I still contend that [Net assets in withdrawal year]/[years of life expectancy in withdrawal year] is better than all these other complicated schemes.
One refinement I would make is to say the above is the absolute limit on what you may spend, but there should be a second softer barrier, the amount you need to live reasonably comfortably. In any year where the hard barrier is above the soft, it then becomes a matter to be judged at the time whether to let a surplus build up as a safety reserve or spend the some of the difference on one-off items.
Adopting the refinement helps iron out one weakness, that at very old ages (mid-nineties) income tended to drop below the long term average.
An additional refinement that also works to iron out that problem is to introduce the effect of putting the assets inside a fair annuity. By "fair annuity" I mean a theoretical one where the provider makes no profit and charges nothing for administration, and at the end of each year each suriving annuitant gets a mortality bonus consisting of his probability of dying in that year multiplied by his average daily investment account balance. According to my data, income from mortality bonuses on their own will exceed the hard withdrawal limit from age 89 for a man, meaning effectively no capital run-down thereafter. (Surely their must be enough retired lawyers and computer programmers on this board to set up an on-line mutual company to provide this idealised annuity?)
The last refinement, which I really don't like because it is ad hoc and inelegant, but which does seem to work well, is to reduce the denominator (life expectancy) by a fixed percentage in calculating the hard barrier. The percentage is found by using Excel Solver to minimise variation in income over the years. This has the effect of moving some of the income from later years to earlier years, which mitigates the low incomes this strategy can produce in the very early years, if the retiree is very young.