They state that the reason complexity and ratings correlate is due to sampling bias, and specifically that: “Complex board games disproportionately appeal to the BGG user base”. They then set out to correct for that correlation, and finally declare it to be a much more reasonable and approachable list than the original one with “inherent bias towards complexity”.
I don’t really know how else to interpret this than that the author thought that the BGG user base had a bias towards complexity (vs the rest of the population) and tried to eliminate that bias.
But OK, let’s say that this analysis was not done with that purpose, but just for the heck of it. What are these results then supposed to mean, how are we supposed to interpret them? I honestly don’t know.
Ratings are statements of preference, and as such each individual rating is correct almost by definition. If an aggregate of ratings is incorrect, it has to be a problem with the sample. But this is not supposed to be an argument about populations or samples, so… ¯\_(ツ)_/¯