The Absolute/Relative/Tactical Momentum PP

[EdwardjK]

Reub,

Assuming you are referring to the strategy I displayed, why do you think it limited to a single asset?  My table shows that all four ETFs are used, albeit with unequal allocations.

[LC475]

But you also can't cry uncle if/when the HBPP drawsdown -25% again.

Bluster, bluster everywhere.

Data, data...............





........




...........




Buehler?


So, informational note: We have seen no evidence that this statement from MachineGhost is true.

[Reub]

Reub,

Assuming you are referring to the strategy I displayed, why do you think it limited to a single asset?  My table shows that all four ETFs are used, albeit with unequal allocations.

9
I was actually referring to MG's proposal to use the strongest momentum asset as the sole component in a 10% VP.If we are going down that road then why limit ourselves strictly to only the 4 HBPP assets?

[MachineGhost]

I was actually referring to MG's proposal to use the strongest momentum asset as the sole component in a 10% VP.If we are going down that road then why limit ourselves strictly to only the 4 HBPP assets?

I think DecisionMoose's terrible performance of the past few years and of late is a huge red warning flag not to be relying on their black box recipe.  I also don't like how their backtest is unrealistic by assuming you sell/buy at the close on Friday, especially when you go public and offer the signals to the world which can promote huge slippage on Monday.

[img width=1024]http://i59.tinypic.com/28usol1.png[/img]

One way you can finesse this sinking ship is by using a trailing stop of exactly 9.62% on every position.  It will reduce the MaxDD from its currrent of almost -30% to -18% and give you a 1.5% or so more CAR (18.43%).  It won't turn losers into winners, however.  The current win rate is down to 64.91%.

Here are the baseline stats (with transanction costs included) for the single-asset relative momentum:

Daily Relative HBPP
14.82%
-43.21%
Daily AbsRel HBPP
15.73%
-33.36%
Weekly Relative HBPP
15.13%
-41.91%
Weekly AbsRel HBPP
13.79%
-41.78%
Monthly Relative HBPP
15.45%
-31.90%
Monthly AbsRel HBPP
13.13%
-27.29%
Quarterly Relative HBPP
14.48%
-28.09%
Quarterly AbsRel HBPP
15.98%
-20.61%

I'll post the Absolute and Tactical tomorrow when I feel like it.

[Kbg]

MG,

Are you doing AbsRelative this way? If something is plus 30% and something else is -31% you go with -31 over +30 or is it Antonacci AbsMo?

[MachineGhost]

Are you doing AbsRelative this way? If something is plus 30% and something else is -31% you go with -31 over +30 or is it Antonacci AbsMo?

Antonacci was rate-of-change right?  So no it is Faber absolute but the ranking is Antonacci.

I wouldn't read too much into the Quarterly AbsRev.  Too few trades and its far easier to randomly skip a bad period just because the absolute didn't line up in some particular quarter.

[MachineGhost]

It didn't seem to be announced, but Peak2Trough's backtester at http://www.peaktotrough.com/hbpp.cgi now works from 03-05-1968, even though it doesn't default to that.  It is still a wonderful time saver to try out different rebalancing frequencies!  In my view, 20/30 rebalancing bands is the best among the limited choices, though I would like to have seen transanction costs included.

The main difference with my results is I use investable mutual funds + S&P500 price return rather than the non-investable S&P 500 total return index.  I don't think the first Vanguard index fund came out until 1976 or so?  So the missing dividends from 1968 to 1972 along with tracking error and fund management fees overall likely account for the 1% CAGR difference between my results and Peak2Trough's.  I tend to prefer more realistic simulations even though the greed in me rather look at Peak2Trough's.

EDIT: Oh, he's only using monthly gold data until the end of 1971.  I don't believe it will matter because the Achilles Heel happened in 1980.  I'm not sure what he's using before 1977 for 30-year TBonds either.  Since the 70's were a bear market for bonds, the returns could be overstated by using 20-year.  Caveats, caveats...

[EdwardjK]

I don't understand the comments about the length of backtesting.  Some commentators insist on backtesting 30+ years before accepting any strategy's results.  I ask, "Why?"

The markets of today bear no relation to the markets 30 years ago.  Todays markets are globally-connected and exhibit high volatility.  In addition, information is available today in an instant compared to the snail-mail of yesterday.  All of these differences, and many others, impact today's markets.

I, for one, believe that a ten-year backtest provides sufficient validation of an investment strategy.  And, although I have not tested this idea, any backtest should give extra weighting to the past 2-3 years.

Feedback welcome.

[MachineGhost]

Feedback welcome.

It's because you need to achieve statistical significance to be confident your results aren't due to random chance or curve fitting to a limited subset of history that isn't representative of all possible future outcomes.  10 years is just way too short to reach any level of significance.  You don't have to use past history; you can use certain techniques that preserve the inherent correlations and volatilities present in past data to generate synthetic data to simulate on, or just outright use randomized data.  When it comes to the HBPP, economic trends last for decades and we barely have enough history to achieve significance.  For instance, the PP has yet to go through a bond bear market as occured from 1942 to 1980....  it only caught the tail end of it.  The PP was based on 100-years of U.S. history when it was first devised in the late 70's as the "PRPFX".

Also, when it comes to yearly mutual fund returns, you have to look at a 20-year track record to be at the 95% confidence level.  For 99%, I think it would be 40-years.  Ever noticed that mutual funds only advertise 15-years max on top of the misleading "average annual return" instead of CAR?  Gee, I wonder why...

[Xan]

I, for one, believe that a ten-year backtest provides sufficient validation of an investment strategy.

I'm going to attempt to echo Craig and assert that no backtest can ever validate an investment strategy.  Backtests can only invalidate them.