PP Inspired Momentum - A Comprehensive Approach

A place to talk about speculative investing ideas for the optional Variable Portfolio

Moderator: Global Moderator

User avatar
InsuranceGuy
Executive Member
Executive Member
Posts: 182
Joined: Sun Mar 29, 2015 1:44 pm

PP Inspired Momentum - A Comprehensive Approach

Post by InsuranceGuy » Sat Apr 21, 2018 2:16 am

The Permanent Portfolio investment strategy is based on the economic cycle which Harry Brown divides into four basic categories:
Prosperity / Inflation / Deflation / Recession

At the risk of sounding heretical, I see the economic cycle as more of an continuum moving towards or reversion from Deflation or Inflation with Recession at either extreme:
Recession <==> Deflation <==> Prosperity <==> Inflation <==> Recession

In this world view, Stock asset classes do well in the middle but shift to Gold during Deflation or to Long Bonds during Inflation. Cash plays more of a recessionary role. This view suggests application of momentum separately for Stock asset classes and Gold/Long Bonds, when they are both showing positive momentum you are likely in an economic oscillation between Prosperity and Inflation/Deflation. By applying momentum separately it ensures the strategy is in a diversified position to capitalize on a movement in either direction. If volatility becomes excessive then things get kicked out to cash.

Disclaimer: This is not meant to be a recommendation of any kind, simply my personal observations as I have tried to develop an easy to use trading system using the PP asset classes as a starting point. For those that may not have read my prior posts, I personally trade a slightly more complicated strategy that is based almost exclusively on the below techniques. Some of the below info is repeated from prior posts in an attempt to summarize my journey.

Both early literature and my personal journey with momentum strategy starts out with ranking schemes, then evolves into various asset allocations going by various names: Absolute, Adaptive, Dual, Elastic, Flexible, Minimum Variance, Modern, Relative, Risk Parity, Tactical, etc. They are all variations an a consistent theme which is variable allocation across assets using returns, correlations, covariances, volatilities, or other "crash protection" schemes.

The problem with these generalized momentum strategies is varying degrees between trading costs and missed returns from late reentry after significant corrections. First, most of the momentum strategies require many trades to be effective. A couple of ways to reduce trading are: using few asset classes, using round numbers for allocating assets, or requiring consecutive buy or sell signals.

Secondly, momentum tends to miss out on returns from late reentry after a correction. It shouldn't be a huge surprise this occurs as momentum is based on winning picks continuing to win, but when there are large corrections you will want to find a way to get in before the bounceback happens. There are a bunch of ways to do this whether by CAPE or other financial ratios, macroeconomic indicators such as unemployment, or at worst long term momentum. There is considerably less literature available available on methods of doing this, I believe because the various methods used affects asset classes differently (macroeconomic effects on stocks vs bonds) or are not comparable across asset classes (P/E for gold?). While I have spend a lot of time developing a macroeconomic indicator, I get nearly the same bang from simply using long term momentum.

Lastly, parameter choices abound. Be careful about overfitting the data especially if you aren't using daily data (or even if you're using daily data). Simple is better as it is less likely to break in the future.

Attempting to remedy these shortcomings, I propose a basic monthly trading strategy with the following assumptions:

Momentum "Prosperity" Building Block: Dual Momentum of Stock Asset Class(es), 12-month lookback of returns over SHY, pick the Top 1 positive
Momentum "Risk Off" Building Block: Dual Momentum of Long Bonds and Gold, 12-month lookback of returns over SHY, pick the Top 1 positive
Value "Prosperity" Building Block: Absolute Momentum of a Stock Asset Class, 5-year lookback using -ln(current month end price/average(54-66 month end prices)
Asset Class Selection: If there are both "Prosperity" and "Risk Off" asset classes, then take the lowest 60-day volatility for any combination of 50% "Prosperity"/50% "Risk Off". Otherwise, take the lowest 60-day volatility of any single asset class.
Asset Class Allocation: Volatility is scaled to a target annual volatility rounded to 0%, 50%, or 100%. In Excel terms this is: Round(If(2 assets,0.5,1)*target*sqrt(252)/60-day volatility*2,0)/2
Risk-Free Allocation: Anything not allocated to "Prosperity" or "Risk Off" asset classes is allocated to SHY

While this basic strategy may seem anything but basic, I am simply picking the best momentum asset classes based on a 1-year lookback separately for "Prosperity" (for returns) and "Risk Off" (for drawdowns) along with a "Prosperity" asset class below its 5 year prior price (for returns) and then selecting the combination with the lowest volatility (for volatility) allocating based on the volatility (for drawdowns). Rounding is done to reduce trade volumes.

The Nominal Returns for Jan 1970 - Dec 2017 are as follows:

Code: Select all

                         CAGR     MaxDD     UPI   Trades/Yr
MOM+VAL GB (8.5% VOL)   15.9%    -17.7%    1.72     8.3 
MOM+VAL PP (8.5% VOL)   13.7%    -15.8%    1.44     6.3
GB (15/25)              10.2%    -21.5%    0.70     3.5
PP (15/35)               9.4%    -21.4%    0.44     1.2
PP (20/30)               9.0%    -19.8%    0.42     3.3
S&P                     10.4%    -55.3%    0.26
Note: trades/year represent an inflow or outflow of an asset, so a trade from one asset to another would count as 2 trades
Note2: my personal mix targets a slightly higher volatility with slightly tuned parameters giving: 18.5% CAGR/-20.1% MaxDD/1.83 UPI/5.2 Trades

The last item worth mentioning is that Antonacci's GEM momentum approach seems to garner a lot of attention. I have personally come to dislike the portfolio the more I play with it as it is does not seem to perform consistently throughout economic cycles. International stocks are highly correlated with gold prices (in USD), but could be included in the "Prosperity" asset blocks like SCV as in the GB approach above. Based on my tests after separating "Prosperity" from "Risk Off" assets, combinations of assets with International stocks no longer rank as favorably as in past attempts.

I'd be happy to recommend academic literature or data sources if anyone interested in learning more. Hopefully some of my ramblings will help someone looking for a practical momentum-based approach that outperforms buy and hold portfolios.

IG

EDIT: I ultimately decided against going down the GB route and ended up moving towards something more GEM-like so I have removed GB from the thread title.
Last edited by InsuranceGuy on Sat Feb 02, 2019 2:46 pm, edited 2 times in total.
HappyMan
Full Member
Full Member
Posts: 74
Joined: Thu Nov 09, 2017 7:01 am

Re: PP/GB Inspired Momentum - A Comprehensive Approach

Post by HappyMan » Sun Apr 22, 2018 7:28 am

IG,

Nice job on outlining drawbacks of Dual Momentum.

People here like to backtest things. Can you link your portfolio to something like https://www.portfoliovisualizer.com/ or at least provide tickers for holdings.

Questions:

- How do you spread percentage-wise a depo among the six areas?
- How did your portfolio do last January?

Academic literature also would be nice.

I am sure more questions are coming.

Regards,
User avatar
InsuranceGuy
Executive Member
Executive Member
Posts: 182
Joined: Sun Mar 29, 2015 1:44 pm

Re: PP/GB Inspired Momentum - A Comprehensive Approach

Post by InsuranceGuy » Sun Apr 22, 2018 5:13 pm

HappyMan -

Portfolio Visualizer does not have the capability to do a backtest for this method, but I did provide backtested results in my original post. As far as tickers, I have pieced together multiple data sources to get data back to 1969:
Large Blend: IVV (2001+), VFINX (1989-2000), ^GSPC+Shiller Dividends (1969-1988)
Small Value: VBR (2005+), VISVX (1999-2004), Fama/French (1969-1989)
Long Bonds: TLT (2003+), VUSTX (1987-2002), FRED 30yr Treasury Yields (1977-1986), FRED 20yr Treasury Yields (1969-1976)
Gold: IAU (2006+), GLD (2005), LBMA (PM) (1969-2004)
Cash: SHY (2003+), VFISX (1992-2002), Average(FRED 1yr Treasury Yields,FRED 3yr Treasury Yields) (1969-1991)

As I was a little vague in the post, here is a more detailed breakdown of how to allocate funds every month for the MOM+VAL GB (8.5% VOL) example. This example assumes you are looking to trade on the last day of December 2017.

Momentum "Prosperity": IVV (IVV won in Dual Momentum applied to IVV and VBR and IVV returned more than SHY)
Momentum "Risk Off": IAU (IAU won in Dual Momentum applied to TLT and IAU and IAU returned more than SHY)
Value "Prosperity": n/a (VBR is higher than it was 5 years ago so it doesn't qualify)

Asset Class Selection: IVV/IAU. Since we only have one "Prosperity" asset (IVV) and one "Risk Off" asset (IAU) they are both selected. If we have multiple "Prosperity" assets with a "Risk Off" asset we'd choose based on the lower variance pair, if only one asset class qualifies it is selected.

Asset Class Allocation: 50% IVV/50% IAU. The formula above had a couple small errors and should have been Round(If(2 assets,0.5,1)*min(1,target/sqrt(252)/60-day volatility)*2,0)/2, so Round(0.5*min(1,0.085/sqrt(252)/0.002993)*2,0)/2=0.5 or 50% each.

Risk-Free Allocation: 0% SHY. Since 100% of our allocation is already taken up by IVV/IAU nothing is allocated to SHY.

For academic literature, you'll have to be more specific what you are looking for. For a start though these introduce many of the concepts in the strategy:
"Risk Premia Harvesting Through Dual Momentum" (Antonacci)
"Adaptive Asset Allocation: A Primer" (Butler)
How did your portfolio do last January?
In January 2017, the MOM+VAL GB (8.5%) example increased 2.7%.

Hopefully that helps, it's hard to get the right amount of content in a post as there is never enough detail but at the same time no one wants to read excessive info.

IG
HappyMan
Full Member
Full Member
Posts: 74
Joined: Thu Nov 09, 2017 7:01 am

Re: PP/GB Inspired Momentum - A Comprehensive Approach

Post by HappyMan » Mon Apr 23, 2018 6:33 am

Hi IG,

I like how more information is coming out. Appreciate that.

Found the first paper online:

http://www.optimalmomentum.com/RiskPremiaHarvesting.pdf

The process of selection seems somewhat complicated. Let's take it step by step. Suppose, one is using the strategy today. (I'll copy and correct for clarity):

1) "Momentum "Prosperity": IVV (IVV won in Dual Momentum applied to IVV and VBR and IVV returned more than SHY)";

Checked in stockcharts, same outcome.

2) "Momentum "Risk Off": IAU (IAU won in Dual Momentum applied to TLT and IAU and IAU returned more than SHY)";

Same.

3) "Value "Prosperity": n/a (VBR is higher than it was 5 years ago so it doesn't qualify)" -

Can you please expand on this component/step? Not sure why it is here and how it works.

4) "Asset Class Selection: IVV/IAU. Since we only have one "Prosperity" asset (IVV) and one "Risk Off" asset (IAU) they are both selected. If we have multiple "Prosperity" assets with a "Risk Off" asset we'd choose based on the lower variance pair, if only one asset class qualifies it is selected."

So, the outcome is either one or a pair of assets?

5) "Asset Class Allocation: 50% IVV/50% IAU. The formula above had a couple small errors and should have been Round(If(2 assets,0.5,1)*min(1,target/sqrt(252)/60-day volatility)*2,0)/2, so Round(0.5*min(1,0.085/sqrt(252)/0.002993)*2,0)/2=0.5 or 50% each."

Do you happen to have an excel file that others can use?

6) "Risk-Free Allocation: 0% SHY. Since 100% of our allocation is already taken up by IVV/IAU nothing is allocated to SHY."

If there were only one asset, then SHY would take 50%?

I see that you take the end of a month as the day for analysis and decision-making. Others prefer the 1st. Is it just a matter of preference? Also, have you thought about better times during a year to enter this strategy?
User avatar
InsuranceGuy
Executive Member
Executive Member
Posts: 182
Joined: Sun Mar 29, 2015 1:44 pm

Re: PP/GB Inspired Momentum - A Comprehensive Approach

Post by InsuranceGuy » Mon Apr 23, 2018 10:04 am

3) "Value "Prosperity": n/a (VBR is higher than it was 5 years ago so it doesn't qualify)" -

Can you please expand on this component/step? Not sure why it is here and how it works.
HappyMan - If you recall from the original post, one of my concerns with momentum is that it tends to miss out on returns from late reentry after a correction. The very nature of momentum is betting on winning picks to continue to win, but again when there are large corrections you will want to find a way to get in before the bounceback happens. While there are multiple ways to do this listed above, I get nearly the same bang as other methods out there by simply using 5yr momentum.
4) "Asset Class Selection: IVV/IAU. Since we only have one "Prosperity" asset (IVV) and one "Risk Off" asset (IAU) they are both selected. If we have multiple "Prosperity" assets with a "Risk Off" asset we'd choose based on the lower variance pair, if only one asset class qualifies it is selected."

So, the outcome is either one or a pair of assets?
Yes, the thought here is that the possibilities after you evaluate the momentum and value building blocks you'll have either only "Prosperity" asset classes, only "Risk Off" asset classes, or pairs of "Prosperity"/"Risk Off" asset classes. If you have any pairs then you choose the pair with the lowest variance, if you only have asset classes of the same type then choose the asset class with the lowest variance of that type.
5) "Asset Class Allocation: 50% IVV/50% IAU. The formula above had a couple small errors and should have been Round(If(2 assets,0.5,1)*min(1,target/sqrt(252)/60-day volatility)*2,0)/2, so Round(0.5*min(1,0.085/sqrt(252)/0.002993)*2,0)/2=0.5 or 50% each."

Do you happen to have an excel file that others can use?
I don't have any that are publicly shareable due to size, but maybe I could paste some values into Google Sheets at some point. The idea of this is that you start with 50%/50% for pairs or 100% for a single asset class. Then you scale it down if needed based on the trailing 60-day volatility.

So, in this case where the target is 8.5% volatility the min(1,target/sqrt(252)/60-day volatility) portion is simply a factor applied when the 60-day volatility is larger than the target. For example if the annualized 60-day volatility is 16% then min(1,target/sqrt(252)/60-day volatility)=min(1,target/annualized 60-day volatility)=min(1,8.5%/16%)=0.53125. If this is applied to a pair then the math is round(50%*0.53125*2,0)/2=50%, but in the case where it is applied to a single asset class the math is round(100%*0.53125*2,0)/2=50% so it scaled it back because of the volatility.
6) "Risk-Free Allocation: 0% SHY. Since 100% of our allocation is already taken up by IVV/IAU nothing is allocated to SHY."

If there were only one asset, then SHY would take 50%?
Hopefully my answer from 5 clarifies this. When there is only one asset class it starts with 100% weight, so SHY would only take 50% when volatility is high.
I see that you take the end of a month as the day for analysis and decision-making. Others prefer the 1st. Is it just a matter of preference? Also, have you thought about better times during a year to enter this strategy?
As some asset classes tested in the various literature out there are only available monthly there is a bias to trade either the first/last day of the month. I have tested various days and some are slightly better or worse than others but I choose the last day both because I tend to do my trades in the afternoons and wish to not have an differences in results with other academic research by choosing a day other than the first/last of the month.
HappyMan
Full Member
Full Member
Posts: 74
Joined: Thu Nov 09, 2017 7:01 am

Re: PP/GB Inspired Momentum - A Comprehensive Approach

Post by HappyMan » Mon Apr 23, 2018 11:58 pm

It is getting clearer.
InsuranceGuy wrote:
I don't have any that are publicly shareable due to size, but maybe I could paste some values into Google Sheets at some point.
Sounds like a good idea.
The idea of this is that you start with 50%/50% for pairs or 100% for a single asset class. Then you scale it down if needed based on the trailing 60-day volatility.
Where can one check volatility of the last 60 days?
So, in this case where the target is 8.5% volatility the min(1,target/sqrt(252)/60-day volatility) portion is simply a factor applied when the 60-day volatility is larger than the target. For example if the annualized 60-day volatility is 16% then min(1,target/sqrt(252)/60-day volatility)=min(1,target/annualized 60-day volatility)=min(1,8.5%/16%)=0.53125. If this is applied to a pair then the math is round(50%*0.53125*2,0)/2=50%, but in the case where it is applied to a single asset class the math is round(100%*0.53125*2,0)/2=50% so it scaled it back because of the volatility.
I must say - well done on formula composition. I'll look forward to that google table with all the formulas in place.

In the case of a single asset, if it is scaled to 50%, the other 50% go to SHY?
As some asset classes tested in the various literature out there are only available monthly there is a bias to trade either the first/last day of the month. I have tested various days and some are slightly better or worse than others but I choose the last day both because I tend to do my trades in the afternoons and wish to not have an differences in results with other academic research by choosing a day other than the first/last of the month.
It'll be interesting to try this on the 30th of this month.
D1984
Executive Member
Executive Member
Posts: 385
Joined: Tue Aug 16, 2011 7:23 pm

Re: PP/GB Inspired Momentum - A Comprehensive Approach

Post by D1984 » Tue Apr 24, 2018 1:39 am

HappyMan wrote:It is getting clearer.
InsuranceGuy wrote:
I don't have any that are publicly shareable due to size, but maybe I could paste some values into Google Sheets at some point.

How big is it? Depfile offers free hosting (for less than 30 days) for any file under 1 GB.. I just did an upload speed test and uploaded a 26.7 MB Excel spreadsheet (which had a bunch of macros) and it took less than 1.5 minutes to upload.

You do have to register with an email address in order to upload but since you get access as soon as you register (i.e. they don't require you to respond to any verification email) you can just input a made-up email address and then immediately upload the file.
User avatar
InsuranceGuy
Executive Member
Executive Member
Posts: 182
Joined: Sun Mar 29, 2015 1:44 pm

Re: PP/GB Inspired Momentum - A Comprehensive Approach

Post by InsuranceGuy » Tue Apr 24, 2018 10:51 pm

HappyMan wrote:Where can one check volatility of the last 60 days?
Just download the closing prices for the last 60 days into excel and then use the STDEV function.
HappyMan wrote:It is getting clearer.
In the case of a single asset, if it is scaled to 50%, the other 50% go to SHY?
That is correct.
D1984 wrote:How big is it? Depfile offers free hosting (for less than 30 days) for any file under 1 GB.. I just did an upload speed test and uploaded a 26.7 MB Excel spreadsheet (which had a bunch of macros) and it took less than 1.5 minutes to upload.

You do have to register with an email address in order to upload but since you get access as soon as you register (i.e. they don't require you to respond to any verification email) you can just input a made-up email address and then immediately upload the file.
It's not just because of size, even though it is almost 500mb. I need to remove a bunch of stuff and make something simple that demonstrates the strategy without the various scenario and what-if testing that I have built-in.

IG
HappyMan
Full Member
Full Member
Posts: 74
Joined: Thu Nov 09, 2017 7:01 am

Re: PP/GB Inspired Momentum - A Comprehensive Approach

Post by HappyMan » Wed Apr 25, 2018 3:11 am

InsuranceGuy wrote:
HappyMan wrote:Where can one check volatility of the last 60 days?
Just download the closing prices for the last 60 days into excel and then use the STDEV function.
I admire your skills with formulas and Excel. The rest of us need little bit more explanation. If we need STDEV function, will PortfolioVisualizer work? For example, I have three different portfolios, one for each most common scenario,

https://www.portfoliovisualizer.com/bac ... ion2_3=100

If it works, then the volatility in February-April is between 3.28 and 6.28% depending on assets' allocation. Hence, no need to adjust. Is this kind of thinking correct?
It's not just because of size, even though it is almost 500mb. I need to remove a bunch of stuff and make something simple that demonstrates the strategy without the various scenario and what-if testing that I have built-in.
People like simple. Something to play with will answer a lot of questions.

In the meantime, thank you for sharing your findings with the rest of us!
User avatar
InsuranceGuy
Executive Member
Executive Member
Posts: 182
Joined: Sun Mar 29, 2015 1:44 pm

Re: PP/GB Inspired Momentum - A Comprehensive Approach

Post by InsuranceGuy » Wed Apr 25, 2018 9:19 am

HappyMan wrote:I admire your skills with formulas and Excel. The rest of us need little bit more explanation. If we need STDEV function, will PortfolioVisualizer work? For example, I have three different portfolios, one for each most common scenario,

https://www.portfoliovisualizer.com/bac ... ion2_3=100

If it works, then the volatility in February-April is between 3.28 and 6.28% depending on assets' allocation. Hence, no need to adjust. Is this kind of thinking correct?
Correct thinking, but unfortunately the returns at PV are monthly not daily so the volatility is way too low. The idea behind using the trailing volatility is that when the market whipsaws you want to scale back your positions, but with monthly returns the volatility is hidden within the months.

I created something simple on google sheets real quick that will give you something close enough. I just used 90 real days which will usually be 59-61 trading days. https://docs.google.com/spreadsheets/d/ ... sp=sharing
HappyMan
Full Member
Full Member
Posts: 74
Joined: Thu Nov 09, 2017 7:01 am

Re: PP/GB Inspired Momentum - A Comprehensive Approach

Post by HappyMan » Thu Apr 26, 2018 8:26 am

InsuranceGuy wrote: I created something simple on google sheets real quick that will give you something close enough. I just used 90 real days which will usually be 59-61 trading days. https://docs.google.com/spreadsheets/d/ ... sp=sharing
Thanks! A titanic job well done.

Unless you are planning to do this and publisch signals every month, there must be a way for others to do it themselves. Where can one find these numbers online?
User avatar
InsuranceGuy
Executive Member
Executive Member
Posts: 182
Joined: Sun Mar 29, 2015 1:44 pm

Re: PP/GB Inspired Momentum - A Comprehensive Approach

Post by InsuranceGuy » Thu Apr 26, 2018 9:02 am

HappyMan wrote:
InsuranceGuy wrote: I created something simple on google sheets real quick that will give you something close enough. I just used 90 real days which will usually be 59-61 trading days. https://docs.google.com/spreadsheets/d/ ... sp=sharing
Thanks! A titanic job well done.

Unless you are planning to do this and publisch signals every month, there must be a way for others to do it themselves. Where can one find these numbers online?
Surprisingly I have had a rough time finding trailing volatility readily available. I could make that spreadsheet update automatically every time you open it, there might be a better way but I'm still looking for it.
Post Reply