I don't want to run out of money either and think we need to be careful about the math behind these analyses. When we use historical data in statistics, everyone recognizes that it doesn't predict the future, but I wonder if people some other issues.
For instance, it overemphasizes some years and underemphasizes others. Each starting year in the data gets one trace, but statistically that's not fair, the end points are under-represented. The first year of data only occurs in one trace, the second year can appear as year 2 in the first dataset and year 1 in the second. By the time you are 30 years later, a given year can appear in 30 different (30 year long) data sets. At the other end, the same thing happens in reverse, data can be used less and less frequently and 2020 data can only appear in one trace. So for 30 year datasets in the 150 years of Shiller's data, the first and last 29 years are represented less often than the middle 92. Personally my approach is to repeat the 1871 and later data at the end of the series, so that all time frames appear equally, though obviously that creates a new problem that 1871 and later results will be different from 2021 and later.
I also wonder whether the oldest data have much significance. It is hard to grasp whether 1871 (Shiller's first year of data) has a bearing on today when Civil War reconstruction was ongoing, the transcontinental railroad was just completed and electric lights and the automobile were not around.
But even if perfectly applicable, 150 years of data is only five fully independent 30 year datasets. So even if the behavior statistics do not change over time, the odds that those few data sets pick up the full range of extremes is very low. Evaluating longer retirements creates an even larger end point problem and has even fewer fully independent datasets.
Most people intuitively understand that they can't fully trust historical results to represent the future so they try to be conservative on the saved amount and try to cut back spending if the market goes down a bunch. That's about all we can do.