When we see "regression" in fantasy, it's usually about a player over or underperforming in a stat relative to the league-wide sample (usually over multiple seasons). When a player is way above or below the average of that sample, he's a positive or negative regression candidate.

It does not mean that every player regresses to the mean of the sample. We have very good reasons to believe some players' true expectation over a big sample would be above or below the league-wide average. For example, we very much should believe Mahomes' TD% is higher than avg.

Where people get confused with fantasy football regression is when a very good or very bad player is a "regression candidate". That doesn't mean they'll regress to the average, but it does mean they're projected to regress *closer* to the mean (ex: Mahomes TD% from 2018 to 2019).

Lets look at Amari Cooper, who scored 8 TDs when my model expected him to score 4 based on his usage.

Amari is good and he plays on a good team. His true baseline is above average, so we should expect him to regress to about 6 or 7 TDs instead of 4. It can be that simple.