I said already. But let’s break down in excruciating detail what “weighted average by the opposing deck’s general popularity” means.
You said Quest Paladin, so let’s say I’m calculating Quest Paladin, T1KL.
First opposing deck: Face Hunter.
Matchup winrate is estimated at 37.4%, because there isn’t published data for T1KL so I use the general Legend matchup in lieu of better data.
That 12.6% from 50%, so that’s the polarization for that match.
Now that 12.6% is put into a weighted average where the weight is not the number of Quest Paladin vs Face Hunter games recorded at T1KL — indeed, HSReplay doesn’t release those numbers if they’re below a certain threshold — but instead by Face Hunter’s general popularity in T1KL. In this case, the weight is 16000 games, because a lot of Face Hunter in T1KL. So in a column on the spreadsheet, 0.126×16000, and in another column 16000. Repeat for all other opposing archetypes (unless no matchup data whatsoever). Sum the former column and divide by sum of latter column.