> Firecalc’s Monte Carlo gives me zero failure cycles. Vanguard puts me at 98% success. It is worth keeping in mind how "failure" and "success" are defined in

> Firecalc’s Monte Carlo gives me zero failure cycles. Vanguard puts me at 98% success. It is worth keeping in mind how "failure" and "success" are defined in the context of these models. Deferring retirement while continuing to work, but then dying suddenly, would be regarded as a "success" - since you died before running out of money. With a broader perspective of what we're trying to achieve (perhaps something like: "make the best use out of our life", but this can be highly personal), that kind of scenario sounds like a failure to me. Similarly, the estimates of those retirement models assume that the past is roughly a representative sample from some distribution that the future will also be sampled from. But there's no guarantee that will be the case. There's a nonzero chance that some largely or completely uninsurable major disruptive event will wipe out some or all of your retirement capital. Looking back over the last hundred or so years: https://voxeu.org/article/inequality-total-war-great-leveller https://voxeu.org/article/inequality-total-war-great-levelle... What am I trying to argue? There are many more failure modes other than running out of money before you run out of life. Focusing on reducing risk of one failure mode while not thinking about other failure modes leads to sub optimal decisions.