What is your strategy
My strategy follows HFEA as I am a young (mid 20's) investor willing to take on increased risk. But I do not want to hold just 100% stocks or 100% leveraged stocks as that increases risks substantially. I also want to be covered in any sort of economic environment, from rising interest rates, inflation, deflation, etc. To do all of this, I set out to complete the following tasks:
- Gather monthly data for Stocks (small, mid, large cap), Treasuries, Gold, Commodities, REITS, etc. Any asset that could be added to a portfolio.
- Estimate leveraged returns for Stocks, Bonds and Gold monthly. See post: Siamonds Simulating Returns of Leveraged ETFs
- Use Portfolio visualizer to calculate the efficient frontier and geometric efficient frontier of a asset mix using the above assets both leveraged and unleveraged.
- Optimize asset allocation to reduce long draw downs. Sharpe Ratio, Sortino ratio, ulcer perf. ratio, etc.
Idealy, I wanted to gather monthly data for as many asset classes back to 1955 as I could. This time was spent scouring this forum, https://fred.stlouisfed.org/, Tyler9000's website Portfolio Charts, and others. I would then backtest what monthly data I collected with an ETF of similar allocation. The data needed to be accurate before leveraging it.
2. Estimate leveraged returns
Next step was estimating what a leveraged return for the specific asset would be. This is where Siamonds Simulating Returns of Leveraged ETFs post was extremely helpful. Using some of these equations, I simulated leveraged (2x & 3x) monthly returns for S&P500, small-cap, mid-cap, LTT's, etc. till 1955. Also, simulated gold back to 1972. Here is the 3x backtest's for S&P500 and LTT compared to relevant ETF's.
S&P500 3x (portfolio 1 is simulated):
LTT 3x (portfolio 1 is simulated):
Here you can see that the simulated leverage LTT's does break from relevant ETF (TMF). Further research is to be done to increase accuracy.
3. Efficient Frontier and mean-variance optimization
Portfolio visualizers historical efficient frontier calculator was then use to plot the difference between an efficient frontier of just 1x assets and an efficient frontier also including 2x/3x assets. Also plotted below is the Geometric Efficient frontier (Kelly Criterion). Both plots only consist of Stocks and bonds, no gold or REITs are used as backtest for gold only went back to 1969 (will be used in section 4.). The "Provided Portfolio" in this plot is Hedgefundies 55UPRO/45TMF Allocation.
Efficient Frontier consisting of Stocks and bonds (1955-2020): Red is 1x Leveraged, Blue is 1x,2x,3x leveraged
Geometric Efficient Frontier consisting of Stocks and bonds (1955-2020): Red is 1x Leveraged, Blue is 1x,2x,3x leveraged
Now I can set my asset allocation to either the max sharpe ratio or anywhere along the curve. So depending on my risk tolerance (standard deviation) or expected return needed, I can select a portfolio that provides that. This is where each individual can select a portfolio allocation based on there specific standard deviation risk tolerance. For the purpose of this post and my own allocation, I have selected to use a portfolio that gives the same return as Hedgfundies strategy during this time frame (17.2% return, 28.2% SD) but with a lower SD. This gives an expected return of 17.2%, SD of 24.6%, and an asset allocation of:
Mando's Portfolio v1[MPv1]: 3x ITT 7-10 yrs - 36%, 3x LTT - 6%, 1x Mid Cap - 18%, 3x Small Cap - 31%, 3x S&P 500 - 9%
Here is the portfolio returns from 1955 till 2020. Portfolio 1 is Mandos Portfolio, Portfolio 2 is HFEA [ 55UPRO/45TMF], Portfolio 3 is 60/40 stocks/bonds
HFEA strategy and this are about equal until 1967 where they split. Around 1975, the basic 60/40 strategy comes close to providing a better return in the short term. Overall, 60/40 unleveraged beats out the other two portfolio's in sharpe ratio during this time frame. Now what really scares me is the 67% and 72% max drawdowns for portfolio 1 (Mando's) and portfolio 2 (HFEA) respectfully.
4. Minimize drawdown
Now this section is optional if you only care about mean-variance optimization. I personally wanted to see if its possible to reduce the max drawdowns without hurting portfolio CAGR by a substantial amount. I stumbled across an interesting paper (PDF) months ago that describes the downsides of only using sharpe ratio as a portfolio metric (See Pg. 9 and 10). It then goes into detail about using the ulcer index and the downsides of just using that(See figure 4 and pg. 14). They then go on to create there own ratio call "Serenity Ratio" and "Pitfall indicator".
As PV does not provide these ratios when simulating returns, I created an excel sheet that would take the monthly returns data from PV and calculate the Serenity Ratio, Ulcer Index, and Pitfall Indicator. I decided to add in bits of gold (1x,2x,3x) and REIT's and by trial and error see if I could reduce the max drawdowns. I also used the solver function in excel coupled with Simba's backtesting spreadsheet with LETF's to help find an allocation of leveraged gold and unleveraged REIT's that would help provide this drawdown protection. This backtest could only go back to 1969 if including gold and 1972 if including REIT's. I would take an equal amount of allocation out of the stocks and bonds and transfer it to leveraged gold or REIT portion of the portfolio. The final result is...
Final Portfolio Allocation - Mando's Portfolio v2[MPv2]: 3x ITT 7-10 yrs - 32%, 3x LTT - 5%, 1x Mid Cap - 16%, 3x Small Cap - 28%, 3x S&P 500 - 8%, 2x Gold - 11%
Rebalanced Annually / Tax Advantaged Roth
Final Backtest (1972-2020): Portfolio 1 - Mando's Portfolio v2, Portfolio 2 - HFEA 55UPRO/45TMF, Portfolio 3 - 60/40 stocks/bonds
Another Backtest (1972-2020): Portfolio 1 - Mando's Portfolio v2, Portfolio 2 - Mando's Portfolio v1, Portfolio 3 - 100% Stocks (S&P 500)
Table of Data/Ratios
The final Mando's portfolio provides better downside protection, Sharpe ratio and Serenity Ratio, compared to HFEA (55UPRO/45TMF) at the same CAGR! It provides a lower Sharpe Ratio than just a 60/40 stock/bond allocation but much higher returns.
This portfolio will be useful to anyone that wants to try and capture some of the increased gains that HFEA's allocation does but with better downside protection. I have been allocating a portion of my portfolio to HFEA for the past couple years and plan on continuing that. I will be setting aside an allocation of 5-10%(or more) to this portfolio strategy. I am currently still in the process of optimizing the portfolio like shown in section 4 through my own excel sheets. For now, I have settled on the asset allocation given above.
This portfolio ended up being very close to what I originally hypothesized. That it would end up close to Ray Dalio's All-Weather Portfolio of 40% LTT's, 15% ITT's, 30% Stocks/SP500, 7.5% Gold, 7.5% commodities but leveraged and with a higher allocation to stocks. In my case, I did not have commodities to include so gold ended up becoming 11% of the portfolio.
This will most likely result in a part 2 where I finalize the portfolio allocation. I wanted to open up this thread as a means of getting feedback before I travel down the rabbit hole on this strategy even more!
What I would like to do next
- See if there is any way to model in a Volatility fund(XVZ) or any other volatility hedge
- Add in commodities, international, or utilities as another hedge. The data used is a very US centric approach and during a time period of increased returns of the US market. I would like to allocate a portion to international to hedge against a "lost decade" scenario like Japan in the 90's.
- Discuss and research the issues with using a 3x leveraged Small Cap fund
- Get better accuracy on some of the leveraged returns (3x LTT's) and get additional backtest data until 1955
- Research using options and futures to leverage this strategy instead of 2x and 3x leveraged ETF's
- Use Solver in excel in conjunction with Simba's backtesting spreadsheet including leveraged ETF's to optimize a portfolios Sharpe ratios, Serenity ratios, etc.