What does a 20th-level dungeon look like?

That’s it. That’s the article. The simple truth: I was thinking about this earlier in the week and thought it would be interesting. If you don’t like this article, I have others!

For most of us, this is essentially a thought experiment. Very, very few games get to 20th level. My Mad Mage campaign didn’t; my Tomb of Annihilation and Curse of Strahd campaigns didn’t come close. In fact, in 20-plus years of playing D&D, only the homebrew ‘Rise of Zargon’ campaign I ran in lockdown did, and that was with fairly generous session-based advancement. For most of us, if we’re playing at 20th level, it’s probably a one-shot: ‘let’s go fight the tarrasque’ or something like that.

I have written before about high-level adventures, but in this article, I wanted to think specifically about dungeons: site-based adventures, the kind we might have played in Tiers 1, 2, and 3, not big, ‘pull-out-all-the-stops,’ campaign-ending finales. What does it feel like to have four 20th-level characters – perhaps a fighter, wizard, cleric, and rogue – and explore some underground location, room by room?

I also thought this question had fun implications for world-building. Who is at the top of the food chain on the Material Plane? Is it even possible to run meaningful adventures at such high levels in the Material Plane? Or, is it assumed in 5e that your 20th-level party are running around the Multiverse at this point?

So that’s the background. Let’s go plan a 20th-level dungeon and see what it looks like.


While it’s possible to build D&D games around the two pillars of exploration and social interaction, most of us build adventures around the third pillar: combat. So let’s start by thinking about the kinds of encounters our party might fight.

For this, I used Donjon’s excellent 5e encounter size generator as a starting point, which is itself based on the encounter-building guidance in the Dungeon Master’s Guide. Yes, the maths might be a bit wonky, and magic items play havoc with it, but let’s go through the motions anyway. (The encounter-building rules in Xanathar’s are perhaps a bit easier to work with but end up with more or less the same result.)

It’s worth emphasizing here that the numbers work better if we account for a series of encounters. It can be tempting to throw a ridiculous encounter at the party when they are fully rested to see if they can handle it, but it’s probably best to avoid this temptation: it just leads to parties resting more often. (I also like the Sly Flourish house rule – if you can call it a ‘rule’ – that long rests should be within the gift of the DM and what makes sense in the fiction. But I appreciate not every group plays like this.)

Large groups will also have a huge impact on party power level. For example, adding a single character at Tier 3 is equivalent to one of those characters gaining five levels. I’m going to assume a party of four players, but if your group is bigger, you will need to adjust accordingly.

I have a few other underlying principles here:

  • Most encounters should be ‘medium’ or ‘easy’ difficulty. Justin Alexander has a good article as to why.
  • Most encounters should revolve around three to twelve bad guys. Solos get kerb-stomped, and mobs are hard to run.
  • Most encounters should make sense for the environment and organization of the creature concerned.

Of course, the key word for all of these is ‘most’. Sometimes you want a ridiculous deadly encounter or a solo fight; sometimes you want wave after wave of low-level enemies; sometimes you want a monster in an unexpected environment, like frost giants in Chult or githyanki space pirates in Undermountain. But these are probably exceptions.

For environment and organization, I tried to lean on existing lore. The 5th edition Monster Manual was a starting point, but I also used the advice in the 3.5 rulebooks. Following this, I pruned out a few monsters that didn’t feel ‘right’ at Tier 4: perhaps arbitrarily, but I think I could defend myself on this. For example, I’m not sure a demilich should be an ‘easy’ encounter; it just feels lacklustre, almost as if we’re disrespecting the creature’s place in the lore.

Incidentally, I limited myself to creatures in the 5e Monster Manual. By my reckoning, there are now well over a thousand stat blocks in official 5e products, and that’s well beyond the scope of this article. Besides, I kind of want to keep it classic: the monsters most readers will already be familiar with.

I generated a list. Honestly, the full list is probably too much for this blog – I can maybe release it on my Patreon if there’s an interest – but the big three that come up again and again are essentially fiends, giants, and slaadi.

More on monsters in a minute. What about adventure environments and loot?


One of the advantages of playing at 20th level is that you can place an adventure pretty much anywhere you want. If you’ve wanted to do a Diablo-style trip to Hell, or a voyage through the Astral Plane, or a jaunt to the Shadowfell, now you can. But for the purposes of this article, let’s imagine we want to keep our adventure on the Material Plane. After all, it’s the world our players have spent the last 19 levels getting to know. However, it might mean that some encounters will make more sense than others. Dragons? No problem! Giants? Great! But demons, slaadi, yugoloths? Also fine – they just might need a bit more explanation.

Anyway, back to the Material Plane. If there’s a 20th-level dungeon on the Material Plane: why? What does it look like? Why has no one plundered it before? Why don’t the creatures within pose an existential threat the rest of the world?

A few thoughts here:

  • Cool dungeons are usually old. 20th-level dungeons might be some of the oldest of all: so old, they’re almost forgotten. No one plunders them because no one knows where they are.
  • Cool dungeons are hard to reach. A 20th-level dungeon could be extremely inaccessible and isolated: on the top of a mountain, at the bottom of the ocean, surrounded by desert, or deep in the Underdark.
  • Cool dungeons are often threatening and made to seem off limits. A 20th-level dungeon should be one of the most intimidating and scary-looking adventure sites in the world.
  • Cool dungeons are usually contested. This is a frequently used trick in Dungeon of the Mad Mage. Sure, an adventure site could be ruled by a single creature type – but isn’t it more dynamic and more interesting to have two or more factions competing for the space?

When thinking about the dungeon environment, it’s also worth thinking about elements like skill challenges, traps and hazards, improvised damage, and utility magic.

  • In terms of skill challenges, most characters at 20th level will have maxed out their primary ability scores and have a +6 proficiency bonus on a range of skills. A rogue with expertise and reliable talent could be rolling +17 on their favourite skill checks, with a floor of 27. An ability check of DC 15 is no longer ‘medium’ difficulty: it’s very easy. For trained skills, ‘very hard’ (DC 25) is the new medium, and ‘nearly impossible’ (DC 30) is, well, not impossible. We should take this into account for locks, traps, secret doors, and so forth.
  • Speaking of traps, I’m not sure official sources even give examples of Tier 4 traps and hazards. Xanathar’s only goes up to Tier 3 with the poisoned tempest (p 120) and sleep of ages traps (p 114). Whatever our traps and hazards look like, they need to be deadlier than these!
  • As a starting point, we could look at the advice for improvised damage in the DMG (p 249). A ‘setback’ at Tier 4 should be dealing around 55 (10d10) damage: equivalent to whirling steel blades, walls crushing together, or wading through a lava stream. To think of it another way: what would be deadly for a Tier 2 character is merely a setback at 20th level. A ‘dangerous’ trap would be more like 99 (18d10) damage: equivalent to being submerged in lava, or the damage one might take from a flying fortress crashing into you. A ‘deadly’ trap at Tier 4 would deal 132 (24d10!) damage. The DMG gives examples like ‘Tumbling into a vortex of fire on the Elemental Plane of Fire’ and ‘being crushed in the jaws of a godlike creature or a moon-sized monster.’

We can keep all of this at the back of our minds for now. One more element to think about: loot!

Magic items and treasure

I’ve written before about magic items in 5e. It’s a pretty stats-heavy post, though, so let’s pluck out the key highlights:

  • By the end of 19th level, most adventurers will have a magic item of each rarity and perhaps two of uncommon rarity.
  • They may well have a variety of consumable items at their disposal, including very rare and legendary items (eg, a 9th-level spell scroll or a potion of storm giant’s strength).

Based on the maths in DMG, I would expect to award the following at 20th level:

  • 1 rare consumable (eg, a 4th- or 5th-level spell scroll)
  • 2 very rare consumables (eg, a 6th- or 7th-level spell scroll)
  • 1 legendary consumable (eg, an 8th- or 9th-level spell scroll)
  • 1 permanent legendary item (eg, the deck of many things, holy avenger, or staff of the magi)

If this seems a bit stingy, bear in mind that a) the party will already have a ton of magic items by this point, and b) you don’t want to award too many magic items at the end of a campaign when the party don’t have any more levels to gain!

Which magic items? That’s up to you and your group. But I have some suggestions.

What about treasure? Honestly, by this point, it’s starting to become a little bit meaningless. Over the course of 20th level, I would expect a party to find maybe 644,000 gp and maybe another the equivalent of another 31,925 gp in art objects.

OK: we’ve thought about our encounters, environments, and loot: what kinds of adventures can we have at 20th level?

Putting it all together

These are just a handful of sketches – hooks, prompts, call them what you will – but I think any of these would work well as a 20th-level dungeon site:

  1. A mighty fire giant fortress inside an active volcano. Enemies: hell hounds, iron golems, fire giants (obvs) and red dragons. Maybe githyanki knights have come here to steal dragon eggs?
  2. A mountaintop citadel, home to an ancient dragon’s hoard but now being raided by frost giants and, for some reason, yugoloth mercenaries. What treasure do they seek?
  3. An ancient desert tomb. Enemies: sphinxes, golems, flameskulls, wights, wraiths, and maybe some blue dragons. Maybe a band of efreets or djinn have also sought out the tomb for what it contains?
  4. A yuan-ti city subjugated by black or green dragons (depending on whether it’s more of a jungle or a swamp). Nagas, giant apes, and various undead could also feature.
  5. Perhaps the easiest: the deep Underdark. Caverns filled with cloakers, hook horrors, grells, aboleths, beholders, mind flayers, stone giants, fomorians. There’s a reason why the Underdark is almost the ‘default’ megadungeon setting: it makes internal sense in a way that few other settings do.

Bonus: demons! An open portal to the Abyss could make a fun emerging threat in pretty much any existing dungeon site. Best of all, it doesn’t even need to really make sense! Maybe the fire giants are having to fight off an attack from a balor and friends.

All of these dungeon sites could work pretty much as they do in Tiers 1–3 with chambers, connecting corridors, traps, treasure, and puzzles. Players have a few tricks up their sleeves to get around them, like teleportation, divination, and safe places to rest like Leomund’s tiny hut, but these dungeons should provide a decent challenge regardless.

Hopefully this gives you a sense of how superheroic Tier 4 can be. Your party are essentially demigods now. Take your normal fantasy and turn it up to 11.

If you like what I do, please support me on Patreon. You can find me on Facebook at scrollforinitiative, Twitter @scrollforinit, and Instagram @scrollforinitiative. And if you want to make my day, you can buy me a coffee here.

Never miss an article

Unsubscribe at any time.

Leave a Reply

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