I have come to believe that all systems which attempt to choose assets based on past price performance are market timing systems.
As against that I still believe a small modicum of “momentum” chasing is acceptable especially when combined with risk parity weighting and a large portfolio equally balanced between high and low risk assets. See: Trend Following Without Terror
On top of that I run scared from any re-balancing system whose results vary too greatly depending on which day of the month one re-allocates. And any sort of on/off binary choice of asset is a dangerous proposition where the re-allocation dates can affect performance dramatically.
Some time ago I did extensive research and back testing on a monthly re-balancing system based on investing in the top “x” US stocks, being careful to include randomly chosen de-listed stocks in the mix.
I read with interest AQR’s momentum prospectus and noted the small pick up in returns over a conventional approach. Cliff Asness has it right: he rotates into the top “x” stocks out of 1000 and “x” is a healthy 300. Fine, drawdown and volatility will likely be no better than the Russell 1,000 or whatever but profit may well be a little increased. And it will be resilient to changes in the rebalancing dates.
All of this lead me to realise that momentum is a fine thing but that concentrated on too small a figure for “x” it is an unstable disaster. Investing in the top few hundred is fine and reasonably resistant to changing the dates of the month on which a monthly reallocation occurs. A small number for “x” such as 10 or 50 is an unstable disaster and leads to wide and wild swings in performance metrics depending on which day of the month one picks.
So any re-allocation method which chooses on or off for a small number of stocks does NOT get my vote. While in back testing it may look good to “go all on or all off stocks / bonds” using TLT and SPY as representative classes, in the long run it will slap you round the back of the head with a lead cosh.
Great danger lies in trying to time the market with a limited number of instruments in an attempt to increase profits and reduce volatility and drawdown. It looks tempting in back testing but will almost certainly leave a sour taste in the mouth in the long term.
These views have been reinforced over the past couple of days by my re-creation and back testing of the Logical Invest Universal Investment Strategy
Their approach is a good attempt to avoid binary “risk on / risk off” systems:
The idea for this Universal Investment Strategy was to develop a strategy which has an adaptive allocation between 0% and 100% for each ETF (TLT / SPY) depending on the market situation.
Unfortunately it still leaves many periods where risk is entirely “on or off”. If you can count investment in a 20 year bond as “risk off”!
Worse, testing different days of the month re-balancing reveals some stress which can add to the instability of a system which is at least in part binary.
A system such as risk parity avoids this problem and is highly resilient and impervious to “day of the month” problems since it is never wholly “in” or wholly “out”. Prediction of “volatility” is a great deal more stable than prediction of “return”.
Here is what I tested. Different “end of weeks” on which to reallocate (13 rolls in a year):
self.week_list1 = 1, 5, 9, 13, 17, 21, 25, 29, 33, 37, 41, 45, 49 self.week_list2 = 2, 6, 10, 14, 18, 22, 26, 30, 34, 38, 42, 46, 50 self.week_list3 = 3, 7, 11, 15, 19, 23, 27, 31, 35, 39, 43, 47, 51 self.week_list4 = 4, 8, 12, 16, 20, 24, 28, 32, 36, 40, 44, 48, 52
And here are the performance metrics for the Logical Invest UIS and a simple inverse volatility weighting strategy using Spy/TLT since 2002 to end 2016 (using a three month lookback period to calculate volatility):
|Settings||CAGR||Daily Sharpe||Max DD|
|Logical Invest UIS|
Good effort Logical Invest and certainly an improvement on “Dual Momentum” but for my choice I will stick to risk parity.