I can't speak for the metaverse proper, but this is the breakdown that was captured on the AltMeta during its nine years of operation:

(Edit: just noticed DMF necro'd...)

Here is all the game data from 2006 to 2015 (when the altmeta was shut down)-

http://altmeta.kryofx.com/altmeta.csv

Every single game on there is a 'real' game that was posted to the metaverse, which does not permit cheat codes or non-cosmetic mods. The data will be by definition skewed because it doesn't include non-metaverse games, which are the only way to play with functional mods and are probably the greater majority of games played. There are no records of non-metaverse games for GC2 (GC3 has better stats since it can track all games played).

Bear in mind also that the data spans all the way back to 1.0; there were strategies at various points that permitted very fast victories (spore rush, etc.) that were subsequently nerfed. At its peak, there was a big focus on getting the fastest military win possible on the biggest, highest difficulty map because that gave the highest score.

Even on the recent engines?

(I think you described something like this in a series of posts early this year, but IIRC you were using DL.)