This is an extremely important point!

As with many things in this series, I did not pay attention to it since it was not important for showing the issues of this approach.

For example, if it even fails with survivorship-biased data, it probably won't work better for unbiased data.