Low-Risk investment strategy ?

Jegersmart

Hi Alicia

Thanks for update, although as I have said all along - this would typically not be a "low risk" strategy or system for most people.

J

grey_gym_sock

there is a danger with data mining, of coming up with the best trading strategy for the available data, when it may not be as good in different trading conditions. which could be a reason to keep it (relatively) simple.

personally, i'd like to have a theoretical reason for each part of the rules, as well as support from back-testing. though this is certainly not a topic on which you should listen to me (as i may have said before).

bowlhead99

aliciathyme wrote: »

Following further modelling I have introduced some changes to the selling procedure in order to refine the stoploss policy and introduce a Take Profit scheme.

In my view, such a good performance justifies the "Low Risk" epithet.

Performance is simply the return you got from deploying the strategy over a certain timeframe. If you are allowed to go back and change the purchase and sale decisions to make invesment choices which work out well for a given set of data (tweaking when to take profits and when to take losses), then the performance can obviously be quite good.

Risk is the danger of things not working out how you would like. The idea that a strong performance from your strategy in some periods means that the risk of not getting a strong performance in other periods is low, seems fundamentally flawed. You have different annualised performances in each of your ISA years which demonstrates that results depend on market conditions. And if you only hold a few stocks at a time, you're more exposed to risk of ruin from a couple of them going very badly (compared to the 4400 funds you compared against, which have broad diversified portfolios).

Operating this new selling policy is more hands-on than the previous simple 40% stoploss, but the results may justify the complication.

grey_gym_sock wrote: »

there is a danger with data mining, of coming up with the best trading strategy for the available data, when it may not be as good in different trading conditions. which could be a reason to keep it (relatively) simple.

personally, i'd like to have a theoretical reason for each part of the rules, as well as support from back-testing. though this is certainly not a topic on which you should listen to me (as i may have said before).

Previously the model was 'sell to buy'. Purchase decisions were driven by technical analysis of momentum of share prices in the index, but the timing and selection of stocks to sell came not by the relative desirability of the holdings or their rate of return but the need to raise the maximum cash to buy a new stock appearing on the buy list.

The stop loss was quite far away and there was no performance limit, so essentially there wasn't much interference in the strategy from from how well a particular holding was doing. No exit was sought until cash was 'needed' and then the selection of which one to exit was just whichever one had the highest absolute amount of pounds profit, disregarding time held, any technical analysis, rate or direction of current price movements, changes in the business, its market, our economy etc. etc.

Now the model has been tweaked to have tighter stops and limits so the winners can't run as long and losses are more frequent but smaller. If all stocks generally 'revert to mean' performance, (which is broadly true in the long long term), there is some logic in saying if something is up over 33% and then starts to fall back, maybe it is time to exit and seek something that has been performing less well. If this is the logic then selling when something is down 33% would seem counterintuitive, except if it's down because it's a dog, which is rather harder to tell.

If you can generally improve the results by selling anything that falls 10% within the first couple of weeks, that's great, but many would say that's too tight. Obviously you aren't continuing the 10% trailing stop after that first couple of weeks so perhaps the reason it works out is that it is tied into the stock purchase trigger process which was looking at recent technical charting movements. If so, there wouldn't seem to be much point anyone else using it with a different stock purchase criteria.

And looking at the historic data and deciding that a 2-week 10% stop is better than a 2.8 week 11.38426% stop, for a given set of data, does not in my mind mean that this will produce a lowest risk or highest return portfolio for the future data that we haven't seen yet.

To be honest, there are a lot of stock picking schemes which have a pseudo-scientific approach, where individual components of them have plausible reasons to believe in them and are 'proven' to have worked for an individual or a historic data set. We can say it's not low risk and the results involve luck, but it seems unfortunately the only way to provide your own contradictory data is to run it for the next 10 years and see if you made or lost a fortune, which I'm not really inclinded to do.

Atlanticus

First of all, are you sure about 332% net profit in 5 years ? because it is hard to get this much of profit within a short duration of 5 years. I would suggest you to do a research before you go for this strategy.

Thank you
Jack

zghbbo

I tried to replicate the strategy then later discovered that the system consists of two processes.

1- The anomaly process.
2- The trend rising process.

While number 2 has been fully covered (both for buy and sell), and I managed to try it out.

However number 1 is still unclear how to test it. You might argue that number 2 is enough and will generate % profit less than the whole strategy.

However I don't think using part of the system can prove the whole system even if it is considered as enough to use.

Aegis

aliciathyme wrote: »

In my view, such a good performance justifies the "Low Risk" epithet.

I've never seen risk measured directly as a magnitude of returns in such a way. Generally speaking it's an accepted fact that the higher the potential return, the more risky the investment.

The fact that you have arbitrarily adjusted your model parameters for better returns over the last 5 years doesn't mean it will achieve better results post-change over the next 5 years. In fact, the introduction of rules based solely on what they accomplished from 2008 to 2013 is likely to have a very different outcome over the next few years.

Please don't think your system is low risk - it's a concentrated equity portfolio based on technical analysis of volatile stocks. It cannot possible be low risk - it's not diversified enough either in terms of sector or asset class.

I wish you the best of luck with it, but it cannot be called a low risk strategy based solely on what you think the likely upside is.

Jegersmart

zghbbo wrote: »

I tried to replicate the strategy then later discovered that the system consists of two processes.

1- The anomaly process.
2- The trend rising process.

While number 2 has been fully covered (both for buy and sell), and I managed to try it out.

However number 1 is still unclear how to test it. You might argue that number 2 is enough and will generate % profit less than the whole strategy.

However I don't think using part of the system can prove the whole system even if it is considered as enough to use.

Yes, this what I have said from the start. The trend process is clear, but as there are other elements involved it is impossible to know what effect those may have......

J

Glen_Clark

As Always, I think it depends on your own financial situation.
Obviously you shouldn't pay £5 for a share if you think its not worth it.
So it makes no sense to sell it just because it has dropped to £4.
Unless you cannot afford to risk losing that £4 - however small the risk may be.

aliciathyme

Aegis wrote: »

The fact that you have arbitrarily adjusted your model parameters for better returns over the last 5 years doesn't mean it will achieve better results post-change over the next 5 years. In fact, the introduction of rules based solely on what they accomplished from 2008 to 2013 is likely to have a very different outcome over the next few years.

All the results I have posted have been the result of rigorous modelling of 5-year periods from a FTSE 100 constituents dataset of daily prices over the last 11 years. The data covers periods of rising, falling and static markets, and I have developed two specific methods to select potential buy targets - the Rising Trend process for selecting an equity for which the share price shows a sustained rising trend, and an Anomaly process which looks for exceptionally volatile shares and works best when the price of an equity is falling.

However, the processes are not geared to the general market rise or fall as such as they work on individual equity prices and these may or may not follow the market.

So, the results of my model for the last 5 years are not based on an arbitrary set of parameters, but parameters which have been carefully selected to produce the best results over the last 10 years.

These are the results of the model options for the current 5 year period to 15 May 2013:

Original Scheme: 40% Stop Loss (no take profit or early sale):
Rising Trend process: Net Profit = 116%
Anomaly Process: Net Profit = 32%
Both Processors: Net Profit = 132%

Revised Scheme: 33% Stop Loss, early sale on Stop Loss 10% within first 2 weeks, Take Profit at 10% fall after at least 33% profit
Rising Trend process: Net Profit = 233%
Anomaly Process: Net Loss = 3.4%
Both Processors: Net Profit = 258%

The point of modelling is to permit continual development of the processing options as new data becomes available, always back testing over a long period (currently 11 years) to confirm any improvements.

As you can see here, operating solely with the Rising trend processor (which I have previously explained in detail) gives exceptional results for the current period (the best of 4400 commercial funds is currently Fidelity UK Smaller Companies at 188% return). Of course there are no guarantees for future performance, but the results have been consistently good over the last 10 years..........................

Alicia

grizzly1911

aliciathyme wrote: »

As you can see here, operating solely with the Rising trend processor (which I have previously explained in detail) gives exceptional results for the current period (the best of 4400 commercial funds is currently Fidelity UK Smaller Companies at 188% return). Of course there are no guarantees for future performance, but the results have been consistently good over the last 10 years..........................

Alicia

Probably why they have currently closed the fund to new lump sum investments.

Any chance of the top ten;)