Screen-Based Portfolio Approach

18th Aug '17 by BankOfChina

13 comments 6699 reads

Hi all - my 2 week trial subscription has just rolled into a one year membership, so pretty new to all this. Fantastic site though and great resource here.

So how to put this information and data into action?

My thinking so far is to create a portfolio (in my ISA) that will track either a screen or a rank. My thinking here is:

- The screens and ranks are outperforming the FTSE over a 3-5 year period. Therefore a better investment than my Vanguard funds.
- I am not a stockpicker, and I don't think I can beat the screens. The screens have been created based on backtesting or other significant research.
- I plan to rebalance quarterly or semi-annually

My plan now is to build a portfolio that will track the performance of a screen. My questions are:

- Is this a common approach, and what success have people had with it?
- What are the behavioral factors to look out for? It cannot be easy to sit back and let the screen pick your stocks.
- What screens are people copying?
- What has not worked well for you in this approach?

Thanks all. Any help appreciated.

BOC

Unlock with Email

Unlock with Google

Already have an account?
Login here

HumourMe

18th Aug '17

This is the process I went through:
I looked through all of the screens focusing on a minimum historic annual return and also 'draw down' in order to identify a list of candidate screens. I rejected all with too few average holdings. To the remainder I used a rule of thumb to nominally add average yield and subtract average spread, based on their current selections and picked the few 'highest'.
This gave me ideas for a starting point to build my own screen. If a share meets my screen I believe that the odds are that it should go up - so I wait for a new high. If this isn't achieved then it has gone sideways or down and I wanted to avoid it anyway.
Part of my screen is momentum based, so after buying, I generally place a 15-20% trailing stop loss, which suits me. After 6 months I re-balance those that still qualify, generally sell those that don't, and set new orders for any empty positions. Weekly if there are 'empty positions' I'll cancel/set new buy orders. In general I'm aiming to have between 10 and 20 positions.

Nick Ray

Have you browsed the Stockopedia University links (from front page) yet? Especially the Stockopedia webinars on the NAPS method.

Personally I have found that you have to find an investing style which matches your personality and which you understand. If you try to follow a screen which is just "black magic" you lose faith in it. It is the same with tips. Unless you have some criteria of your own to judge them you end up chasing every idea going with no quality control.

So having said that, here's a simple screen which filters on volatility-adjusted momentum and quality. Feel free to use and modify at your own risk - or ignore completely!

http://www.stockopedia.com/scr...

(Actually I don't know whether links to screens can be seen by others, so let me know if the link doesn't work.)

Howard Adams

19th Aug '17

Nick

Yes the link to your screen works. An interesting mix. All got nice graphs though. Good for you.

Regards
Howard

iwright7

Nick

I agree about finding an investing style that suits your personality. Indeed there is questionnaire in "Selecting Shares that Preform" that purports to direct you into different investing styles depending on your scoring. This particular quiz didn't help me much and I wonder if other Stocko readers have come across a personality questionnaire that point you a particular direction?

For myself I found that I had to start with a shotgun approach (many different Guru/Factors) and then narrow down the field into a style(s) that produced overall winners and that I was most comfortable to hold.

I like the output of the screen you suggest and hold 4 of 14 companies it selects, (BUR, GAW, TET and BVXP). Out of interest why have you used; %Price change 1Y >3.5 x 1 year Vol – It’s the 3.5 part that intrigues me? I have also noticed that for me 1Y Vol<35 seems to work well and if you drop from <50 you would only lose 2 companies, (which happen to be the most volatile). Ian

Howard - don't be too seduced by the "nice graphs". The rules around volatility are designed to select a nice graph (i.e. smooth upward movement with not too much reversal), but of course the future is not always like the past. It is quite common for a really nice graph to go bad after you buy the stock!

Ian - the personality thing is really difficult isn't it? I thought I was a trader by nature when I started. But I have gradually completely gone in the opposite direction. So the problem is that (in my case anyway) you probably don't really know what kind of style will suit you initially. And it can be quite expensive finding out!

On the screen, I am trying to capture the effect of maximising (% Price Chg 1y - 25)/(% Volatility 1y) [called the 'Sharpe Ratio'] but you can't do this calculation using Stocko screens. So instead I selected for (% Price Chg 1 y) > 25 and separately chose (% Price Chg 1y) > 3.5 * (% Volatility 1y) as a weak compromise. The 50 limit on (% Volatility 1y) is perhaps a bit arbitrary to be honest.

You can do better by relaxing the 3.5 to maybe 2, optionally change the 50 limit on volatility, and then copying the resulting table from Stocko into a spreadsheet and computing an extra column to find (% Price Chg 1y - 25)/(% Volatility 1y) explicitly and sorting by that column. You want the stocks with the biggest value in that column. (Personally I also drop any stocks with a Mkt Cap < £100M. I find enough stocks with the filter without taking on the extra risk of very small companies.)

To head off the next question: "why 25", well it is a rough target for the annual return you want to better. If you make it too aggressive it will tend to go for stocks which had a fabulous year they probably won't repeat and if you make it too small then you will get reliable but rather underwhelming stocks. However any value in the range 20-35 seems to work pretty well.

Valley Exile

Appreciate your generosity in sharing your screen, and then taking time to explain it.

However, I am interested by your reference to the Sharpe Ratio, as I have come across some strong criticism of this indicator as a measure of risk.

The main problem with the Sharpe ratio seems to be that it fails to distinguish between "upside" and "downside" volatility. Therefore, perversely for example - a strategy with large upward "jumps" in returns (which presumably every bear trader seeks) and no draw-downs, would be deemed a more risky strategy than an alternative strategy which had smaller fluctuations in both directions (i.e. including draw-downs) - but arrived at the same endpoint.

I guess the presumption that markets and share movements are also "normally distributed" would also concern those with a greater statistical background than me, but that's a different discussion I suspect.

In short, have you found a way to modify or augment the Sharpe indicator to mitigate these issues or do you not view these issues with the Sharpe ratio a significant problem to worry about?

It's an area where I have found my own approach (which is set out below) but no doubt everyone will have their own view. (btw I've just typed this all in again because the browser lost my first attempt!)

The Sharpe ratio occurs in Modern Portfolio Theory. It measures the angle of the line joining the risk-free rate to a given portfolio in a plot of return against volatility. As you move along this line it shows how the return and volatility change as you gradually move your wealth between cash and the portfolio of stocks. (If you extend the line beyond the portfolio the claim is that this reflects to return/volatility for a leveraged portfolio.)

The highest value of the Sharpe ratio occurs when you draw a line from the risk-free rate to a tangent of the "efficient frontier" (see link for a picture). The claim of MPT is that you cannot do better than that so if you want a higher return or a lower volatility you should change the mix of cash and the tangent portfolio to achieve it. (It is conjectured that the tangent portfolio is the market portfolio.)

Sometimes volatility is equated with "risk" and then people start to worry that volatility does not measure risk (which is true.) The most common modification is to replace the Sharpe Ratio with the Sortino Ratio and this forms the basis for Post Modern Portfolio Theory. If you are interested in this view of risk/volatility that is one direction to go.

However, I use a hybrid Sharpe/Sortino ratio (Sharpino?). I use a minimal acceptable return (MAR) as Sortino does, but I stick with standard deviation (like Sharpe) rather than use the so-called semi-deviation.

So my version goes like this: I want a 'guaranteed' minimum return R (with some specified low probability of failure), but all I have are historical mean return 'u' and historical volatility 's' as estimators of future mean return and future volatility.

If I knew the probability distribution, then I want u - k * s > R for some value of 'k' based on the probability of failure I am prepared to accept.

However if we solve for 'k' we get (u - R)/s > k. So even if we don't know the probability distribution, then we can still reasonably say that the best probability of success will come with the stocks or portfolios with the highest value of (u - R)/s which is just the Sharpe ratio but using R instead of the risk-free rate. (Just as in MPT, you should really compute 'u' and 's' for the portfolio as a whole rather than individual stocks. But that is beyond the screening abilities of Stocko at present.)

There are some further complications you can get into. The historical values of 'u' and 's' are not very accurate estimators of future 'u' and 's' so you need to make the methods 'robust' -- which means that the method should cope with errors in the estimators and still give broadly similar results. It is also slightly better to work with log(Share Price) rather than share price ratios (returns).

It is also true that using only mean and volatility means that we are ignorant of higher moments of the probability distribution. In particular, Taleb is always banging on about fat tails (high kurtosis or fourth moment). But in my experience the estimates of future 'u' and 's' are so inaccurate anyway that worrying about higher moments doesn't help very much computationally. You just need to be personally aware that any individual stock can go to zero at any time!

20th Aug '17

Thanks for the interesting response.................

Before some brief comments below, i have just noticed that one of the intended edits in my original response is missing, so to avoid confusion please substitute the following sentence........

"Therefore, perversely for example - a strategy with large upward "jumps" in returns (which presumably everyone but a bear trader seeks)…………"

For me:

I am still trying to work out my system, and which indicators to use - but realize there is no "golden bullet", so i need to find those mix of indicators which will provide enough volatility, skew & kurtosis that every successful trade / investment needs, at acceptable risk.

Most indicators appear to be based either implicitly or explicitly on a normally distributed probability view of the market (even when in theory they could be used with other distributions) so I just need to take this on board, as you suggest. (For example see https://www.spcforexcel.com/kn... )

Though we often focus on how we enter an investment or trade, it seems to me as least as much effort should be put into portfolio management and position exit rules in order to protect any gains once they have been made. I guess we all think we can avoid the Northern Rock and other market crash situations in "real time" - but only those who have survived these events in the past will have some evidence to back up their confidence.

Many indicators can work great with carefully selected historical data, so now i'm more interested in forward testing and so will be interested to see how your screener works in practice, and how I can develop something that works for me that doesn't provide spurious assurance.

Your link is right. The results are very sample dependent. And for stock returns don't add much info.

If you compute skewness (third moment) for stock returns you usually find that it is negative when the stock is trending downwards and positive when the stock is trending upwards - so roughly the same info that you get from comparing whether the stock is above or below an MA, and just like an MA the lag in detecting the change in direction depends on how many days you compute the skewness over.

Kurtosis (fourth moment) is even more amusing. It will be about 3 (the value for a "normal" distribution) almost all the time and then jump dramatically if something unusual happens such as a price crash. Great! It tells you that something unusual has happened just after it happens. You can of course compute the kurtosis over a much longer period but then it still isn't actually telling you anything you don't already know - namely that unusual price moves occur more often than a normal distribution would predict (but it still doesn't tell you when).

Your other comments suggest you are currently in a trading mindset. If your indicators use periods of less than about 200 days you are trying to exploit volatility. Once you get beyond 200 days the mean tends to begin to be more useful but at that point people also usually abandon indicators (and a trading approach) in favour of mean-variance optimisation. But it is all down to what works for the individual.

The Third Man

Hi BOC

If, as you say, you are not a stock picker and you don't think you can beat the screens my advice would to be to study the rationale behind NAPS and Stock Ranks. Nick Ray has already referred to it as being something to consider.

I have spent a good many years trying to pick stocks using a wide variety of methods with some degree of success. But for the last two years my portfolio has been almost entirely based on NAPS and the results have been significantly better.

Mechanical Bull

21st Aug '17

- Is this a common approach, and what success have people had with it?

I have been following a purely rules based approach since April 2013. For my UK stock portfolio I am currently getting a 33% annualised return - This includes dividends and all trading costs. For much of this period I have been holding a fair bit of cash in my various accounts during this period, so my "invested" performance would be even higher. For the last year I have been running an equivalent approach for Australia/NZ stocks, which has made a small loss - However, this has a fairly small number of stocks and my back testing suggests that I have been quite unlucky.

- What are the behavioral factors to look out for? It cannot be easy to sit back and let the screen pick your stocks.

I don't find it at all difficult to let my system pick my stocks. I recognise that I am a terrible stock picker whereas my system is based on loads of academic literature, most of it conveniently put together by Stockopedia – I trust it much more than my own judgement. Highly skilled analysts of the calibre of someone like Paul Scott may be able to use screens to narrow down their stocks to a selection, perform additional analysis and improve their returns. However, if you are no better than an average stock picker, this approach will not help you. Indeed, there is some evidence that stocks that seem unattractive are the ones that actually go on to outperform. This is why rules based investing is great for useless stock pickers.

The behavioural aspect that I find most difficult is suppressing the urge to be doing something. I cannot stop myself checking my portfolio every day (even multiple times per day). This does not achieve anything, but I simply cannot help it. Therefore, although there is no real advantage in reviewing a rules based portfolio any less than quarterly, I review my portfolio every month. One thing I have learnt to keep my fiddling to a minimum. If you are going to fiddle the best thing you can do is to cut your losses and reinvest in something else. It is also very important to run your winners.

Another important factor is to be is to be aware that no rules based strategy will work all the time. My progress over the past four years looks like a staircase - periods of big gains followed by months of not much happening. Paradoxically, the main reason why good strategies continue to work over the long term is that they stop working some of the time. People then start to give up which then helps creates the conditions for the strategy to start working again. If a strategy always worked, then the market would recognise it and arbitrage it away. There is no substitute for patience.

- What screens are people copying?

The main screen I use currently is trade secret but what I will say is that is leans heavily on Momentum. Academic research shows price momentum to be the most important single factor in explaining out performance. If you want to read more about this I can strongly recommend “Your Complete Guide to Factor Investing” by Andrew Berkin and Larry Swedroe: http://amzn.to/2uY1Rh2 . This explains not just what factors work, but also why. A lot of these reasons are behavioural and getting insight into these is really valuable.

The importance of momentum is backed up by my own back-testing (I use the Sharelock Holmes backtester) - My best performing screen has momentum as the core ingredient with just a couple of other filters. This screen has outperformed the FTSE100 on average by 17% p.a. over the last 60 quarters and has beaten the FTSE 80% of the time. The probability of this being down to luck is over 600,000 to 1 so these are extremely favourable odds.

My general advice is to not speculate about what might work but instead follow documented strategies that are proven to work or even better strategies that work that you have discovered yourself.

- What has not worked well for you in this approach?

I am repeating myself here, but my biggest mistake has been not cutting losing positions quickly enough and then getting out of winning positions too quickly. It may seem obvious but you not only need a good strategy for buying stocks but you also need one for selling them. My initial sell strategy was way too sensitive which pushed me out of winning positions and produced excessive churn. Even after four years, I do not quite have my selling strategy completely nailed down. However, broadly the rules are:

• Sell on a 15% drop
• Sell on a profit warning
• Sell if the StockRank goes below 75 and the Momentum Rank is under 90

I am not sure whether this is a mistake or not, but one important reason that I have not made a lot more money over the past four years has been my pessimism around the wider economy my risk aversion and large cash holdings. Several times I have been convinced that another financial Armageddon was just around the corner, which of course has not happened … at least, not yet.

Hi MechanicalBull

Thanks for an excellent and informative post.

To add to your points about use of the Stock Ranks and rule based strategies.

On 29/11/17 I constructed a pure NAPS based portfolio. I was curious to examine how the Stock Ranks would perform if I selected stocks from my newly extended subscriptions beyond my UK only cover, now into Europe and US cover.

I elected to select 30 stocks.One from each market UK, Europe and US and one from each of the ten Sectors.

The one tweak to this I made was to only go for the three highest ranking stocks. Thus in some cases I had one from each market, but in others three from one market.

Since then I have not touched the portfolio. Over this period (8.5 months approx) the FTSE All share is up 8.69%. This UK, Europe & US portfolio is up 16.13%.

Now, this does not really come close to the 33% annualised returns you mention. But, it can illustrate another important point you have offered. Selling falling stocks and holding winning stocks.

For example, (today) there are 5 holdings with losses greater than 25%. The worst being 52.7%. If I had intervened as would be my normal practice and sold with drops of greater than 15% to 20% (taking into account context with a detailed scrutiny), then the overall return would have been higher than 16.13%.

But also, there are 5 holdings with gains greater than 50%. The highest being 192.2%. Now, I would have remained comfortable with a gain of 50% to 80%, but without following the - don't touch anything - rules as I did, I'm not sure I would not have top sliced this winner once it reached > 100%.

So, in short, the rule based approach has worked quite successfully across the three markets, which was my primary test.

Importantly, it helped me run winners.

But also, I am now convinced that a pure rule based approach needs to be supported by at least a re-balancing strategy (often encouraged by Ed, Roland Head, the literature and other SP subscribers), but probably also a strict exit strategy. As you point out in your piece above.

Regards
Howard

Hi Howard
I totally agree with the need for a strict exit strategy and, in my view, a rule based investment system encourages that. I buy based on the rules so it seems be obvious to sell on the rules! Where you set the stop is a matter of opinion.

Screen-Based Portfolio Approach

Unlock the rest of this article with a 14 day trial

About the Author

BankOfChina

13 comments

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

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