That’s very interesting!
If we restrict to games that people ‘care about’, we can get rid of a lot of automatically generated games. One measure of ‘caring about’ is having at least 5 ratings. Then the numbers are:
2005-2009:
281
2010-2014:
644
2015-2019:
690
Having at least 1 review might be a different measure. But I’ve written about 1/4 of all reviews, so to remove my personal interests from the equation, I’ll do 2 reviews:
2005-2009:
225
2010-2014:
495
2015-2019:
557
And again, we’d expect all these numbers to go down a bit over time if growth was stagnant, since older games should be accumulating more ratings and reviews.