Reconstructing Antonacci's Dual Momentum Systems

Research
1

I’ve spent a better part of a week trying to construct/reconstruct some of Gary Antonacci’s Dual Momentum models in P123 simulations.

And I’m not coming anywhere close. To keep this simple, I’ll narrow in on a single data point of monthly returns.

On Gary’s website, he posts the returns for another (proprietary) model, the Dual Momentum Fixed Income model. The stated asset selections are:

Barclays Capital US Credit Bond
Barclays Capital US High Yield Corporate Bond
Barclays Capital US Mortgage Backed Securities
90 Day US Treasury Bills

The closest ETF proxies I can find are, respectively: CRED, JNK, MBG, BIL

From the author’s performance page (http://www.optimalmomentum.com/trackrecord2.html) we see that the April 2015 return was 1.2%.

Here are the April 2015 returns (data from Yahoo Finance) for the corresponding ETFs.

CRED -0.90%
JNK +0.87%
MBG +0.04%
BIL -0.02%

How can any combination of those four asset classes generate +1.2% when the highest single asset generated +0.87%? (I know, it can’t, unless he’s shorting something. Which if he is, he makes zero mention of anywhere.)

And this is not the only example, I chose to highlight a single month for clarity’s sake. I’ve found many months of difficult-to-explain stated returns on the published performance record.

Perhaps it’s ETF tracking error? If so, such a gargantuan tracking error vs the underlying index reveals a fatal flaw to the strategy itself. And if it’s not tracking error, well, then what in the world is Antonacci’s secret sauce to turn 0.87 into 1.2?

Anyone else tried their hand at implementing Antonacci’s ideas here?

2

Hi Remmerde,

Two points which explain the monthly difference in April:

  1. Mistake in your return calculations. Looking at Antonacci’s returns, the model must be invested in JNK. This ETF was up 1.0% - including the monthly dividend.
  2. Substantial gap between JNK and the benchmark index. Please check out this link: https://www.spdrs.com/product/fund.seam?ticker=JNK

Two points which can explain longer-term differences:

  1. Fees & Slippage. It’s been some time that I read Antonacci’s book, but I think he considers fees. His fee structure might be very different from the fee/slippage assumptions made by P123/you.
  2. Buying date/time. Again, I do not remember all the details, but Antonacci might buy at the close, whereas you buy on the next day’s open.

Hope this helps!

Cyberjoe

3

Thanks Cyberjoe, my JNK calculation was indeed wrong. I had HYG instead of JNK in that column. And thanks for the link - the tracking error is pretty substantial for JNK.

All my testing so far is zero fees, zero slippage. Just to get a baseline.

I think Antonacci’s got some secret sauce for DMFI. I notice in the fine print, he says it’s “a proprietary, rule-based approach”. Whereas for GEM, the same fine print calls it “a rule based approach”. GEM was fully described in the book. His DMFI is not.

So maybe my efforts are futile - I found that trying to apply the dual momentum rules to his DMFI basket of fixed income assets (the presumed equivalent ETFs) is significantly off from his performance. At least for the short time (~4 years) the ETFs have been around. The straight Dual Momentum method resulted in -4.3% for 2011. His DMFI result was +15.7%!

The spreadsheet attached has my manual attempted reconstruction of DMFI

-Mike

Antonacci DMFI reproduction w JNK.xlsx (44.4 KB)