Momentum with Month Skip

[Nick Radge]

I see a lot of literature for momo strategies that exclude the latest month of data. Theory is that in such a low window the momentum effect actually reverses, i.e. false breakouts tend to occur more often.

Anyone tested skipping the most recent month data?

[SaidBitar]

I remember that i tested this before and no added value was found any how i will test it again tonight

[Nick Radge]

Thanks Said. Was my finding as well – just making sure I wasn’t missing something.

[LeeDanello]

Are you guys talking about this?
A reader asked: “In reviewing your various posts on momentum-based trading, I noticed that many impose a one-month delay between momentum calculation and actual trade implementation.

Why the Skip-period in Momentum Strategies?

[Nick Radge]

Yep. Most academic papers cite it now.

[ScottMcNab]

Related info…..showing same

http://backtestwizard.com/testing-different-momentum-rules/

[Anonymous]

I looked at eliminating the most recent month from the momentum calculation and also found no value in doing so for my system.

That said, I looked at the idea of the most recent month being mean reverting in a different way, and did find value…

What I do is I look at my momentum score for the most recent month and eliminate candidate trades above a certain threshold.

E.g.

MoMoScore = ROC(260)
MoMoScoreShort = ROC(20)

Cond1 = …
Cond2 = …
Cond3 = MoMoScoreShort < Threshold LE = Cond1 AND Cond2 AND Cond3 score = Ref(MoMoScore,-1) Score = iif(LE,score,0) So for my ranking score, the most recent data is included, however, candidate entries that have a very high level of recent momentum are screened out. This increased risk adjusted returns for me and was robust across time periods.

[JulianCohen]

I found the following gave a nice improvement.

Code:

MoMoScore = ROC(260);
MoMoScoreShort = ROC(90);

Cond2 = MoMoScore > 0;
Cond3 = MoMoScoreShort > 0;

This helps to remove the tendency to mean revert in the most recent three months.

[Anonymous]

Julian – keep in mind that there are ~20 days in a month, so you are looking at 4.5 months for your short term score above

[JulianCohen]

Brent Hause wrote:

Julian – keep in mind that there are ~20 days in a month, so you are looking at 4.5 months for your short term score above

Thanks Brent. I found anything from 50-90 gave the same result. I’ll test it further to see the effect. 20 for me doesn’t make too much difference…..possible curve fit I guess in my case but I’m thinking that in theory 0-3 months should be mean reverting and periods after that trending.

[Author]

Posts