Stephen Bland says mechs are long dead, but maybe they will be fashionable again at some point.
Going back to the earliest days of the Fool UK, and I've been around almost from the start, mechanical share-trading schemes were all the rage. It's hard for anyone who wasn't here then to appreciate just how popular they were. So much so that a newcomer could have been forgiven for having the impression that this was just about the only sensible way to trade shares.
Now, and for a long time, they have been all but dead judging by the absence of traffic on the two boards we have for this purpose. I'm not totally sure, though, whether mechs were discarded because mostly they didn't work in the end or just because there is always some new idea to tempt those with insufficient perseverance to stick with any method. The latter approach is just fashion, to which investment seems just as prone as many other walks of life.
Fashion-following investors are unlikely to profit much in the end from always changing strategies with the mood of the crowd. Better by far to find one that works and stick with it. The problem with that idea, though, is that you'll never know until it's probably too late whether you made the right choice or not.
Just about any strategy will have its bad patches and it can be hard to stick with it in such times, particularly if you see some other method that seems for the time being to work while you are doing poorly. But changing at that point means you will forever be doing so at a stage that will usually be too late to capture the good returns of your new approach, more like just when it too is about to hit its inevitable bad patch.
One method of establishing quality that I have repeatedly condemned is back testing. It became notorious with the popularity of mechs in those early Fool days when frequent new ideas were put forward by people who claimed that it worked in back testing. It did, but what they nearly always forgot is the huge unreliability of this because the market conditions of the back test they considered just did not extrapolate to the future.
Further, they were often very unscientific in the use of data mining, which became a pejorative term round here back then for seeking past patterns that fitted an idea. The scientific approach to a theory is to try and disprove it, not to seek confirmation which you'll always find due to the natural inclination of people to want to be proved right. Few though appreciate this I've found, with people regularly mistaking for the real thing, the apparent patterns that sometimes occur during certain periods in randomness.
Time for a revival?
I don't know if mechs are due for a revival right now, it's been a while, but if it's true that fashions tend to be circular then perhaps their return may be imminent. And if they do, expect more of the same. Data-mined back-tested results that misguided people use to claim their method works. Don't fall for it.
All of which makes me appear to be anti-mech, but in fact I'm not. I even designed one of my own back then after guessing that the Fool's official mech, which we called 'Beating the Footsie', was likely to be a poor performer and long before it actually turned out to be just that. I called mine the 'pyad26' but for many years now I've ceased to follow it. I'd be interested to hear from any readers who still do, whether the results are good or bad.
What attracted me to a mech then was not back testing under market conditions that likely are not of a very long-term nature but what I called investment logic. If a scheme possessed the likely characteristics of winning in the stock market as used by non-mech investors, then I believed it had a decent chance of winning. My chosen approach for scoring from shares was value. Consequently, a mech based on value, I figured, had a better than most chance than most other mechs of working.
The same two personal qualities would be required for a mech value investor as a non-mech one, namely extreme patience and strong stomach to ride out the high volatility that can occur. The difference between the two is principally in the sell decision, though there is some in the buy decision too.
What I mean is that a value player sells when sufficient value has been outed regardless of profit or loss but may not set precise criteria for this, judging each case on its merits. The mech version would set predetermined quantitative exit criteria for a sale which could be relative to a sector or market etc. or absolute, thereby excising emotion out of the decision.
Similarly when buying, qualitative factors would tend to be ignored by the mech player and instead all shares matching the purchase criteria would be bought. All of this means laying down strict rules for your mech then sticking to them and not being tempted to override them.
Some versions of value mechs can be very stripped down, concentrating at the barest level on just one criterion such as low P/TB or a couple such as Yield with a minimum market cap for example.
For the avoidance of doubt, note that a value mech based on yield has nothing to do with my High Yield Portfolio approach. I'm talking of trading strategies here, not the buy and hold style of HYPs which are about income. It is a fact, however, that the same shares could easily qualify for a value mech based on yield if it had a big-cap rule, as for an HYP.
So how would a very simple value single-criterion mech work? Easy. Let's say the buy rule is P/TB < 0.9. You buy all the shares which have this feature, hold until P/TB rises above a predetermined exit level then sell. I reckon this would work quite well over time for those inclined to mechs.
The same applies to any other single value test mech. With yield you would buy all yields over a certain level and then sell when they fall below your much lower exit level, same with P/E or whatever.
That's oversimplified and I'd advise more tests in some styles. P/TB might be okay on its own but with yield as one example, I'd introduce in addition a minimum cap filter so as to pick only from the FTSE 350, say. The reason is that dividends from larger caps are probably more reliable on balance than those from smaller companies.
There are a large number of other variations you could adopt such as more quantitative filters or a timed holding period. But I'm not so sure that time is an apposite test for a value mech because there is no reason to suppose that the value will out in the chosen time frame. The outing period is a very indefinite quantity in value.
Are mechs due a revival? If so or if you just fancy a mech, whatever its method, don't be bamboozled by back testing. Rather, look for the investment logic which is the most likely indicator of potential long-term performance. Also, know yourself that you won't become easily disillusioned when it fails to perform at times because I can tell you now that this will happen.
If you're interested in high-yield shares from the FTSE 100, download the free Motley Fool report "8 Shares Held By Britain's Super Investor". The identities of the UK's leading dividend stock-picker favourite blue chips are revealed in this updated report for 2013. To have it sent to your inbox immediately and completely free of charge, simply click here now.