Benchmarking random NPC generation (3.5e DMG vs 5e actual data)

I set off to answer 3 questions: - how to the populations of characters running about map to what the DMG would generate - is there variation between player preferences in different (real world) geographies - are there observable changes over time This is inspired by the large datasets released from DnD Beyond - this shows us the make of the PC population for 5e, broadly speaking. Correct me if I am wrong but the last 'fractions of NPCs in the gameworld' rules that I can spot are the 3.5e DMG so we shall compare to those.


Taking the first - how does what players (blue) do compare to what the 3.5e DMG (grey) tells us is out there? 

- DMG thinks we are playing lots more rogues, clerics and wizards, much less paladins 

- Lumping warlock and wizard accounts for that difference between wizards in both cases (3.5e had no warlocks, both similar magic users) 

- DMG seems to think all the paladins we are playing should be clerics

 - Barbarians, rangers, fighters, druids are broadly the same 

- All the rogues the DMG thinks should exist are actually running monks, bards and sorcerors 

Of course this compares 5e players to 3.5e DMG; I hope to check this with 3.5e data to see if it holds up. 

The orange attempts an estimate from a small (N = ~50) poll in 2003 which indicates there could be a more interesting story: that not all the warlock players are wizards by another name; the theory that they are an arcane flavour of cleric e.g. drawing powers from a greater being could have something to it. More data could refute or affirm this.