My plan in theory was as follows:
1. Assume the ~4% rule was Truth, but pick whatever slightly lower rate would result in 100% historical success.
2. Every year, either follow the 4% rule on your previous withdrawal rate by increasing your withdrawal for inflation, or retire again and take 4% of your new balance. In the latter case, you could conceivably also reduce the duration of your plan by 1 year.
This approach is sometimes referred to as the retire again and again model or, if you dig back to ancient history, the payout period reset model.
If you pick a rate that is not 100% historically safe and follow this model, then I do think it increases your risk over the standard model, because you're effectively playing Russian roulette with a firearm with 20 cylinders and 1 bullet (5% chance of failure every tie) and you're playing it repeatedly.
If you pick a 100% historically safe rate and believe that past is perfect prologue (i.e., Truth) or at least the future will be no worse than the past, then you can follow the model and in theory be perfectly safe.
...
Another theory is to look at the balances over the stereotypical 30 year periods for each starting year and try to find a point in time where most of the lines that end up as successes have diverged upward from most of the lines that end up as failures diverging downward. I played around with this once; I don't recall exactly but I think after 10 years you can pretty much tell which way things are going. I think there is some research along this line out there as well which quantifies the risks.
...
Ultimately there really is no 100% guarantee of either life or finances in the future, so I think the real answer is you look at the data, place your bets, and take your chances. Two factors that seem to weigh heavily for many people is if they have children or heirs that would inherit their excess, and how they feel about running out of money in their old age.