Very interesting and I appreciate your comments and threads; but it is overwhelming as long as your approaches and all the different approaches that I mentioned, including the valuation based approaches that I think motivated this thread, are not reconciled in some way.I agree price movement is the key. I personally do nothing with valuation based investing, that's a whole world that I have no knowledge of.
"This is like trend following in some ways, except that future volatility is more reliably predicted than future price movements." - Price movement is all that matters, however; of course adjusted for the possibility of drawdowns. We can "eat" i.e consume the price i.e. the terminal account NAV as the end result, but not the volatility or lack of it.
So it all comes down to predicting risk and return, in particular geometric rates of return which are adjusted for drawdown risk, and the question is to what extend can volatility do that. I think you demonstrated your excess returns in your other thread; I have a feeling that your approach is very similar to a trend following overlay, if one were to do a factor decomposition of returns with trend as one of the factors. Trend following can be backtested on testfol.io with symbol KMLMX.
If both "trend" and "value" or valuation based investing result in excess returns, then the asset allocation problem becomes a question of how to combine "value" with "momentum", which some factor ETFs try to do - on a security basis at least, but similar to what one could do with asset classes. Here I view "trend" as a particular use case of "momentum".
The volatility part is risk management. If volatility picks up, the chance of a drop increases, take profits. Volatility drops, the chance of a climb goes up, take more risk. There doesn't need to be an adjustment based on expected return.
I think of trend/momentum as something additional. The classic dual momentum approach considers both absolute momentum (is the price trending up) and relative/cross-sectional momentum (which assets are doing relatively better). If I understand correctly, the momentum factor is basically cross-sectional momentum with the top third minus the bottom third. Classic trend following is more extreme, it tries to identify and take advantage of the extreme events (the 5% tails) where assets are making large prolonged movements up or down. I don't do that - that's the Turtle traders and their ilk, but I think KMLM and similar funds have low-vol versions of that trend approach.
I don't really predict return per se, although I suppose that one might argue that it underlies my risk budget approaches.
I've been exploring both a risk budget approach (just a fixed risk budget and volatility determines allocations) and a momentum-based risk budget approach in my other threads.
The pure risk budget approach tends to give much smoother returns when it works. All that it does is have a predefined fraction of the portfolio variance (or volatility) assigned to each asset (the risk budget), using recent volatility. Backtesting suggests that just this volatility scaling would have worked 1980s through 2022, when both stocks and bonds were simultaneously in a long-term bull. As a rule of thumb, I let equities take 4 times the risk of ballast. I think we may be more back to the 1960s/1970s now, implying something other than long bonds would be better for the ballast.
The momentum-based approach starts with a universe of assets to select from, calculates momentum for each, weights the risk budget by momentum, calculates the risk budget weights, and winnows down the smallest allocations until around 40% remain.
The momentum-based one tends to give higher CAGR and backtesting suggests that it would have navigated 1960s to present pretty consistently using nothing but 3x LETFs plus -3x bonds, but with peaky fits and starts that give frequent spike/drawdown pairs. Almost every asset tends to have higher returns than its underlying when accounting for the holding time of each asset (for example, the equivalent CAGR for an asset held 40% of the time is larger while it's held than the same asset bought and held throughout).
Statistics: Posted by comeinvest — Fri Nov 22, 2024 2:49 am