Benchmarking random NPC generation (2e DMG vs 5e actual data)

In a deep, deep cut to the first days of this blog - how do the NPC groups generated by the DMG compare to what see at actual tables of players?

My first number-crunching post looked at 5e stats and a group generated by the 3.5e DMG in "Benchmarking random NPC generation (3.5e DMG vs 5e actual data)" - and I just now stumbled upon the AD&D 2e equivalent in the back of the Monstrous Manual (Appendix III:NPCs).

Reading through this we already see a *massive* difference to contemporary parties in that an NPC party should be 2-12 members, with 2-5 leveled classes and the rest henchmen and men-at-arms. We will come back to this but to start, the key comparison we want is the classes. These are massively different to anything we've seen since - assumption is that 40% of an NPC party are fighters, 20% wizards, 20% clerics, 10% Thiefs (Rogues) and 10% for all the rest.

Testing this against the roster for my drop-in game characters back in the mid-90s, surprisingly it seems to hold up.

Breaking out this roster plus another cleric that played with us and we get:
50% Fighters
21% Thieves
14% Clerics
14% Wizards

A touch higher on the thieves and Fighters than expected but still Fighters in the majority by a good margin.

This is compared to the rules about 'fractions of NPCs in the gameworld' from the 3.5e DMG which we previously tested against a 2003 survey and found *broadly* representative. We also compare to the 5e D&DBeyond figures to get a sense of what is actually played at table today since there are no NPC generation rules in the 5e DMG.

Hugely different to what we see today with its must more even spread among classes and more classes to boot. There were no Monks, Warlocks or Barbarians as classes at the time. Specialist Wizards were a thing which I am comparing to Sorcerers (which did not exist as a class) but they too were a tiny proportion. The major assumption was that a party was 'swords first' through fighters, where as I would expect a modern 5e front line to be Fighter/Cleric/Paladin/Monk/Barbarian. If we take this, then the total fraction of 'front-liners' is similar - Fighter + Paladin was 40% in 2e compared to Fighter/Cleric/Paladin/Monk/Barbarian being 35% from the 5e DnDBeyond figures.

Share of Clerics and Wizards in 2e were roughly the same as in 3.5e (DMG and survey) at 15-16% each. Both of those drop to ~8% in 5e as Warlocks and Sorcerers take their share of Arcane casting and Druids get a share of Divine casting while Paladins become accessible to all when the Prime Requisite requirements are dropped.

The big step was the 2e:3e release from class restrictions due to whatever you rolled as your stats. Once people could choose what they liked you got a much broader spread of classes. The 3e:5e change shows much closer similarity. Altogether, if you squint, it sort of looks like a broadly similar long-haul trend for preferences between sword-swingers, spell-slingers and sneaky types that just get chopped up differently between the classes available in the different editions - "unavailable" here including both the class not existing or it being locked behind improbable stats rolls.

The practical outcome those was that we really were playing a different fighter-led game back then and it did look like those random roll Appendix III NPC's.