The Absolute/Relative/Tactical Momentum PP

[MachineGhost]

Here's what an absolute momentum PP's performance looks like:

[align=center][img width=1024]http://i59.tinypic.com/2pzl7rt.png[/img]
7.50% CAR, -8.84% MaxDD[/align]

Weights are 33.33% each since cash allocating is built-in.  An excellent reduction from the HBPP's -25% MaxDD, but the CAR takes a hit.

[MachineGhost]

What are the rules?

Monthly buying (selling) if excess return above T-Bills > 0 (<0) and/or C > MA (< MA); 50% scale-in (out) for either.  I also optimized the lookback periods for each asset from 1 to 12 months.

EDIT: I've just updated the stats to account for dividends where available and take out modern-day commissions.  No dividends for 1968-1972.

[EdwardjK]

MG,

The folks over at logical-invest.com have just published a similar strategy.  It's fully described in the blog "Will We Ever Kill this Bug?"

Part of my comments on their modifications to Harry Browne's Permanent Portfolio centered on trading frequency.  Browne developed the PP as a "sleep well" strategy that needed only annual rebalancing.  In the pursuit of better returns, your modifications require monthly trading.

Although some may disagree, I don't think the additional return and the anxiety that comes along with higher trading frequency are worth it.

Have you considered a higher weighting based upon the better performing asset?  For example, if gold is generating a higher return based upon a 60-day moving average, then the portfolio might weight gold at 30% instead of 25%, with the offset in the lowest performing asset.  This approach retains the core PP strategy and should not increase trading frequency all that much (I think).

I created a new post a couple days ago injuring what changes Browne might have made to the PP strategy if he were alive today.  Last I looked, everyone said "none".

Why change what is working well?

[Kbg]

Do you use Amibroker for your studies/

[MachineGhost]

The folks over at logical-invest.com have just published a similar strategy.  It's fully described in the blog "Will We Ever Kill this Bug?"

I checked it out.  Like similar studies that I've seen, they use such a limited history that I can't give them any credence.  I also tested the bonds back to 1943 to make sure the parameters was stable.

Although some may disagree, I don't think the additional return and the anxiety that comes along with higher trading frequency are worth it.

Have you considered a higher weighting based upon the better performing asset?  For example, if gold is generating a higher return based upon a 60-day moving average, then the portfolio might weight gold at 30% instead of 25%, with the offset in the lowest performing asset.  This approach retains the core PP strategy and should not increase trading frequency all that much (I think).

Monthly is not "high frequency".  Once a month is a long time between check-ins.  I can't even remember what I ate last week, nevermind a month ago.  I specifically made the decision not to use daily or weekly because of the stress and whipsawing.  Would you be happier if it was quarterly?  There's usually very little difference between monthly and quarterly with momentum.

I created a new post a couple days ago injuring what changes Browne might have made to the PP strategy if he were alive today.  Last I looked, everyone said "none".

Read it again.  I gave some suggestions he may have been wise enough to have thought of.

Why change what is working well?

Changing the strategic portfolio weightings gets into the tactical allocation forecasting ballgame, which I think is a fool's game and clearly goes against the spirit of the HBPP.  Nothing has proven to beat equal-weighting schemes historically, including overweighting outperforming asset classes...  which is again against the spirit of the HBPP.  I and many others have literally tried them all.

So I ask you, why change what is working well?

Do you use Amibroker for your studies/

Yup!  And R, Perl or Python when applicable.

[buddtholomew]

MG is the smartest person in the world, at least he thinks he is..what was the name of that hedge fund full of PhD's? What is their track record?

[Reub]

The numbers are so small that I can't read them, sonny!

[MachineGhost]

The numbers are so small that I can't read them, sonny!

Click on the image to enlarge it, four eyes!

[MachineGhost]

MG is the smartest person in the world, at least he thinks he is..what was the name of that hedge fund full of PhD's? What is their track record?

Long Term Capital Management.  And I know exactly how and why they failed.

[MachineGhost]

EdwardjK read this thread where we explored the question you have in mind and then some: http://gyroscopicinvesting.com/forum/pe ... -and-risk/

[buddtholomew]

MG is the smartest person in the world, at least he thinks he is..what was the name of that hedge fund full of PhD's? What is their track record?

Long Term Capital Management.  And I know exactly how and why they failed.

It was a rhetorical question as everyone knows why they failed.. Overconfidence.

[MachineGhost]

It was a rhetorical question as everyone knows why they failed.. Overconfidence.

Hmm, I wouldn't say that.  It was ignorance.  They were certainly overconfident and arrogant about their ignorance.  PhD's don't live in the real world.  Very few academics have any real trading or investment experience.  They don't know what they don't know.  If we always waited for them to catch up with the real world, nothing would ever get done.  How long has it been now that they've been beating the EMH drum and being flat-out wrong?  60 years?  LTCM is a spectacular example of putting EMH into practice and failing.

Anyway, I've updated the stats again to reflect the best case scenario.  For this particular situation, worse than best case would up the CAR but also the MaxDD.

[Reub]

What are the rules?

Monthly buying (selling) if excess return above T-Bills > 0 (<0) and/or C > MA (< MA); 50% scale-in (out) for either.  I also optimized the lookback periods for each asset from 1 to 12 months.

EDIT: I've just updated the stats to account for dividends where available and take out modern-day commissions.  No dividends for 1968-1972.

        2
E=mc ?

What's C? What's MA?  What's scale-in? What's excess return? Buying/selling of what?

[EdwardjK]

EdwardjK read this thread where we explored the question you have in mind and then some: http://gyroscopicinvesting.com/forum/pe ... -and-risk/

MG,

Ok, I read through the 10 pages of posts.  This was certainly an exhaustive and exhausting discussion.  But can you say that any of the approaches described resulted in a meaningful improvement over the basic HBPP?

The HBPP has the benefit of being easy to understand, easy to implement, and easy to maintain.  Transactions are limited to once a year, maybe twice if all hell breaks loose.  Alternatively, what I read requires monthly transactions and, I think, more than four assets to manage.

Are the results so superior to the basic HBPP that it warrants implementation?

[MachineGhost]

Are the results so superior to the basic HBPP that it warrants implementation?

If you can live with the reduced return and increased shortfall risk.  See my post in Other Discussions... "The Cult of Emotion".
I think you can probably ameliorate the reduced return / shortfall risk by being more judicious in your use of investments that benefit from Prosperity.  There's a whole world out there beyond just the secondary market for [overpriced] public stocks.

If you don't want the hassle, no one is forcing you.  But you also can't cry uncle if/when the HBPP drawsdown -25% again.  Look at last year (2013) when the PP had a minorly bad year...  pretty much everyone gave up on it around here except for the diehards.  That's probably a good lesson for not checking it more than once a year.

[KevinW]

pretty much everyone gave up on it around here except for the diehards

Damn straight.

[EdwardjK]

Are the results so superior to the basic HBPP that it warrants implementation?

Look at last year (2013) when the PP had a minorly bad year...

MG,

Absolutely a fair point.  Which is why I think a momentum-based allocation across the four asset classes is a reasonable modification of the HBPP strategy.  A simple weighting towards the highest performing asset "should" improve performance without significantly increasing trading and increasing risk (although performance would not reach anywhere near the S&P 500 30%+ return).

I will play with this idea and report back.  It'll take a bit of time since I do not have your programming skills and will need to download data for manual manipulation.

[MachineGhost]

Absolutely a fair point.  Which is why I think a momentum-based allocation across the four asset classes is a reasonable modification of the HBPP strategy.  A simple weighting towards the highest performing asset "should" improve performance without significantly increasing trading and increasing risk (although performance would not reach anywhere near the S&P 500 30%+ return).

I will play with this idea and report back.  It'll take a bit of time since I do not have your programming skills and will need to download data for manual manipulation.

I have actually been backtesting a relative momentum allocation among the four asset classes (14.13% CAR, -31.93% MaxDD).  That's riskier than the HBPP.  So I think your Tactical idea may have merit after all, i.e. allocate 5% extra to the outperforming asset class, reduce cash by 5% or from all others proportionately.  It's just a question of whether absolute momentum or fully invested is superior for risk management within such a framework.

Adding a 25% cash allocation brings the MaxDD up to the HBPP but reduces the CAR to only 12.05% (includes transanction costs).  That's about 5% more CAR than the HBPP.

[EdwardjK]

MG et al,

Following up on my earlier post, I want to share the results of two momentum strategies using the HBPP components.

I pulled daily data for VTI, TLT, GLD and SHY from Yahoo Finance.  GLD is available starting only in 2005, so I pulled sufficient data for all four ETFs to make the calculations comparable.

In Excel, I calculated the 63-day return of each ETF and then weighted the ETF percentage allocation using the 63-day returns.  The four resulting percentage allocations were then rounded to the nearest 5%, with any adjustment to SHY to get it all to equal 100%.

Using December 31, 2005 as the starting point, I allocated $100,000 to the ETFs and applied the daily returns.  In one scenario, I assumed monthly rebalancing and in the second scenario I assumed quarterly rebalancing.  NO commission fees were deducted.

Subject to anyone who has better programming skills than I, here are the results compared to the HBPP (HBPP data from ETFReplay):

                        Monthly            Quarterly              HBPP

CAGR                9.2%                9.3%                  7.2%

STDEV-P            5.5%                5.3%                  5.1%  (Standard deviation of annualized returns)

2014 YTD          9.7%                9.8%                  9.8%
2013                  0.6%                1.8%                (2.0%)
2012                  8.1%                8.1%                  6.5%
2011                14.3%              12.4%                11.5%
2010                14.4%              14.2%                14.5%
2009                11.7%              13.1%                  7.9%
2008                (1.2%)              (1.2%)                2.1%
2007                14.2%              14.5%                13.4%
2006                12.0%              12.2%                10.7%

Based upon this exercise, it appears that using a basic momentum strategy does improve the results of the HBPP.  However, quarterly rebalancing yields a slightly better return, minimizes the cost and effort of monthly rebalancing, and lets you sleep better at night.

I welcome anyone who wants to validate these results and use proxies for older historical data so as to expand the historical backtest.

Editing in the allocations by rebalancing date:

Date         TLT VTI GLD SHY
12/19/14 25% 25% 20% 30%
10/1/14 25% 25% 20% 30%
7/1/14 25% 25% 20% 30%
4/1/14 25% 25% 20% 30%
1/2/14 25% 30% 20% 25%
10/1/13 20% 30% 20% 30%
7/1/13 20% 35% 15% 30%
4/1/13 20% 35% 15% 30%
1/2/13 20% 35% 15% 30%
10/1/12 20% 35% 15% 30%
7/2/12 25% 35% 15% 25%
4/2/12 20% 40% 15% 25%
1/3/12 20% 40% 15% 25%
10/3/11 30% 30% 15% 25%
7/1/11 25% 35% 15% 25%
4/1/11 25% 35% 15% 25%
1/3/11 20% 40% 15% 25%
10/1/10 20% 35% 20% 25%
7/1/10 25% 35% 15% 25%
4/1/10 25% 35% 15% 25%
1/4/10 20% 40% 20% 20%
10/1/09 20% 40% 20% 20%
7/1/09 15% 50% 15% 20%
4/1/09 15% 40% 20% 25%
1/2/09 30% 25% 25% 20%
10/1/08 25% 30% 20% 25%
7/1/08 25% 30% 20% 25%
4/1/08 25% 25% 25% 25%
1/2/08 25% 25% 25% 25%
10/1/07 25% 25% 30% 20%
7/31/07 25% 30% 25% 20%
4/2/07 25% 25% 25% 25%
1/3/07 20% 30% 30% 20%
10/2/06 25% 30% 25% 20%
7/3/06 25% 25% 25% 25%
4/3/06 20% 30% 30% 20%
12/30/05 20% 25% 30% 25%

[MachineGhost]

I welcome anyone who wants to validate these results and use proxies for older historical data so as to expand the historical backtest.

This is an excellent start, but you left out the risk metrics.  We know that in general that relative momentum is riskier or does nothing to reduce risk.  Any increased return would have to more than offset any increased risk.

I noticed you weighted multiple assets by their momentum rank (how did you prevent TBills from being minuscle?).  That adds a layer of complexity that isn't so simple to do in real time, but it does make backtesting a lot easier.  I wanted to avoid a situation like that this time as compared to "Engineering Targeted Returns".

Lets say that you have a VP at the traditional 10% of assets and trade a single asset relative momentum PP strategy in it.  Along with the other 90% of assets to the HBPP, any asset that is evidencing momentum in the VP would actually contribute what I calculate is a formal 2.25% allocation increase to the same asset in the HBPP.  So you might as well dispesne with the VP and just use the HBPP tweaking whatever asset is currently experiencing momentum.

[EdwardjK]

MG,

Appreciate the feedback.  I added the standard deviation of the annualized returns and the ETF allocations by rebalancing date.

[MachineGhost]

Can you explain more your weighting scheme?  Especially this:

7/1/09  15%  50%  15%  20%

[EdwardjK]

MG,

On July 1, 2009, the 63-day return was:

      63-day return      Calculated Weightings      Rounded

TLT      1.04%             15.34%                          15%

VTI      3.32%              48.99%                         50%

GLD    1.17%           17.27%                          15%

SHY    1.25%              18.40%                            20%

In this case, VTI has the highest 63-day return and is assigned the highest allocation.

[MachineGhost]

Okay, just checking.  I can easily code this up but not the other strategy of adding 2.25% to just one asset.  Stay tuned.

[Reub]

If we are going to use a DecisiojMoose type of VP strategy  (which has been fairly unproductive lately) then why not just use DecisionMoose's highest signal for our 10% VP instead of limiting ourself to just one of the four HBPP assets?