That’s not how it works at all. The number of games of matchup data is not equal to the simple product of total games × opponent side popularity. I recently wrote a post explaining how it actually works.
Please note that even this is subject to random variations. It’s less true with smaller samples but becomes more accurate with larger ones.
Speaking of…
… would imply that as a sample size grows, subtracting wins from losses would yield a result which also grows multiplicatively. You don’t understand the law.
He said around 1k, unless you asked him off of the forums somehow. I don’t know how that telephoned into above 2k but somehow it did. The climb to 550 was not surprising.