Investing as a profession requires a degree of prediction. And yet, before considering your ability to predict an outcome, it is useful to first think about whether the outcome of a system is inherently predictable. Some systems, i believe, are inherently so complex and dynamic that making predictions of their outcomes with a reasonable degree of certainty is not possible.
It isn’t about the degree of expertise required or available – certain events may be predicted correctly by a few experts, but can the same experts repeat their performance over a period that is long enough to generate returns consistently?
Science has a term for these systems and refers to them as Complex Adaptive Systems (‘CAS’). The systems discussed in the following paragraphs exhibit some of the following characteristics[1]:
- Distributed Control – No single centralized control mechanism
- Co-evolution – elements in the system change based on their interaction with each other
- Connectivity – a decision or action by one part within a system will influence all other related parts but not in any uniform manner
- Sensitive dependence on initial conditions – Changes in the input characteristics or rules are not correlated in a linear fashion with outcomes
- Emergent order – From the interaction of the individual agents arises a global property or pattern, something that could not have been predicted from understanding each agent individually
Survival of the ‘Expert’
My family enjoys horseracing and over the years I have spent a fair degree of time at the tracks. Like other sports, horse racing has layers of experts who specialize in every step of the process. The breeders, trainers, jockeys, managers and book makers are passionate about the sport and often come from long family lines in the profession.
Yet, despite the presence of experts, outcomes in horse racing are notoriously hard to predict. Bookmakers predict the winner of a race only 35% of the time on average in most racing centres around the world. This data is anecdotal of course, but it is in line with my own experiences betting on and losing money at the races.
Confirmation bias, led by bookmakers, is rampant and is part of the reason that unexpected outcomes are in fact so common.
Even with such a low success rate, the experts at the track continue to make their predictions and the industry continues to function (and thrive, depending on the location). It is interesting that the experts don’t lose money when they get their predictions wrong – it’s the owners and punters who do. Naturally, the survival rate of the experts is far higher than that of the owners and punters.
What compounds the problem is the binary nature of the returns for the punter – you either pick the winner or you lose all your money. Do the odds on offer compensate for the CAS at play combined with the binary outcomes? I don’t believe they do.
Rule #1 – Don’t lose money
Investing, in part, is a negative art – what you choose not to do can be as important as what you choose to. By staying away from inherently unpredictable situations, you make fewer mistakes and hence increase the chances of your survival, which in turn is a necessary precondition to success.
Several such situations present themselves to investors/ traders everyday – elections and weather predictions are examples of non-market events that impact the markets significantly.
Now, expertise is readily at hand to make predictions in these events. However, it is important to step back and wonder if the system is a CAS and hence if the ‘expertise’ is useful.
Often wrong, never in doubt
The weather is famously hard to predict and is a straightforward example of a CAS. Yet, predictions of the extent, spatial distribution and volume of Monsoon rains are tracked minutely by traders and investors all over the world. To be fair, the monsoon rains have a tremendous impact on the economy of India. However, the predictions of the competing weather agencies are never consistently accurate and so add no value ex-ante. It would be tremendously useful if it could be done, yet, we must recognize the fact that it cannot be done.
Also take for example trades that are based on predicting the result of an election. There is no shortage of punditry during an election, yet, as the last few years have shown us – the ‘experts’ have gotten it wrong repeatedly. What issues are voters basing their decisions on? Do endorsements make a difference? How do voters react to debates? Is there such a thing as ‘winning’ a debate? Even data based exit polls have a terrible track record of predicting eventual winners.
Experts can present (mostly) rational frameworks upon which voters should base their decisions. Yet, the decisions of millions of voters are fundamentally unknowable because we do not know how people decide on emotive issues. There is of course a rush to explain any outcome (especially the unexpected ones), but the exercise is fundamentally flawed because the underlying system does not work within the boundaries the experts define. You can justify any outcome, but to be able to understand it and apply it to the next election is another task altogether.
Price movements of currency pairs are a CAS and present an incredible challenge to economists and traders alike. The variables that go into determining the prices of a currency pair seem to be finite and predictable (to a degree), yet the final outcomes have confounded most analysts. Interest rates and inflation should be the primary drivers, but political risks, central bank action and the reflexivity of flows makes price movements hard to predict. Just pick up a report from a major bank in the last two years and you will see the flip-flops that have been made on major currency pairs such as the EUR/USD and USD/JPY.
Yet, Banks and Brokerages make these predictions because they must – they are the supposed experts and cannot possibly admit their inability to predict prices of an asset class they encourage their clients to trade.
Systems that depend significantly on the outcomes of a CAS are also prone to random outcomes – consider a global conglomerate that sells and buys goods in dozens of currencies. Its business is complex, but by itself is not a CAS. However, it’s hedging of currencies will eventually throw up a surprise – because the company is required to take long term calls on inherently unpredictable outcomes.
Binary Outcomes – It is better to be directionally right than to be specifically wrong.
As exemplified in the case of horse racing, the risk presented by a CAS may be exacerbated if the payoff is binary.
In my view, binary payoff scenarios add a layer of risk that is seldom compensated for. Payoffs are adjusted but at a certain point you must ask – does this payoff compensate me for the chance that I may lose a substantial portion or all my capital? (somewhat like high PEx ‘beat and raise’ stocks – you have to ask what margin of safety you have in case of a miss)
However, markets sometimes offer a way to mitigate a binary prediction. Instead of predicting the exact level of a currency pair one year from now, it is possible to step back and say that a currency pair should be higher or lower than its current levels, in a certain amount of time.
Instead of making a trade based on the whether a Central Bank will raise interest rates in the next policy meet, it is possible to take a view that in the next year, interest rates should be higher than current levels and in a range.
Is time a friend or enemy?
The impact of time in this context is a bit difficult to pin down because it plays different roles in different systems. The weather is not predictable, but climate is. What is the key difference in the prediction? Time.
For currency pairs, time takes on an even more complex role. Depending on the weighing of variables at play, longer time horizons may mitigate risk but sometimes it has the opposite effect because the variables may change in an unexpected manner (central banks may withdraw special measures, war may break out).
But, broadly, time can be a tool to mitigate risk. Of course, buying time costs a lot of money, but it is worth exploring if paying for time value can help lower risk.
Can you cut through the clutter?
It is possible to stipulate that a system is not complex per se, and that the complexity is only in the analyst’s mind. This requires a sceptical approach and testing through some patient prediction-outcome analysis. Cutting through a Gordian knot with a stroke of genius is alluring, but not something I would count on doing.
Conclusion – betting on the roll of a dice?
To sum up, I ask you to consider betting on the roll of a dice. Now, you may know a lot about gravity, air resistance, and friction of the surface, but assume your chance of being right is only 1/6. To start with neutral expected value, the odds on offer have to be at least 6x.
Ordinarily, I would ask you to stay away from the bet because the outcome is unpredictable and binary. However, if you were to consider the bet, you should demand odds far greater than 6x and the opportunity to place the bet multiple times.
But, before calculating the odds or adjusting the payoff – you need to recognize that all the experts in the world cannot determine the outcome of the roll of the dice.
Vishal Gupta
[1] http://web.mit.edu/esd.83/www/notebook/Complex%20Adaptive%20Systems.pdf by Serena Chan
Directionally Speaking 