05 August 2023

Class/Race archetypes in 1.2 million D&DBeyond characters

tl;dr: giant dataset from D&DBeyond shows us that over half of characters created are the same few (30) race/class combos.

dice-scroller.com pulled over a million character sheets of D&DBeyond to give us Most popular D&D Classes and Races and I wanted to revisit some work I did last year on Player choice of unusual races within D&D classes. Caveats about all this at the bottom of the post but short version - this is probably not Big-T-Truth, no dataset can be, but I think it is still worth looking at what shapes and patterns might become visible.

Just eyeballing the work that dice-scroller did we can see the same concentration as before - this is not a gently-fading colour map, we have points of heat speckled across our race/class grid that shows strong preferences for certain combinations. Interestingly for such a large data set we *still* have some race/class combos missing saying that eight years after the game launched not all things have necessarily been tried.

The significantly lower frequency of Artificers and to an extent Bloodhunters says that those being gated behind paywalls on DnDBeyond is having an affect, as could be expected.

Practically what can we do with this? What I read is that there is a strongly perceived match of race to class for some classes; nearly 1/5 Rangers are elves, over 30% of Druids are elves, 1/3 Fighters are human. We see 2/7 of Paladins are human, Warlocks are tieflings, Bloodhunters are human, Wizards are elves. The archetypes jokes are rooted in facts it appears.

Looking at the top 30 choices overall we see 53% of characters fall within that and 21% are the humans alone. This is 30 out of 686 possible race/class combinations with this set of races. This is people sticking to a ~4% subset of the options available which is a pretty remarkable concentration.

If you used the table below to generate a group of NPCs it would be pretty typical of what people are apparently playing at tables. I wrote up a version of this before in Data driven Appendix P: 3xd100 NPC party generation tables. I think the data we see here is sufficiently similar that if you use the tables in that post you'll get close enough results.

Class Race share
Fighter Human 0.04
Ranger Elf 0.03
Rogue Elf 0.03
Wizard Elf 0.02
Rogue Human 0.02
Paladin Human 0.02
Wizard Human 0.02
Cleric Human 0.02
Druid Elf 0.02
Warlock Tiefling 0.02
Cleric Dwarf 0.02
Rogue Halfling 0.02
Barbarian Half-Orc 0.02
Bard Half-Elf 0.02
Ranger Human 0.02
Fighter Dwarf 0.02
Monk Human 0.02
Barbarian Goliath 0.02
Paladin Dragonborn 0.02
Rogue Half-Elf 0.01
Warlock Human 0.01
Bard Human 0.01
Fighter Dragonborn 0.01
Barbarian Human 0.01
Warlock Half-Elf 0.01
Barbarian Dwarf 0.01
Sorcerer Half-Elf 0.01
Wizard Gnome 0.01
Sorcerer Human 0.01
Monk Elf 0.01

The flip side of this being that if you as a DM want to surprise your players you could go with other options, popular but not top tier - this implies a workable build but not an archetypal one.

One of the other things I find interesting is the apparent perception that there are unworkable builds - Kenku Barbarians and Paladins, Halfelf Barbarians Firbolg Rogues and Druids, Lizardfolk Bards - this is where there are other classes much more popular for those races. The Firbolg Druid is surprising to me.

Noting the caveats in the comments on dice-scrollers post and adding that all this has probably been rendered meaningless by Tashas. I think this is probably more of a window into a past set of behaviours (pre-Tasha). No data set is perfect, this one has flaws but given the sheer volume of data within it we can poke at these big trends with some confidence. Nothing above should be taken as a view on what is best or how one should play, this entire piece is a commentary on what we might be able to see in this data about players perceptions of the optimum race/class combinations, whether they are right or wrong we cannot tell from this. We do not know for sure why any of these decisions were made.

All the following posts on this topic come off surveys mostly so take them with as much salt as you deem appropriate.

Player choice of unusual races within D&D classes - April 2022

Revisiting D&D race preference by source - June 2021

Changing popularity of PC races with age & experience - April 2021

Comparing range of race choices across surveys of D&D players - October 2020

Sources can be found in the Big list of TTRPG/D&D surveys


  1. Interesting that paladin is still a human class. It's been 23 years since paladins were restricted to only humans, but it had apparently shaped the archetype enough so that people don't really imagine a paladin as anything else.

    1. Paladins seem to be a popular choice if you go dragonborn - or at least, the most popular non-human paladins are dragonborn

    2. Pre-Tasha's, if you wanted to play dragonborn, paladins were one of the only combinations that had obvious mechanical synergy.

  2. most common dwarf class: cleric?

    1. Just a bit above dwarf fighters but apparently yes.

    2. Hill dwarf life cleric was considered one of the best combos you can play in phb.

  3. Dwarf cleric and firbolg druid are both notable class combos in a popular DnD actual play podcast (The Adventure Zone)

    1. Good point - I am surprised Firbolg Druid is not more popular.

  4. My criticism is that the "top 30"-based analysis really assumes all of the races and classes are legitimate in every game, which isn't true. (And is far less true of races than classes.)

    I'd like to see this analysis redone for PHB-only races and classes. No artificer or bloodhunter, and only PHB classes. Maybe pick the top 8 combinations. Do we still see a massively unequal trend? Looks like yes, and this would be more meaningful.

    1. Can you elaborate on what you are thinking in those first lines? I am not quite following what we're looking for. The Top 8 of the 'PHB only' selection is the same as the the top 8 of that top 30 list.

      I think what we have got here won't change much:
      - PHB only races are 83% of the whole set
      - PHB classes are 99% of the whole set

      As best I can tell taking "PHB only" knocks Genasi out - they are more popular than Gnomes - and pushes Goliath Barbarian out of the top 30 letting Tiefling Sorceror slip in.

      I think it will still look broadly similar - heavily weighted to humans and elves - but let me put some stats on that.

    2. To duplicate the concept on 'PHB only' - you get 50% of player choices included in the top 21 class/race combos; that would be 21 of 108 potential options ~ half of all players going for ~ 19% of the available options.

      Top 8, or 7% of available options for 'PHB only', would cover 25% of players.

    3. I don't think it's PHB only as much as the PHB is what most people have when they start, and so it probably is owned by the most people. And thus that's what most of the variables come from, there's only 2 classes outside of PHB and the most famous races are all in it. I myself didn't collect books for around 7 years after starting, so it's logical to assume other's didn't either

  5. Love that used my data for this!

    One small thing: This is only a subset of all characters on dndbeyond so saying "Interestingly for such a large data set we *still* have some race/class combos missing saying that eight years after the game launched not all things have necessarily been tried." might be incorrect.

    I really want to improve on what I did here when I find the time. Maybe I can even get all the characters if possible.