weapon, which they should cast on their staves or daggers when an enemy gets within melee range of them. Warlocks have expeditious retreat, hellish rebuke, hex, misty step and counterspell.

Any bonus-action spell that’s applicable to the situation, plus a damaging cantrip, is at least as good as a leveled spell that takes an action to cast. Spiritual weapon, I have to mention again, is truly spectacular, because it keeps going without concentration, providing a bonus action every turn, until the combat encounter ends. (Technically, until 1 minute has gone by, but that’s a 10-round combat encounter. Combat never lasts that long.) And wizards and sorcerers should always keep a spell slot in reserve for shield or counterspell.

The value of spell slots

I value spell slots not by their level but by their scarcity. The rationale for this is that a fireball cast using a 4th-level spell slot is as good as any 4th-level spell. However, higher-level spells tend to have effects that lower-level spells don’t duplicate. In other words, if all you want to do is hurt someone with fire, you can use fireball, or you can use scorching ray, or you can even use fire bolt. But if you want to become invisible and stay that way even if you’re attacking or casting spells, you’re going to have to cast greater invisibility.

A level 9 wizard has four 1st-level spell slots; three 2nd-, 3rd- and 4th-level slots; and one 5th-level slot. That wizard will only use his or her 5th-level slot to cast a 5th-level spell, period. Why? Because there’s no other slot he or she can use to cast that spell. It’s unique.

On the other hand, those three 4th-level slots can be used to cast 4th-level spells or 2nd- or 3rd-level spells boosted to 4th level. This is where expected damage calculations come into play, because, for instance, the 3rd-level fireball spell does more damage when boosted to 4th level than the 4th-level ice storm spell does. Ice storm is 4th-level because it produces an additional movement-impeding effect; if you don’t need to impede movement, you may as well just boost fireball for the extra damage. Wizards tend to be smart cookies, and they know what their spells can do. They’re good judges of which spell suits the occasion better.

The wizard won’t waste a 4th-level slot on boosting a 1st-level spell, though. Those are simple, dime-a-dozen utility spells. There’s nothing they can do, even boosted, that a higher-level spell can’t do better. And the wizard’s four 1st-level spell slots aren’t good for anything except casting 1st-level spells. So anytime the wizard needs to cast a 1st-level spell, he or she will use a 1st-level slot for it.

Note that warlock spells are always boosted to the caster’s highest spell level, no matter what. Warlocks also have very few spell slots, so they have to make every spell count.

Special features

Each NPC spellcaster archetype has a special feature related to his or her specialty. These I’ll look at individually, because they affect the spellcaster’s tactical decisions in different ways.

  • The abjurer has Arcane Ward, a recharging magical barrier that soaks up damage that would otherwise affect him or her. The main effect of this ward is to allow the abjurer not to retreat if charged by a melee attacker. This ward, plus a high Constitution, lets the abjurer survive in the middle of the fray for longer than a spellcaster normally would. Since the Arcane Ward is recharged by the casting of abjuration spells (helpfully marked with asterisks in the stat block), an abjurer who’s under attack by a melee opponent may cast abjuration spells (e.g., mage armor) specifically for the purpose of recharging the ward. However, the spells he or she casts in straightforward self-defense (such as shield and stoneskin) may be enough to accomplish this by themselves.
  • The archdruid has Change Shape. One word: mammoth.
  • The blackguard has Dreadful Aspect, which affects every enemy within 30 feet. He or she therefore has a strong incentive to wade right into the middle of the fray. Combined with the high-Strength, high-Con ability contour of a brute, this suggests a pure tank style.
  • The conjurer has Benign Transportation, allowing him or her to teleport up to 30 feet as a bonus action each time he or she casts a conjuration spell (marked in the stat block with an asterisk). Useful as a way of maintaining distance, but also useful for getting out of trouble by swapping places with a tougher ally.
  • The diviner has Portent, which is a bailout measure that doesn’t affect his or her tactics in any way much more useful as a risk reducer than I realized way back when I first wrote this. In particular, see Fuligin’s comment below.
  • The enchanter has Instinctive Charm, also a bailout measure.
  • The evoker has Sculpt Spells, which lets him or her semi-safely cast damaging area-effect spells with allies inside the area of effect—kind of a big deal, especially since the evoker has a lot of those spells.
  • The illusionist has Displacement, which he or she should be using all the time, especially considering that it’s a non-spell bonus action, so it can be combined with a leveled spell in a single turn.
  • The necromancer has Grim Harvest, which gives him or her an incentive to rely on necromancy spells (marked in the stat block with asterisks) as much as possible.
  • The transmuter has Transmuter’s Stone, which has one of several different effects, and it’s cheating to decide that it starts off with exactly that effect that will be most effective against your player characters. Pick one that suits the transmuter’s stats and personality. That being said, a transmuter with a speed boost stone will use it to Disengage from melee and maintain a safe distance from enemies; the other options won’t affect the transmuter’s tactics. However, since the transmuter can change the effect of the stone on the fly as a free action after casting a transmutation spell, it can make informed decisions once it sees what the PCs are capable of. In a subterranean setting, against a party comprising only humans and halflings, it may make sense to choose the darkvision effect, then take out the lights. Against chargy melee fighters, the speed boost is helpful. Against one or more rogues, the Con save proficiency offers resistance against poison. Against a magic-user casting spells that do elemental damage, or against magic weapons that do the same, resistance to the appropriate element is de rigueur. The transmuter is intelligent enough that he or she can make an educated guess about the PCs’ predilections after one round of combat, without having to actually suffer damage from any of the aforementioned situations.
  • The warlock of the Archfey has Misty Escape, a handy bailout measure, but also one that allows him or her to make a subsequent attack from hiding, an opportunity he or she should make the most of.
  • The warlock of the Fiend has Dark One’s Own Luck. The time to use this is when the warlock is moderately wounded (reduced to 54 hp or fewer).
  • The warlock of the Great Old One has Whispering Aura, a passive effect. Since it has a range of 5 feet and affects everyone the warlock wants it to affect—and since the warlock also has a high Constitution—he or she has every motivation to wade right into the midst of his or her enemies.

When to flee

It takes a lot of work to become a wizard, and that’s not an investment that wizards take lightly. They’re also, usually, fragile flowers. They need only be moderately injured (reduced to 70 percent of their maximum hit points or fewer) for discretion to seem like the better part of valor. They tend not to have great ACs or move unusually quickly, so their best choice is to Dash (action) away without further delay, unless they have some kind of spell or feature that can cover their escape, in which case they’ll use that spell or feature, then Dash on their next turn.

Other spellcasters live closer to the edge (and also tend to be more durable). It takes being seriously injured (reduced to 40 percent of their maximum hit points or fewer) to induce them to run away. A paladin probably won’t run away at all; he or she will bravely fight as a rear guard while his or her allies retreat, and only once they’re safely removed from the battle will he or she Disengage and withdraw.

Next: yuan-ti, revisited.

18 thoughts on “NPC Tactics: Magical Specialists

  1. Keith,
    I cannot tell you how useful this all is. I’m running a session coming up where I’m using a conjurer as a lieutenant (or captain?) of the BBEG, along with some orcs. I am desperate to use Conjure Elemental, but it’s so delicate because of the Concentration aspect, and when you lose concentration (probably a 50/50 proposition if hit) the elemental becomes hostile.

    So I solve this (I hope!) by tweaking the conjurer’s spells a bit. Gave him Banishment and Greater Invisibility (and got rid of two other spells of the same level) for if/when the elemental goes berserk. Since he’s going to conjure a fire elemental, the conjurer can use movement and benign transportation to move so that some/most of the PCs are between the elemental and him. The elemental will move directly through the PCs and set them on fire. I’ll do this as often as I can, essentially leading the elemental through the enemies. And once Concentration is gone, I can use other Concentration spells if need be.

    Greater Invisibility will confuse the elemental enough that it will likely go back to hitting whoever is hitting it (likely the PCs and definitely not the orcs).

    Banishment is the last ditch emergency option if nothing else works. Thanks for getting me to think about the tactics in a new and deeper way.

  2. Possible tactic In regards to the archdruid- boosted call lighting to level 5 or 6, concentration, wild shape to an invisible stalker. Fast, nigh undetectable, and dealing 5d10 or more lightning damage each turn

  3. Hello! Love your blog; stumbled on it a week ago and have been reading it obsessively ever since.

    On the subject of spell casting NPCs, I’d like to raise one small point. And before I do, I want to say that I understand your blog is specifically about *tactics* – behavior in combat – and not building encounters in general. But I feel it still merits mentioning.

    When preparing an NPC spell caster, I think it’s important to consider what they were planning to do that day BEFORE they crossed paths with the player characters. Specifically, does it make sense for their entire list of prepared spells to be combat-oriented?

    PC spell casters often walk around with at least one non-combat spell taking up space in their inventory. Detect Magic, Comprehend Languages, Knock, Identify, etc.: these and many more.

    Unless the NPC spell caster woke up that morning and decided “I’m gonna go kill some player characters today”, it doesn’t make sense for their entire list of prepared spells to be combat-oriented. Of course there are exceptions, like the war mage on the battlefield. But generally, most spell casters should have a few non-combat spells prepared, reflective of what they’re doing that day.

    Thanks for your time and for all your work maintaining this fantastic blog!

    1. Absolutely! I agree 100 percent. When I say “cross it off” with respect to a spell, I don’t mean, “Don’t give the caster this spell.” I mean, “Once the fur starts flying, you no longer need to think about this spell at all—so don’t.”

      1. Very late to the party, but I was wondering about this recently. I’ve been preparing a partial spell list for NPC magic users based on a sort of logic like this (i.e. assuming that only some of their prepared spell list is useful in combat and that they have other spells that just aren’t relevant). Up to now I’ve been doing so at roughly 2/3 of their normal total with a slight preference for higher level spells. I was thinking of cases – as above a military battle mage or retired scholar perhaps – who might go further one way or the other. Any thoughts on how that could/should be handled (or on whether its a good idea at all – I admit I’m partially doing it as a time saver for me).

  4. Hey, great stuff!

    I wanted to pick your brain about the Enchanter’s 5th level spell: Hold Monster. This is his only 5th level spell, but he’s also got Hold Person below and the only difference is Monster works on “creatures” instead of being limited to “Humanoids”.

    Would it be worth using this spell on a party of presumably humanoid races? Would you keep it in reserve for effecting one low-wis summon from an enemy spellcaster? Or would it be more effective to Upcast the 3rd level Hold Person and maybe get 3 low-wis PCs? My gut says the latter, but maybe I’m missing something about Hold Monster.

    1. It would be a waste to use hold monster against even a single humanoid, unless a spellcaster were boosting it by at least one spell level and targeting a nonhumanoid creature at the same time. If an NPC enchanter is casting hold monster against the PCs’ party, it will be primarily to paralyze a nonhumanoid ally of the PCs, most likely either a summoned creature (such as an elemental or fiend) or a mount. And frankly, I think it would also be a waste to use it on a mount unless the mount were something extraordinary, like a unicorn.

  5. Hey Kieth,

    Nice work, just found your site and it is definitely scratching an itch I’ve had for a while! I love the tactical nature of D&D and most of my friends don’t talk shop like this very much.

    One note on the Diviner’s “Portent”. This can be a really powerful offensive mechanic as well to essentially impose disadvantage on another creature’s saving throw. Might be nice for getting a Mass Suggestion to stick on the tougher enemies.

    Just a thought, keep up the great tactical analysis!

  6. Why would the diviner’s Portent only be a bailout measure? It can be used offensively to great effect.

    It enables the diviner to, depending on whether their Portent rolls are high or low, make an ally’s high-risk, high-reward attack automatically hit, or make an enemy automatically fail a saving throw against one of the diviner’s own spells in a “spell-must-stick” situation – especially a spell which the enemy might’ve otherwise saved against easily. Specifically low Portent rolls enable the diviner to use spells that require saving throws on opponents who have a high saving throw bonus in a particular save REGARDLESS of that defense, if the Portent roll is low enough. That is pretty significant, if you ask me – spells that are save-or-suck and/or don’t allow saves after the initial one profit from this the most. Getting a Suggestion or a charm effect to stick on a character with high wisdom, such as a cleric or druid, or nailing a squishy rogue with a Disintegrate or trapping them in an Otiluke’s Resilient Sphere despite their ridiculous dexterity save bonus is awesome, plain and simple. This feature essentially allows the diviner to ignore their enemy’s ability score distribution (which would otherwise dictate which spells could be used against them effectively) twice or three times (depending on level and if they rolled low enough) a day, granting them far greater versatility, due to being able to use a very effective spell against an enemy it couldn’t normally be used against, and automatically imposing that spell’s effect. How is it that you dismiss a feature capable of THAT as a mere bailout measure?

    However, of course there is a small caveat in the form of spells that allow multiple saving throws – you wouldn’t want to use, say, Maximilien’s Earthen Grasp on a fighter with maxed strength regardless of the diviner’s Portent feature, because after the initial save they’d just break out of it. It does however cause them to waste an action breaking the restrainment; whether that’s a worthwhile trade depends on the other spells the diviner can use, and the action economy of and threat posed by the resistant enemy, as well on whether there are allies of the diviner near the resistant enemy who get to take turns before them and can take advantage of the, well, advantage granted by the resistant enemy being restrained.

    Dan also mentioned this, however I disagree with their comparison that it’s like imposing disadvantage – you outright replace whatever the enemy would’ve rolled, and depending on your Portent roll and the enemy’s save bonuses, if you play your cards right it’s an automatic fail, not mere disadvantage.

    Yes, Portent does allow the diviner to grant an auto-success on an ally’s saving throw against a save-or-suck effect, or make an attack that would kill the diviner miss automatically, but it has so much more potential than that. I’m surprised and disappointed you missed or ignored that, considering you’re usually so perceptive, creative and efficient in how you utilize monster features.

      1. Thank you!

        I am planning on playing a Diviner in my first ever game once I find a group, and I’ve been wondering whether legendary resistance can be used to succeed on a legendary monster’s failed save if it was “failed” by a Diviner using Portent to replace whatever the monster would’ve rolled. Nothing I’ve found so far specifies anything about how these features interact with one another.

        Can the resistance be used, because technically all the Diviner did is make them fail the save, or would the Diviner be replacing the automatic success with the Portent roll? Having Portent be overridable by legendary resistance makes the feature much less strong against a monster who has such resistances – until they are used up, that is.

        How would you rule in this situation; both Keith and anyone else who reads this? Any other thoughts?

        1. I think Legendary Resistance has to win. Portent can impose a numerical result, but Legendary Resistance says, “If —— fails a saving throw, it can choose to succeed instead,” which is independent of the result of a die roll.

          1. Yeah, that’s what I thought as well, I just wanted to check. Thanks for the near-instant response!

    1. Hi!

      I was re-reading through the Portent Ability on the NPC Diviner, and the Player’s Divination Wizard Portent Ability, and mechanically they are written quite different. You can correct me if I misunderstood, but it seems like for the player’s feature, they roll at the beginning of the day and can use the ability tactically offensively or defensively. But, the NPC Diviner can only roll the d20 when they see an another creature make an attack roll, saving throw, or ability check. So they can’t plan ahead. Instead of using this ability judiciously, it’s more akin to a disadvantage as Dan described. It’s still strong, but not nearly as strong as the player’s version of Portent, since it has to be rolled and used in the same instant.

  7. For the Blackguard, I have one fighting on a train, so she won’t need a steed. What should I replace find steed with?

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.