Has a sea-change occurred this year in the time-frame to consider when selecting stocks?

26th Sep '18 by Nick Ray

14 comments 3328 reads

Something happened in June

Many of us running a rules-based portfolio have noticed a recent slump in performance since June which has also afflicted the NAPS portfolios which are well documented on this site. NAPS portfolios have performed well over the last few years, but this year something interesting has happened, as illustrated in the graph below.

This graph shows the 2018 NAPS and SNAPS portfolios (together with two rule-based portfolios called PF.256 and PF1024 which we will discuss in a moment.)

All portfolios perform in a similar way until about mid-June. But then the NAPS portfolio and PF.256 start to decline while PF.1024 continues to improve. When the SNAPS portfolio is created it tends to follow a similar path to NAPS and PF.256.

So obviously the question is:

How are the two PF portfolios constructed and what is the 'secret' ingredient in PF.1024?

The answer is that each PF portfolio selects the 20 stocks on January 1st 2018 which had the best Sharpe Ratio. But PF.256 computed the Sharpe Ratio over the previous 256 working days (about 1 year) and PF.1024 computed the Sharpe Ratio over 1024 working days (about 4 years).

The Sharpe Ratio can be thought of as a risk-adjusted momentum measure. Alternatively, it computes a value which is proportional to the probability that the stock's annual return will exceed a target rate (in this case 25% p.a.) Another way to look at it is that it is a "momentum plus quality" measure since the risk adjustment tends to penalise momentum which is not backed by quality. Or yet another view is that we are looking for stocks on the efficient frontier as defined by Modern Portfolio Theory.

How repeatable is this effect?

The initial conclusion might seem to be that it is better to compute the Sharpe Ratio over longer periods than one year. We can test this by sweeping the period over which the Sharpe Ratio is computed from 100 days to 1500 days and plotting the results.

The grey guideline shows the trend of the return as the number of days is increased. It looks as though there is a step change somewhere around 600 days. So do we have a quick and easy way to improve returns?...

Unfortunately there is a catch!…

Unlock with Email

Unlock with Google

Already have an account?
Login here

ISAallowance

26th Sep '18

Interesting article. The obvious question is what market are you seeing this in: UK or global? And if it's the UK, do you see the same thing in other markets?

Nick Ray

This is all based on UK market for stocks with Mkt Cap>£90M. (I don't have any data for other markets.)

timarr

I must admit that a big learning point for me is that what had seemed like a "rule" to me ("you only need to look back about 200-400 days") based on the last seven years performance data turns out to be a case of generalising from insufficient data.

There are lots of examples of long term historical trends over much greater periods than seven years failing. Take LTCM, for instance, which was undone by hitherto unseen market correlation. Although, in fact, it wasn't hitherto unseen, it was just that they didn't look far enough back in history.

But in general, there's not enough data to ever prove any specific model - all of stockmarket history simply constitutes one gigantic data point. You'd need to see the evolution of the same market over thousands of multiverses to have anything like reliable data. It's the butterfly effect, writ ever so large.

timarr

schober

So, what stocks were in the two PF portfolios?
What happened in 2017, 16 etc?

27th Sep '18

Just to clarify something. The two later graphs both show the result of running 150 portfolios of 20 stocks each. This was an exercise in looking for patterns in the data. So talking about exactly which stocks were involved does not really add to the point I wanted to make. However there are three simplified screens if you want to look at typical stocks meeting the 1y/3y/5y criteria.

iwright7

Nick,

Interesting article and the Sharpe Ratio is a metric I have never previously considered. I don't know if it has any relevance to your observation, but I see Goldman Sachs have a high US Sharpe Ratio/short lookback screen, which they suggest works best in periods of high volatility.

June - 2018:

Investors are facing mounting uncertainty and volatility in the stock market, driven by factors such as high equity valuations, escalating trade wars and the tightening of credit by the Federal Reserve. Goldman Sachs Group Inc. (GS) suggests that investors look to their High Sharpe Ratio basket of stocks with the highest prospective risk-adjusted returns in each sector, a strategy that has delivered significant outperformance versus the S&P 500 Index (SPX) since 1999.

Selection Methodology

After computing expected returns, as described above, Goldman also calculated the 6-month implied volatility percentage for each stock, based on their analysis of options pricing. They then picked the 50 stocks in the S&P 500 with the highest ratios of expected return to implied volatility, with the proviso that each industry sector would be represented. The median stock in the basket has a ratio of 1.1, versus 0.5 for the median stock in the S&P 500.

Outperformance 70% Of Time

Goldman indicates that its High Sharpe Ratio basket has outperformed the S&P 500 in 70% of the semi-annual periods since 1999. Its average excess return versus the S&P 500 was 334 basis points over the typical 6-month period, or about 675 basis points annually. Goldman notes that this strategy performs best during periods of high volatility.

https://www.investopedia.com/n...

Would be interested in your comments? Ian

HollandSmith

This is excellent research. I current use this on a 200 day range to identify stocks with strong recovery after big pull backs.

My question is - why have you limited the vol to 55%? Are you not cutting off some big winners by doing this?

Rusty2

I have some European shares, I have seen a similar pattern there although it started earlier than the UK. If you look at the guru screens where Europe has usually outperformed the UK in previous years, they also have generally underperformed the UK this year. I think NAPS is too small a sample to due many conclusions, I did one the same day as Ed picked his and half my selection was different. If you look at the Stocko performance figures for 90-100 QVM, the EU peeks on 22 Jan 2018, the UK 11 June 2018.

andrea34

It's a very detailed analysis and I have quite a simple answer... isn't it Brexit worries?

My question is - why have you limited the vol to 55%? Are you not cutting off some big winners by doing this?

Good catch! I put that in to try to simulate what is called a "geometric efficient frontier" which tends to droop for stocks with very high volatility. Obviously you can take it out if you prefer. But the thinking is that stocks with very high volatility and very high returns (especially over one year) are much less likely to repeat that behaviour going forwards.

After computing expected returns, as described above, Goldman also calculated the 6-month implied volatility percentage for each stock, based on their analysis of options pricing.

They have cleverly tried to work out the future expected return and future volatility. If you could do that accurately it would create a beautifully smooth portfolio. Very sophisticated, although I see they note that it can sometimes over-estimate the future return if a stock is lagging a bit.

I believe that more generally, optimising the SR of stocks or optimising the SR of the portfolio as a whole is becoming quite popular, especially in the U.S.

30th Oct '18

I thought I would follow up on this post because of "what happened next" as they say on TV sports quiz shows.

The original post was made at the end of September, which we can now see was just before a major drop in the markets.

In the original post I noticed that a portfolio made up from stocks selected with a Sharpe Ratio chosen by looking at the last 1024 working days was outperforming a portfolio where we only look back 256 working days. The suggestion being that maybe a longer time-frame tends to put more emphasis on consistency and reliability rather than agility.

But now we see another aspect of the different strategies for building the portfolios. The NAPS portfolio, although it is down, has taken a rather more gentle path down than either of the Sharpe Ratio portfolios which are basically momentum-driven and hence have a high beta.

However the 1024-day portfolio is still just about in positive territory for the year whereas the 256-day portfolio is now well down on the year. (However, last year a 256-day momentum-based portfolio would probably have gained well over +50% so partly what we are reacting to here is partly higher volaility of the 256-day portfolio.)

It is still too soon to see how this story ends but I think we may be learning a lesson in real-time in the difference between volatility and mean return. Over twenty years maybe these approaches would return similar gains, but the higher beta portfolios will have higher volatility and give you both big gains and big drawdowns along the way.

Jack Brumby

2nd Nov '18

This is great stuff! Really interesting because my natural inclination is to approach investing with a very different mindset.

I wanted the Sharpe for my RiskRating post but it was taking up too much time as it was.

What do you use for your charts and graphs?

3rd Nov '18

I use R for the analysis and the ggplot2 package for the charts and graphs.

Although R is rather long in the tooth now, it is an incredibly good tool for experiments and exploration. Much more useful than spreadsheets and with a style which has been honed by people who clearly were using the tool themselves to try to solve statistical problems.

Has a sea-change occurred this year in the time-frame to consider when selecting stocks?

Something happened in June

How repeatable is this effect?

Unlock the rest of this article with a 14 day trial

About the Author

Nick Ray

14 comments

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

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