Your last paragraph is why I did using MC simulations based on credible data. Once you pick a return and a volatility, the output just restates your inputs and you can't tell a wrong answer from a right one. Historical data doesn't fix that, but it moves the assumption somewhere you can argue about, and shocks arrive as correlated bundles that actually happened.

Dying with roughly $0 is a real mode in the app: floor and ceiling, or Yale style smoothing, where spending flexes instead of staying fixed.