Trend Followers recognise that our ability to predict outcomes in a complex adaptive system is a futile exercise. There is no enduring permanence to these dynamic systems that allow us to predict outcomes with any degree of certainty. What may appear to be predictable, or a permanent structural feature of a complex system turns out to be a temporary emergent phenomenon held together by the inter-relationships between a system’s agents. As the agents evolve and adapt, so too does the emergent fabric of that system, but not in a linear and controlled way, but in sweeping and extreme non-linear ways.
This dynamic feature of complex systems prevents us from ever quantifying any outcome such as expectations of profit or losses with preciseness. A Trend Follower accepts this fundamental feature of all complex systems and therefore drops any desire to predict or quantify future outcomes.
When it comes to future profits or losses, we have no opinion on the anticipated timing or magnitude of these events. We simply recognise that these events will happen in the future. We avoid the use of any statistical Gaussian metric that attempts to assign a degree of probability to such events. In fact, we are prepared to base our whole approach on a method that disregards the Gaussian predictive premise. Namely our approach is couched on a premise that the ‘real’ market is more likely to ‘deliver more extreme profits and more extreme losses’ than what we can ever hope to exactly quantify with traditional statistical descriptive methods. This feature of price extension during trending conditions, is what we base our entire ‘Trend Following’ approach on.
While the reader may feel that by disregarding standard statistical treatment, that we are left rudderless and exposed without a crystal ball to predict the future, we have a contrary view. We believe that the adoption of predictive methods in trading models themselves leaves the trader vulnerable to far greater risk than what is experienced by our non-predictive trend following approach.
An assumption of predictability relaxes the way we deal with future risk. By attaching a degree of confidence to a single future path, means that we are more likely to be blindsided by events that we never saw coming. We see this all the time in the TV reality show of “Survivor”. The confident ‘leader’ of the group will assume based on their predictive weighting of their perceptions of how they rate in their ‘support group’ that the risk at a particular Tribal Council is low….but then *Kaboom*….out comes the blindside, and the ‘Leader’ now becomes the victim. More often than not, the risk event (in this case the risk of ruin) was caused by the very limited view the victim possessed on the systems interrelated context. The victim assigned an incorrect weighting of risk to the Tribal Council Event based on their limited viewpoint of the risk that existed in the system. If the victim was able to map a risk landscape that united the minds of all participants, then the blindside would have been avoided….but such is the limited perspective that any single participant can afford in a complex system.
As Trend Followers, we believe there to be significant risk underpinning the assumptions of all predictive models. This is the risk associated with the inevitable error of predictive certainty. We refer to this risk as Model Risk and this risk is bound within a model’s assumptions made about future market behaviour which leaves the trader vulnerable to ‘unforeseen risk events’.
So, for example couched in the predictive assumptions of a Buy and Hold Investor in the US Equity market is that the Index in the long term will always reach new high watermarks. The Model itself does not have any risk mitigation measures embedded in the method that allow for any error in the predictive logic. The assumption is that you buy a passive Fund that mimics the Index and Hold it for the Long Term until you elect to retire wealthy from the market.
Now while post War history has certainly given substance to these predictive assumptions and that the method of Index construction itself provides a long bias to the underlying market data, the trading method itself holds extreme Model risk. This is the risk that the assumptions do not hold over the term of the participants investment term.
Trend Following Models however have a far wider appreciation of risk and adopt a method in their system design that mitigates the risk that arises when any assumption used in their method proves to be false.
So how do we as Trend Followers understand risk?
We view risk as the uncontrollable factors that reside in a trade event that exists between the trade entry and the trade exit. There are only two times during a trade transaction that we can exercise direct control. The entry point and the exit point. At all other times our trades become victim to what the market decides.
During the intermittent period between trade entry and trade exit, we are exposed to capital risk. While we can adopt techniques that transfer or conceal this risk (referred to as methods that warehouse or conceal risk), we can never actually eliminate this risk until we exit the trade.
The market data itself provides an historical context to understand how price has recently played out at the time of making a new entry decision. From recent price action we can define whether a market exhibits a trending condition and by using an Average True Range of price over a recent look-back we can define an optimal initial stop and an equal dollar risk bet for any of our trades that we decide to take at this point in any of our diversified markets. We also define at the entry point the price point in which we define ‘our trend to end’. In other words, we know our exit in advance before we enter our trade.
By configuring our system entry point, initial stop and trailing stop from the outset you can see that we have purpose built the boundary constraints for our system in which all future prices MUST lie within, for us to continue as players in this Trade Bet. Our system therefore defines the constraints within which future price must unfold for us to remain as trade participants.
Our bet size which is the dollar risk between our entry point and our initial stop is what we are prepared to commit to this trade event. We are prepared to risk all of it. While we recognise that our backtest suggests that we are more likely to risk only some of it as our method has optimally placed our initial stop based on recent historical volatility, we recognise that the future carries with it far greater risk than what is presented in the historic record….so we are prepared to accept a total loss on our single trade…..however we deliberately configure that loss to only be a small fraction of our finite realised equity.
It is the small bet size that reduces the material impact of any adverse risk event of a single trade on our total portfolio. In fact, it is this single reason that is perhaps the most important feature of a diversified trend following system. Many would say that correlations are important, but I would like to argue that diversification as a method to reduce the bet size is far more important than diversification as a method to reduce correlation across a portfolio. The reason I say this is that correlation is once again a moving feast in terms of complex adaptive systems and what may be correlated at one point in time might be uncorrelated in another. There is no causative guarantee to correlation. There is however a causative guarantee with co-integration which is a method that Trend Followers use….without possibly knowing that they use it…to force drawdown offsets into a portfolio using design principles….but that is perhaps for another Blog post.
What we can say with confidence however is that, if we spread our finite capital over as many liquid markets that appear to be unrelated, then we force a smaller bet size to each position of our portfolio if we make each bet size equal in relation to realised equity at the time of entry.
Now the small bet size is our ultimate method to reduce adverse risk events of any single trade event, but the reason for the insistence of the application of our initial stop is for a different purpose than what most of us think. An initial stop is not a method to guarantee risk protection from a single trade event, but it is a method that provides a risk release valve that removes risk from the portfolio. You know how we said that risk can only be released from a portfolio by a trade exit. Well, this is the mechanism we deploy to release risk from a portfolio. If we did not have an initial stop for each trade event, we could find that our portfolio builds significant warehoused risk over time that makes it far more vulnerable to future unpredictable risk events. Our initial stop ensures that no single return stream will materially impact the overall risk signature of the global portfolio. At any point in time our portfolios are effectively optimised to carry more ‘future’ risk as our risk release valves have already “de-risked” the remains of historical risk that resides in our portfolios.
Like our Initial Stop our Trailing stop is also not what it appears to be to the naive trader. It is not a risk mitigation mechanism though it may appear that way. The Trailing stop is our method of ensuring that we let profits always run. We need a method to close the trade event when the ‘trend is deemed to end’ and our trailing stop as opposed to a constraining profit target is our preferred method. By having a trailing stop, we allow for unbounded future profit potential when we are fortunate to be riding an outlier.
So to recap….our Trend Following Systems are the way we constrain the ways future price can move in the future to achieve our performance outcomes. A trend follower deliberately restricts the possible risk events that can unfold in an uncertain future through their design principles that are applied to all their trend following models.
As Trend Followers, we treat risk management seriously. We understand that predictive techniques significantly underestimate risk, and we deliberately deploy risk mitigation mechanisms in our system designs themselves that are specifically configured to address the far wider gambit of possible risks that are present in complex adaptive systems. Rather than attempting to predict all possible future risk events, we use risk mitigation mechanisms in our models to restrict the domain of possible risk states in which we are going to participate in the future.
Trade well and prosper
The ATS mob