[johnnyanmac]

>Then it seems that "survivorship bias" is some kind of unfalsifiable and self-fulfilling belief.

It's very falsifiable. Just not by us, as we have no acccess to the sales data, nor enough public sales points to make a proper statistical analysis. The best ones out there are either based on estimations (especially reviews to sales ratios) or non-public data you pay the NPD or someone similar thousands to access (and obviously you're not allowed to share that data). So someone truly curious can pay a lot of money to get an answer.