PVI is a flawed metric that assumes the two major parties natural state is equal in the presidential popular vote. This is not true. A non adjusted average spread over the past couple elections is a far better metric.
PVI is a good metric because it exists in a vacuum, independent of the cyclical nature of politics. If you took an average after 2016 for the previous 4 cycles, then the Republicans seem like they are perhaps the majority party. If you took it after 2012, then the Democrats are. The Democrats are the larger party, but if you average the last 20 years, including midterms, they only get about 1% more than Republicans.
PVI doesn't just exist in Presidential years. If 2022 GCB is R+4 or something like that, being like "Democrats have won WI by 2 on average over the past 3 Presidential Elections" isn't helpful when Trump barely lost in a D+4 year.