This is very interesting - do you mind sharing your MC code? I have been playing around with your earlier analysis and it has been very helpful.millennialmillions wrote: ↑Sat Aug 14, 2021 11:51 am I ran the same strategies as above through bootstrap-sampled Monte Carlo simulations. The fund performance data includes 1955-2020. I assume each year is independent but that the returns across various funds within a year are dependent. E.g. a year in the sample may correspond to historical year 1980, in which case returns for all funds reflect returns from 1980. I ran 10,000 simulations of 35-year periods for each strategy.
Unsurprisingly, there is more spread in the percentiles than the straight historical analysis. The only meaningful difference I see is that the 10th percentile is much closer across strategies.
Looking at both the historical period analysis and the bootstrap analysis, the constant 2x 70-30 portfolio seems like the winner to me. The 10th percentile is close to the best result, and the median result is significantly higher than anything except a constant HFEA.
The biggest thing I'm grappling with is Rob's point about bond performance being inherently different in the future. However, even looking at periods that begin in the mid-80's, the performance of the 2x 70-30 portfolio is significantly better than strategies that do not hold bolds while leveraged. I understand it's not reasonable to assume that high of inflation and bond yields going forward, but isn't it reasonable to assume the risk premium of equities over bonds will behave similarly to the past and the strategy that performed best across history is most likely to perform best going forward?