Dancing with the Gorilla



Managing change in today’s world is like dancing with a gorilla. You don’t stop when you get tired. You stop when the gorilla gets tired! – Stephanie Thompson

Nowhere is this statement truer than in the financial markets, an ever changing world where we trade and earn our living. As soon as you think you understand the world, it changes.

Consider that in the last 12-months’ alone markets have been spooked by Brexit, migratory/immigration issues in Europe, Nene/Pravin-gate, Credit downgrades, “#FeesMustFall”, the US presidential “reality show” and now President Trump.

This year (2017) we will also see a large number of elections in Africa (13), Asia (10), Europe (11), Oceana (4) and South America (4) [1]. A potential increase of a nationalistic vote (especially in Europe - most notably France, Germany and Holland all have general elections this year) together with an unsure picture of what effect “Trumpenomics” might have on the global economy, could keep volatility very much alive.

Below the radar we see the quiet, but ever increasing impact of automation on all aspects of the global economy. Automating many aspects of “knowledge work” threatens to put many people out of work, especially in South Africa! [2]

It would, however, seem that the more we tamper with things, the more we lose control. George Soros coined the phrase: Reflexivity, which describes the self-reinforcing juxtaposition between market interference and market outcomes. “The participant’s views influence the course of events and the course of events influence the participant’s views” [3]. Therefore, by interacting with the market we influence or change the expected outcome and we continually fall short of our goals.

Considering the above, it might sound as if the market gives us no clues to take our guidance from, or rather that it is a completely random process, as some might contend. As a fund manager that uses hedging strategies, gearing, and a large amount of automation, these cues or hints of what the “Gorilla’s next steps” might be, are essential in allowing us to manage our risk; or control the levels of exposure we might want to take. This helps us to get ready for an acceleration in the dance or to break for a turn, all without actually leading the dance.

While we will not get into the issue of whether markets are random or not (for this you can read my colleague Stuart’s blog [4]) there are some tools that help us in guiding our exposures and directional preferences.

We will however say at the outset, that we don’t always get the dance right. Excuse the euphemisms, but “Dancing with a Gorilla”, means we take risks. Possibly more than the average fund manager and from time to time “the Gorilla steps on our toes”! That said, if we can get things right more times than we get them wrong, manage the size of the mistakes, and most importantly learn from our mistakes; we will outperform and add value to our portfolio over the long run.  Furthermore, with the automated dance we don’t have to get emotionally involved, but we can instead (hopefully) enjoy the dance as well while focusing on our next steps!



Over the past year we were all flummoxed by the runaway performance of the Rand vs almost all of the developed market currencies. This was however not so much a country specific story or a positive story around emerging markets, but rather a negative story around certain developed markets. To add perspective, the Russian Ruble and the Brazilian Real each appreciated in 2016 by 16% and 17.8% respectively vs the dollar and similarly against the Euro and the British pound (GBP). Furthermore, it might look like South Africa’s ALSI was a clear underperformer with respect to developed market indices, but the picture looks markedly different when we consider these global market indices in Rand terms (from the perspective of an SA investor). In Rand terms the ALSI was a clear winner, even though the returns were muted compared to previous years.  All thanks to the Rand.


The table above shows the impact that the Rand can have on a South African investor’s portfolio and the need for diversification (or not).

Will the music change?


While exchange rate prediction is one of the most difficult things to do we can consider the current state of our currency valuation relative to other currencies. The famous Big Mac Index [5] is one way of getting to grips with how over/undervalued our currency is. In short the index looks at comparative prices of (say a Mac Donald’s Big Mac hamburger) in different countries. The assumption is that a similar hamburger will capture the relative cost differences of labour and ingredients for the same product in various countries, thereby capturing the relative purchasing price differentials between countries. Below is a table for various product price differentials as on the 20th January 2017.


These anecdotal examples point to the fact that the rand still remains undervalued.

A more accurate and accepted way of calculating the relative over/or undervaluation of our Rand is to use Purchasing Power Parity or Inflation differentials between us (SA) and say the USA. The graph below shows that the rand is currently still around 55-60% undervalued, even though there has been substantial revaluation over the past year.

These calculations are however greatly affected by the date at which the calculation starts. Something else to bear in mind is that while the per capita income or even per capita GDP for countries differ vastly, differential balances of various economic factors will necessitate some of these over or under valued numbers for certain countries over the medium to long term, as in the case of South Africa. The Economist [6] offers a reworked version of the Big Mac Index that also takes into account these relative per capita income or GDP levels. Based on the histogram below it would seem that our currency exchange rate (R/$, while having improved by over 10% over the past year) still remains heavily undervalued (dare I say “under” sought after?) – but with possible room for further strength.

Capital Markets

So what about valuations for various markets? The typical metric used to value equities would be the Price Earnings Ratio (or PE), which has become one of the accepted metrics to measure whether a company or a market is relatively cheap or expensive. In our analysis we use the CAPE metric devised by Robert Shiller [7], which uses the price of a market divided by the average of the inflation adjusted earnings over the past 10 years. (i.e. the same PE but with the earnings adjusted for long term inflation) The current standard PE ratio is indicating that the SA market is getting a bit on the expensive side compared to its historic average.

However, considering the CAPE, we see that the market valuation has been running at a slightly higher values since the global meltdown of 2008, but that we are no-where near the overextended valuations of 2008. Whether you are using CAPE or Trailing PE, current levels would suggest that our market does not offer much for further multiple expansion. (This is even further reinforced by the “Rule of 21” which states that the trailing PE and the inflation rate should add up to around 21 to allow for any potential PE expansion)

What is potentially more insightful would be to consider how the market performed over the following 12 months, given a specific (exit) PE ratio today, as shown below. Given the higher valuations in 2016 and the graph below, it seems to be no wonder that equity returns were muted. Looking ahead and considering todays almost unchanged CAPE or slightly higher trailing PE’s we should also probably dampen our expectations for 2017 as well. (Note our reference and statement is for the All Share index and not for any specific sectors)

Higher valuations and lower returns were further reinforced by the fact that foreign portfolio flows were cumulatively negative in South Africa. In other-words foreigners were net cumulative sellers of both bonds and equities in South Africa.

Risk appetite

So what about risk appetite? A standard measure for global risk appetite is the Chicago Board of Exchange (CBOE) VIX (or Volatility) Index. This index shows the market's expectation of 30-day volatility. It is constructed using the implied volatilities of a wide range of S&P 500 index options. This volatility is meant to be forward looking and is calculated from both calls and puts, and is a widely used measure of market risk, often referred to as the "investor fear gauge." The reality is that the bottom has completely fallen out of this index and it currently trades at level as low as those last seen in 2014/2007 and 1995 before that. It would seem that risk appetite could possibly be driven by the “America First” euphoria in the USA.

It is interesting to note that the previous time the VIX had reached these low levels, it rebounded by upwards of 60% within 3 months of reaching the low. (It would seem like an opportune time for us to load up on volatility?)


Lastly, it has also been interesting to note the recovery of value based strategies over the past year and a bit. These strategies were decimated during 2014/15 and caused many a headache for investors and managers alike. More recently these strategies have recovered well, with many breathing a sigh of relief. Our own technology stack has also developed well and we now track these and a view proprietary factor strategies for inclusion in our portfolios in the near future.


New Blood

We are very pleased to have Ph.D candidate Ryan Sweke joining our team for an internship for the early part of 2017. Ryan has completed a B.Sc. degree in Applied Mathematics and Physics at UCT, followed by a M.Sc. and Ph.D. in Theoretical Physics as a member of the Center for Quantum Technology at UKZN. In particular Ryan's M.Sc. and Ph.D. research [8] has focused on developing algorithms for the simulation of physical systems on quantum computers. Motivated by the growing applicability of tools from the condensed matter and quantum information context to machine learning and AI, Ryan has recently begun to focus his attention on developing novel machine learning techniques inspired by this rapidly developing correspondence. Ryan’s focus in the short medium term will be to develop a better way for us to do feature selection and dimensionality reduction techniques which we will hopefully open source in the next few months to come.

Google Tensorflow

We have made some good progress in the development of our technology stack. Over the past 6 months we have switched our code base from R to Python/Cython. This has been partly necessitated by the ability to include/incorporate Google’s Tensorflow machine learning (ML) framework into our technology stack. We use the Keras python package as the interface to Tensorflow which also allows us access to Theano, another ML framework.

Microsoft Azure ML

We have also done some initial testing of the Microsoft AZURE ML studio. The studio allows us to test certain predictions strategies in an easy to use, graphical environment. In essence it allows us to simulate and fail very fast, thereby being able to test the viability of certain algos or their calibrations, very quickly. Its still early day, so more on this in the future.

New Algos

We have also done some in depth work on fundamental factors over the past 3 months and we are hoping to implement a proprietary factor model in the next month or two.  In January we also implemented a volatility strategy which shows quite a bit of promise in the long term, but which could have some larger drawdowns in the short term. This contributed to some of our drawdowns in January.

We are also working very hard to add additional predictive models that will help us with regime identification and to some degree market timing (the holy grail!). Again more on this when we have this finalized.

Assets Under Management

Our assets have steadily grown over the past year, even though we have made very little effort on the marketing front (all our efforts have been on the technology front to get this right and to improve our performance) We are now almost at the R40m mark and we still have around R50m of capacity left in our Class A & D fee classes that are at a reduced fee for early investors.


While our performance for 2016 was broadly in line with the JSE’s AllShare Index (-0.13% Class D & -0.65% Class A vs -0.08% on the ALSI), it is nowhere near where we would like it to be. It is however interesting to note that our exposure at all times was below 50% net and we have only at times had gearing of the portfolio to a max of 120% gross. Also take into account that we started the year with an 11% drawdown in January 2016 when our short positions were still not in play. In August with the fall out of Brexit, our fund was up 7,8%. Looking forward, we do believe that the number of strategies we have now, together with our technology stack, puts us on a strong footing for the future.

Thank you

We want to again thank our investors for their loyalty and patience with us. We appreciate that you have put your faith in us as a team and we hope that in 2017 we will do better than just perform in line with the market.  I again want to emphasize however that our strategy will be volatile at times and that investors should be aware of this. While our exposures have been low, we do take on risks that in the shorter term could see us in drawdown. As our fund grows and our number of strategies grow, we believe that the combination strategies should mitigate some of the shorter term drawdowns.

Thanks again for your interest and loyalty and we wish you all the best for 2017.

“….and now for the two-step!”

The NMRQL Team