A Mechanical Portfolio: Does Mechanical Investing Work?

13th Jun '11 by siestainvestor

10 comments 15749 reads

There may be some merit in a mechanistic approach to portfolio adjustment for the following reasons:

- Set at suitable review intervals, trading frequency and related costs can be minimised

- The inertia effect of “falling in love” with stocks is avoided

- The system can allocate assets on a rational basis (asset allocation is the biggest determinant of performance)

- Emotional components (herd behaviour, fear and greed, reluctance to admit mistakes etc) which bedevil individual investors are minimised.

- The system can mandate sales, the hardest element of portfolio management

- The system can be engineered to be biased towards capital preservation

I set out below the structure of a simple system which attempts to tackle the issues above.

Elements

This system determines which of four diverse asset classes are held at any given time. Each asset class has a single proxy chosen (a) to be easily traded on small spreads and (b) to have a good financial credentials (balance sheet, size etc). The four asset classes are:

Developed market equity
Real estate
Emerging market equity
Hard commodities (gold, mined raw materials)

Clearly there could be more than four asset classes but these four seem to me to cover much of the waterfront and do not correlate much in normal market conditions.

Decision process

This system is based on a quarterly review period. The decision dates are the first trading Tuesday after each of 15th January, April, July and October. (The dates and frequency could vary but these suit me as I am usually not travelling then). The decision review answers the question “should this asset class be held for a further three months?”. If the answer is No then that portfolio element is switched to cash for the following three months.

Basis of Decision

The asset class is held for a further three months if the 50 day moving average price on the review date is above the 200 day moving average. Otherwise the asset is sold and cash is held until the next review date.
This is a simple “the trend is your friend” methodology that can be back-tested through the last five years - a period of considerable volatility. I have done this and the results are below.

Proxies

I have chosen single proxies for the four asset classes on the basis that I know the particular assets. My proxies…

Unlock with Email

Unlock with Google

Already have an account?
Login here

alano20

12th Jun '11

The table below shows the results of a back test of this system for a five year period (from July 2006) - Is this missing?

siestainvestor

13th Jun '11

The table was pasted in as a .tiff image. It may be that your browser can't parse it.

Thanks, got it now. Neither Firefox nor Google Crome could parse though.

I changed the table to a jpeg and it now parses in Firefox. Thanks.

Dave Brickell

Very interesting post. Although this particular strategy didn't perform so well vs. a straight buy and hold strategy, if you look at the rise of quant funds in the City, I think it shows the power of some of these semi-automated investing styles.

We're very interested in the whole area of algorithmic investing and the use of screening as a guide to improved investment decision-making (I am less keen on the term "mechanical investing" as I don't think it recognises the way different levels of judgement can and usually should be applied to the results as no algorithm is ever likely to be perfect given the constantly changing market environment).

Anyway, this is all very much in line with some of the development work we are doing for Stockopedia Premium which will be launching this summer, so hopefully you'll find that to be of interest when it comes into beta.

snickers

I'd like to know how robust this method is.. can you generate an average return for the last X years, or with a random start date? does that average tend towards a simple buy-and-hold return value?
also i'm not sure you escape the psychological traps: i presume you have to monitor the investments and consequently watch peaks and troughs come and go.. On the other hand, if you do succeed in disassociating from the process, you're possibly losing out on one neglected plus of shareholding: which is to identify those parts of your life where your analysis actually means something and has a sure footing... invest in yourself, Etc.

marben100

Unfortunately, to the best of my knowledge, I don't think any mechanical or semi-mechanical method has been shown to work consistently. I even tried a semi-mechanical method (based on voting by supposely knwledgeable investors) over a 4 year period on a portion of my capital. Whilst when I first examined and back-tested it, it appeared to produce superior returns, it failed miserably (and was the poorest performing part of my portfolio) over the period tested.

Though "quant funds" may be popular in certain quarters, have any produced consistent & not erratic returns? If there are any, I imagine their agorthims will constantly adapt to changing market conditions.

In my opinion and experience there is no substitute for hard work and serious research. If you don't have the time available, you are probably better off entrusting your capital to a skilled manager (or collection of fund managers), who spends all day, every day doing that. There are one or two who post here. ;0)

Cheers,

Mark

14th Jun '11

i spent the evening (face it kid, half the night as well..) trying a bit of momentum analysis. the beauty of computers is that you can run a program over and over with slightly different inputs. what i've done is, using monthly average prices for antofagasta, to generate a rolling moving average of various time periods, and use share price divergence from these to trigger buys and sells. it's a simple buy-on-the-dips, sell-on-the-peaks strategy. so in the following grids horizontal is the % divergence, vertical is the moving average time slice. zero is top left for both scales. the dot size shows the cumulative return. [i've not included the cost of trading or of missing dividends].

the main feature here, apart perhaps from incomprehensibility, is the big wedge at top right: this is data from the last 4 years and that wedge, i think, represents buying in march 2008 and selling today; ie waiting for a massive dip.. it also looks like there's an small benefit from using longer moving averages and demanding a bigger peak/dip to trigger a trade. the empty top right corner shows how being too greedy means you never buy at all. but is there another wedge in the top left? does this pattern show up using weekly data? here are 2 periods, from 2004-6 and 2009-now.

the older data shows you need momentum to make momentum trading work. preferably upward momentum. the recent data shows... i'd say no more than that a long horizon and non-itchy (but not totally insensitive) fingers gets better returns. if there is something going on on the left hand side it's a small effect... it's probably a bug in the code.

next, here's a weekly graph for a volatile stock: desire

with this one the volatility seems to skew better returns towards a long-term-moving-average/low-divergence tactic. but you could easily miss the fat dots.
hope that helps! i can't guarantee my maths: i may well have written code which confirms my current thinking. perhaps the best strategy is to wait for that generational dip, to invest only once and try to accumulate cash in anticipation; trade the economic cycle.
as marben says productive research must lie elsewhere: biographies of management etc, analyse the product. i like the approach of brontecapital: the man there identifies frauds using anomalies in the balance sheet and then has private investigators sniff round the company's premises. easier said than done of course.

Full marks for artistic merit snickers. It seems you have to set trigger parameters high and review cycle times long and use long moving averages to get into your wedge. This will require considerable patience.

I know mechanical investing was discussed at length in another place a few years ago and generally rubbished by those that tried it in practice, although I did have some success with a system based on Profit Warnings for a couple of years (I gave up in the end because it was hard work).

I don't disagree with marben when he says there is no substitute for research - particularly for finding undevalued assets to buy. But the hard bit is deciding when to sell specific assets and indeed when to take money off the table completely. Some routine that forces you to consider something that can be objectively measured (eg 50ma>200ma) on a regular (but not too regular!) basis seems to me to worth attempting. Another method I use is to see where the current valuation lies against the previous five year's PE ratios (easy using SharelockHolmes). (ANTO went from a PE of 8 to well over 40 for example. I baled out around 20 and missed a big bit of momentum but I slept nights).

As I listed in my "Reasons" in the main post, asset allocation seems to be everything. The real problem at present is sitting it out in cash earns you nothing - one shoud probably be in German sovereign debt, index linked bonds and maybe gold all of which are harder to buy and hold for the private investor. It is not a happy time. It looks like all the proxy assets are about to switch to Sell. Anglo Saxon government debt repayment is still being pushed onto future generations or will be worked out by debasing the currency. When Property shares dive under again it might indicate that the next (Government Balance Sheet) recession is upon us.

Maybe I should have titled the thread "Towards a system for triggering a Risk Off posture" as that for me is the Holy Grail. How do others achieve this? After all, if you avoid losses you will eventually make money!

Siesta

Full marks for artistic merit

ta.
yes, patience. graft all you can, index-link your savings, and bide your time. i'm torn between selling up non-dividend bearing shares for now until a China-triggered correction has passed [it didn't happen today anyway], and staying in the game, for the excitement, interest; and it makes me feel relatively more intelligent in my other activities.
Stock screens can be deceptive, i find. a good showing is often a flag for some kind of problem. kazakh miners etc always pop up. but it seems more scientific than dart chucking, to pick a company to investigate further. picking an intelligent guru is possibly a good stategy, and sticking with them. i feel sure Christopher Wood's Greed & Fear must be a mine of good sense, since i've never seen it leak onto the internet..

A Mechanical Portfolio: Does Mechanical Investing Work?

Elements

Decision process

Basis of Decision

Proxies

Unlock the rest of this article with a 14 day trial

About the Author

Siestainvestor

10 comments

Unlock the metrics for all these stocks and more...

Unlock the metrics for all these stocks and more...