So, being a bit of a spreadsheet nerd, I have a results tracker of my battles. In 2022, I added win differential to the data points I am tracking, and thought I'd share it here. This is manually tracked, and I only track battles tier 5+, solo, and no stock modules, but still a sample size of 2951, so it should give a pretty good look.
The ultra close games (0-2) were more common than the total steamrolls (13-15), but the average trend does show the snowball effect pretty dramatically, since the most common 3 results were pretty large differentials (8-10). I had thought going in the middle of the pack would be more like 7-9.
Either way, I do think the "every game is a steamroll" that is bandied about quite regularly is not quite as accurate as people might think.
Source: https://www.reddit.com/r/WorldofTanks/comments/1014u64/win_differential_2022/