Creating your own Strategy to improve an existing strategy

[Mark Vincent]

One of the things I am constantly looking for is how to improve the existing strategies if that is possible. I’m hoping other members will give their thoughts on how to create a strategy that improves your existing strategies. What tools do you use other than LI? Are you betting inflation will come in the future and want to create an inflation strategy? We are lucky that LI has done all the hard work and we don’t have to spend hours figuring out how to create new strategies. LI gives a 3 main categories of assets that are uncorrelated and a combined Hedge. There are other assets like tips and currency but the four main ones are below:
1. Equity
2. Bond
3. GLD
4. Hedge

What if your crystal ball tells you that when inflation happens hard assets like commodities and precious metals will rise in price. You might want to create a Hedge strategy to lower volatility that uses precious metals or metals. Here is a process to create the strategy but I am sure there are better ways and lot’s of questions:

1. Select a precious metals ETF. I will start with PALL, GLD, SLV
a. A big question is will PALL rise when inflation hits? It is mostly used in gasoline catalytic converters. Therefore it might not.
b. Am I curve fitting by adding PALL since it has done so well over the last couple years?
c. Should you add more assets like URA?
d. How do you choose what assets are best for a strategy?
2. Look at the correlations between them and against Equities and Bonds using the following:
a. https://www.portfoliovisualizer.com/asset-correlations
3. The correlations only point you in the right direction. GLD, TLT are the most uncorrelated to SPY but PALL:SPY = 0.37 and SLV:SPY = 0.38.
4. Create the Strategy “Precious Metals” And optimize the parameters over 3 years.
a. Monthly Rebalance
b. 122 days lookback
c. Use top 2 ETF
d. Max allocation = 60%
e. Volatility Attenuator = 2.5
5. I choose these parameter settings based on the optimization giving stable returns across many lookback periods.
6. Add the strategy Precious Metals to the Strategy “0 LI Strategies of strategies”. This is done in the strategy manager.
a. Duplicate 0 LI Strategies of strategies
b. Rename the strategy “0 Test1” or anything you like.
c. Save it
d. Add Precious metals
7. Optimize 0 Test1 and choose parameters that have stability over a 3 year time period.
a. The optimization I only changed the lookback in the original strategy from 166 to 155 and Volatility Attenuator to 0.5.
8. Compare the results.
a. The New model 0Test1 has a better return but a slightly higher volatility.

Questions and Next steps:
1. Did I curve fit by adding PALL? PALL has done very well over the last 3 years. If I added the uranium metal ETF URA it would not do as well. Selecting what ETF’s go in the strategy is critical. The objective of the strategy should drive this.
2. Does the increase in performance justify the increase in volatility?
3. Was 3 years a good time period to optimize across? Or should I use 5 or 10?
4. Let the model run OS for 3 years and then reevaluate it. This is the only way to really tell if the strategy works.
5. If there is no inflation it is possible that commodities and precious metals prices fall. I need to work on the crystal ball.

In this process my goal was to lower volatility of an existing strategy, 0 LI Strategies of strategies. That goal was not achieved but it would be worth exploring other strategies that help reduce volatility in the future. If anyone would like to post their process for creating a strategy and how the assets are chosen please feel free to do so. LI seems to group similar assets into the strategy like US equity vs World Equity. Bond strategy vs Gold strategy. Then they combine them to optimize the KPI.

Cheers,
Mark V.

[Mark Vincent]

Let me ask this a different way.

Does anyone have a logical process to add a strategy to another strategy? For example if you create a strategy why would you add it to the 0 LI Strategies of strategies? I have added a few of my own with very good results.

I’m looking for others thought process.

Cheers,
MV

[Richard Thomas]

Hi Mark,
If you read the Modified Permanent Portfolio article that I wrote you will see my thought process on this topic. My objective was to find a way to improve the Permanent Portfolio by increasing the CAGR and reducing the MaxDD and I was able to achieve this. In the process I was able to create two additional new strategies; Precious Metals and a stock version of the US markets strategy.

When I developed these strategies and wrote the article I too was hoping to get some feedback but so far have received very little. I would be grateful if you could give me your comments on it.

I chose to use a 5 year optimization period purely based upon gut feel. The key, I believe, is to make sure you have a period covering as many market conditions as possible (rising, falling, flat, crash etc for all of the ETFs in your strategy) but over a reasonably tight period of time. 3 years felt too short and 10 years felt too long so I settled on 5, but would welcome everyone’s thoughts on this, especially LI. The recent crash in March and subsequent recovery has helped enormously with providing good data to use in strategy development.

When I optimize I look at all the light squares in the optimization window and try to find the “best” one. The way I do this is to look at the performance (primarily CAGR and MaxDD) of the 8 squares around the selected square so I have the best block of 9, if that makes sense. Ideally there will not be significant deviations in the performance of the surrounding squares. I will then feel I have a stable set of parameters (look back and vol att) to use.

I usually start with using SR and 1 to get a baseline position in a new strategy and will then look to increase the number of ETFs and different combinations of max and min % to see if this improves the results. In the case of my Precious Metals it did not. I also looked at adding CPER, JJC and PPLT to the strategy but these all gave worse performance.

When I have selected my look back and vol att parameters I will tinker with various combinations of the mean rev weight and % values to see if that improves the results. If it does then I will select these and rerun the optimization to see if this changes anything. This feels a bit like curve fitting to me, what are your thoughts?

The big unknown is how well the strategy will perform in the future. I have tried fixing the look back period to be 6 years back and optimize over 5 years and then see how it worked for the next year, but this part of QT is very unstable and it often crashes with exception errors.

The other major question I have is when should you reoptimize your strategy? Every year or just when there is a major new event?

Hope this helps.
Richard

[Mark Vincent]

Thanks Richard I had a quick look and will comment later. How did you paste the screen shots in? Those are very helpful. Saves you describing every parameter.

Cheers,
MV

[Richard Thomas]

Hi Mark.
I had real trouble doing this. Eventually I sent everything to Alex and he posted it for me!

[Mark Vincent]

I really wish LI would get a better forum that can paste with proper formatting and pictures. Or post a quick tutorial on how to post things.

[Author]

Posts