The Darkest Dungeon enemy file carries 99 entries across 19 class tags, and those tags are not lore dressing. Unholy enemies average 156.4% bleed resistance across the eleven entries with full stat blocks. Eldritch/Beast enemies average 8.6%. Same party, same skills, and one of those groups takes your bleed build and the other ignores it entirely.
This page is a reading guide to that dataset, with the derived averages shown so you can reproduce them. Everything comes from darkestdungeon-enemies.json, parsed from darkestdungeon.wiki.gg.
The class distribution
| Class tag | Entries |
|---|---|
| Eldritch | 17 |
| Unholy | 13 |
| Human | 13 |
| Bloodsucker | 11 |
| Eldritch/Beast | 11 |
| Beast | 9 |
| Human/Beast | 6 |
| Eldritch/Human | 3 |
| Human/Eldritch | 3 |
| Husk | 3 |
| Cosmic | 2 |
| Everything else (8 single-entry tags) | 8 |
“Everything else” includes the odd ones — Carpentry for the Barrel O’ Bombs, Ironwork for the Brigand Pounder, Crockery for the Hag’s Cauldron, Poor Soul for the Miller, Stonework/Eldritch for the Garden Guardian. Those are object enemies, not creatures, and they behave accordingly: the Cauldron has 100% PROT, the Barrel has 200% resistance to every status channel.
The dual-class tags matter. An enemy tagged Eldritch/Beast can be affected by anything that keys off either, which is why those entries tend to have the softest resistances in the file.
Resistance averages by class
Averaging Champion-tier bleed and blight resistance across every entry with a stat block produces the most useful table on this page:
| Class | Entries sampled | Avg bleed res | Avg blight res | Better channel |
|---|---|---|---|---|
| Unholy | 11 | 156.4% | 45.5% | Blight |
| Beast | 6 | 68.3% | 93.3% | Bleed |
| Human | 8 | 34.4% | 46.3% | Bleed |
| Bloodsucker | 7 | 40.7% | 54.3% | Bleed |
| Eldritch | 15 | 65.0% | 46.3% | Blight |
| Eldritch/Beast | 11 | 8.6% | 22.3% | Either |
| Human/Beast | 5 | 66.0% | 94.0% | Bleed |
Read the Unholy row first, because it is the one that decides party composition for an entire region. 156.4% average bleed resistance is not “resistant”, it is immune with margin. Individual Bone entries run 200% to 245%. The undead counter guide breaks that down enemy by enemy.
The Human/Beast row is the mirror image: 66% bleed against 94% blight means bleed works and blight mostly does not. Swine enemies live here, and it explains why the Warrens rewards a bleed party while the Ruins punishes one.
Eldritch/Beast at 8.6% / 22.3% is the softest group in the file, eleven entries deep. If you are unsure what to bring to an unfamiliar region, this is the class where almost anything works.
Size, and why it changes the fight
| Size | Entries | What it means |
|---|---|---|
| Small / Small (1) | 20 | One rank, dies fast, often acts early |
| Average / Average (1) | 47 | The standard body |
| Large / Large (2) | 18 | Two ranks, usually armoured or high HP |
| Gigantic (3) | 4 | Three ranks |
| Unimaginable (4) | 2 | Fills the enemy formation alone |
Average-sized enemies are 46% of the file — 47 of 99 — which is why most fights feel like a line of interchangeable bodies. Large enemies are where the trouble concentrates: the Ghoul at 57 HP and 40% PROT, the Bone Captain at 68 HP and 33%, the Swinetaur at 88 HP and 40%.
Size also drives targeting. A Large enemy occupies two ranks, so your rank-3 and rank-4 damage can reach it without movement skills, but it also means there is less room for the enemy to hide a caster behind it.
PROT is the real health bar
Raw HP in this file ranges from 6 (a Maggot variant) to 500, but PROT rewrites the ranking. The formula is damage-to-kill = HP ÷ (1 − PROT):
- Bone Defender: 31 HP at 45% PROT = 56 damage
- Bone Captain: 68 HP at 33% PROT = 101 damage
- Pelagic Guardian: 51 HP at 50% PROT = 102 damage
- Uca Major: 82 HP at 50% PROT = 164 damage
- Unclean Giant: 137 HP at 0% PROT = 137 damage
The Uca Major is the clearest case: 82 HP costs more damage to remove than 137 HP, purely because of armour. That is also why damage over time is so strong in this game — blight and bleed skip PROT entirely, so a 164-damage Uca Major is an 82-damage target for a blight party.
The enemy priority guide builds a full kill order out of that table.
Speed separates the cheap threats from the expensive ones
The file’s speed column correlates hard with cost. The 16–27 damage-to-kill enemies move at speed 3 to 10. The 101–164 damage-to-kill enemies are nearly all at speed 2.
“Bone Defenders are enemies commonly encountered within the Ruins. Their large shield leaves them well-armored and capable of taking a beating, and they can use this resilience to protect the other undead soldiers” — darkestdungeon.wiki.gg, Bone Defender stat block
That quote is the design philosophy of the whole bestiary in two clauses: the expensive enemies exist to protect the cheap ones. A party that kills in HP order does exactly what the encounter wants.
Reading a new enemy in four columns
When you meet something you do not recognise, four numbers answer almost every question:
- Size — how many ranks it blocks, and whether your back-rank damage reaches it.
- Speed — whether it acts before or after your party.
- PROT — whether raw damage is worth attempting.
- Bleed and blight resistance — which damage-over-time channel you own actually works.
Dodge is a fifth, and it matters more than it looks: a Crone at 42.5% dodge or a White Cell Stalk at 38.75% will eat a quarter of your actions before resistance even enters the calculation. Accuracy trinkets and multi-hit skills beat single big swings against those.
Where the numbers come from
The dataset is a parsed snapshot of darkestdungeon.wiki.gg stat blocks, cached locally so the site can render the enemy database without hitting the wiki at request time. Bosses live in a separate file of 37 entries — see the boss guide — but several bosses also appear in the enemy file, because the wiki catalogues them in both.
The averages on this page are computed across entries that carry a full resistance block; entries with only partial fields are excluded. That is why the sample sizes in the table are smaller than the class counts.