Momentum trading

This is something totally different from what we do with SQX, but also very interesting (if automated) and still related to trading in general.

At first I will honestly add that I have no practical experience in momentum type trading and I do not know if this can be applied to other markets, however I like this particular idea very much since it is based on ‘everlasting’ economical world-wide dependencies, between US and the rest of the world, greed and fear and cash. It has a very strong fundamental bias.

This idea is called Dual momentum.

Sometime ago while listening to a podcast, I’ve learned about “dual momentum” longterm trading (investing) strategy. It’s a very interesting and extremely simple way of ‘beating’ the market and it could be ideal for busy traders since it requires only few minutes of your time per month!
The mastermind behind this is: Gary Antonacci
Website: http://www.optimalmomentum.com

Idea is explained here:
https://vimeo.com/164047189

Podcast (you need to skip first 5min to get to discussion in English):
https://www.youtube.com/watch?v=rahV0NaUZY4

The rules are pretty simple. Each month we need to analyze most recent performance (last 12months) of (large-cap) US equities (like S&P500), bonds (or out of market cash) and World wide markets (like MSCI World ex USA Index). Simplified version of this algo has only two steps:

1. Absolute momentum: Analyze (large-cap) US stocks performance over last 12 months. If US stocks performance is < 0, then put your capital in bonds.
2. Relative momentum: If US stocks are performing well (>0) over the last 12 months, then compare performance of US stocks VS global world market (ex. US) and put your money in the best performing index.

1+2 = “Dual momentum”

Interactive backtesting example using VFINX (US index 500) VGTSX (International index) and VFINX (Bonds, out of market asset)
https://www.portfoliovisualizer.com/test-market-timing-model?s=y&outOfMarketAsset=VBMFX&endYear=2019&movingAverageType=1&windowSize=12&timingModel=6&startYear=1998&assetsToHold=1&multipleTimingPeriods=false&outOfMarketAssetType=2&symbols=VFINX+VGTSX&singleAbsoluteMomentum=true&absoluteMomentumAsset=VFINX&rebalancePeriod=1&benchmark=VFINX

The original strategy is backtested since 1950:
http://www.optimalmomentum.com/gem_trackrecord.html

Has anybody any experience with this or something similar?

 

 

P.S. I’m currently less available, my PC time is limited 😉

stockpicking…this is not for beginners

hankeys wrote:

stockpicking…this is not for beginners

Just to be clear, the original ‘dual momentum’ idea is based on global indices, not on single ‘stocks picking’ method.

Basically we are comparing US economy with World-wide economy and using cash/bonds/treasury bills as our escape to ‘safety’, out of the market.

The only picking required is to find a proper global instrument like an ETF or something similar.

 

I have experience in stock performance ranking on a weekly basis based on price momentum.

However momentum trading, is a very wide topic, maybe this or something similar, could be adapted to other instruments = the second part of this discussion. I wonder if someone already used something similar in the past.

Wow, that backtest performance looks great! Any idea how such a thing could be incorporated in SQ? Could it be done in Algobuilder if not the main program?

Yes I’m also amazed by the simplicity vs performance ratio of this method and the fact it has a strong fundamental reason to be valid. The absolute momentum part could be easily programmed in the algowizard or even in strategy generation template, since it is only based on validating the performance of given asset over the last X months. (B.T.W. the work of Gary indicates, that the value of X should be somewhere between 6 and 12 months). He says he has also smart proprietary ways of selecting the proper X value…I do recommend to watch his presentation and listen to the podcast (= great stuff).

I do not know how to model the relative momentum in SQX, since it would require some cross-asset action, so I added this task sometime ago:

https://roadmap.strategyquant.com/tasks/sq4_4740

Please vote for it.

So basically, for point 2. I would simply compare every global Index by relative momentum or relative strength then put my money in the best performing one? I wonder what the result would be if a hedge was built in: so the top 20% would be bought and the bottom 20% would be sold.

As I understood, in this method the US economy is always dominant one, and should be validated first using its absolute momentum over X months. If the performance of US economy is <0, then there is no point of using relative momentum and compare the performance between US and world-wide, because US is still the biggest economy and when it goes down, it’s only a matter of time and world-economy will get into the problems as well.  Remember where 2007 crisis started? (Big short is a great movie…btw).

Thus, your top 20% idea looks very interesting, but I’m afraid that way you will always have an absolute momentum performance of >0, so you will be always ‘in the market’= so there will be no reason to escape to bonds/T-bills. However still very interesting to see it backtested. E.g.: top 20% S&P500 vs MSCI ex US.

 

 

 

 

 

So if you google around here and there you will find some traders that tried to improve this original idea by introducing “accelerated dual momentum”. Basically the idea is based on taking an average momentum of last 3, 6 and 12 months instead of only looking at one fixed lookback period of 12 months. Below a 150years backtest of the accelerated momentum trading strategy vs original dual momentum (GEM):

See also the following link:

https://www.quantopian.com/posts/accelerating-dual-momentum-150-year-backtest

On this website there is a source code (in Python) provided for this particular strategy together with an interactive backtester for those who want to play with the code.

It would be nice if someone could translate it to easy language for TS 😉

 

 

For this would be necessary to implement

1. Portfolio trading on more than 1  instrument  ; But this feature would be great to testing algos on a basket of markets  ( this is well implemented in quantshare, https://www.quantshare.com/ ).
2. Able to rank instruments ( rank function )

I m voting too.