Player Class Stability from 1M D&D Beyond Sheets

Regular readers will know I do love a good D&D Dataset and Dice-Scroller gives us a good one! Through some Python wizardry they scrapped ~1.2 *million* character sheets off D&DBeyond and made them available for our wonder and delight!

You can find the dataset on Kaggle and see their own neat little dashboard and analysis here. What I want to do here is take this as the new baseline for our understanding of the 'general gamer' and see what it tells us.

The 'general gamer' is a pattern I have been finding in lots of datasets - where they are from forums, twitter, anything that implies an engagement with the hobby away from the playing table we see a pattern of favouring spell-casting classes. Where we get the data from tools that help people at the table we see a different pattern with a heavier favour on fighters and rogues. I want to update the comparisons with this new giant dataset - starting gnarly and clarifying from there.

Start with our 'everything in' star-plot - we see the general/forum gamer split; oranges are the app-derived datasets, blues are the surveys and polls.

Distill that down and we have a new 'general gamer' baseline from this monstrous dataset; our story remains the same (Fighters, Rogues beloved by those people who are appear in these giant datasets but not in the forums)

Ranking the delta between these two groups preferences shows this clearly.

What is especially cool is that we now have a good set of general gamer sets over time that allow us to see the stability of class preference - the general gamer has not seen their preferences change with the release of books like Tasha's which updated the classes.

Compare this with the choppiness and changeability in forum respondents where we see more variability in favourite character classes. [Edit: for the avoidance of doubt; I am saying this data is noisy, as we would expect from small surveys drawn from lots of different places. This contrasts with the stability of the larger surveys.]

We poked at this before (see Differences in 5e class preferences by source, Ranking 5e classes by preference within player populations, Updated class preferences, incl. Twitter, Facebook and Player Stats for AideDD Character Builder 2020-2021) but the sheer scale of this dataset gives us decent confidence we have a solid window on what is going on.

Putting on my 'bringing more folk to the game' hat - I think the takeaway from this is to anticipate that people who walk up to the table will probably be more happy to play a martial class more than you might anticipate.