There are many definitions of robustness but for clarity sake and in order to do some generalization by robustness here we assume the ability of a trading strategy to at least withstand and at best adapt to changes in the trading environment. And by trading environment we assume all factors that eventually affect the trading quality, and these factors can essentially be classified into three major important groups: infrastructure, liquidity and regulations.
If put a bit simpler, robustness assessment is the analysis of a trading system’s stability that aims to predict the possible reaction of the system to the changes in market conditions. This is perhaps the most complex domain of systematic trading, and yet its importance is hard to overestimate as otherwise we are not able to know if we need to stop exploiting a particular system when something went wrong, or it’s just a temporary drawdown and we simply need to wait a little till it recovers.
Let’s briefly see how changes in the market environment may cause changes in the behaviour of market participants and consequently price behaviour in general — and therefore affect the performance of many algorithmic trading strategies.A good illustration of how the three aforesaid factors affect the fx market is the rapid burst of new technologies around mid-2000’s: infrastructural changes like the development of ECN, STP and integrators reformatted the fx market allowing unprecedented access to liquidity.
Picture 1 shows how these new opportunities opened doors to high frequency strategies and in turn caused volatility to fade to historically low values. Many volatility-based strategies died in that period, and it could have been possible to avoid or at least minimize losses in this global process given the proper analysis of the infrastructural changes had been done in a timely manner.
Another lesson that we can learn from the same picture is that market processes that have somewhat similar appearance can be (and most of the time are) caused by different reasons. While the fade in volatility around early 2007 was caused by the major changes in infrastructure, the most recent historical lows in volatility were mostly caused by serious changes in market regulations that aimed to reorganize the whole structure of the interbank market with their flagship requirement of better transparency.
Along with the decreased volatility these new environment conditions affect general money flows redistributing liquidity between exchanges, traditional informational systems, ECNs and other trading venues. Proper analysis of the market environment changes and the assessment of robustness of a particular trading approach against these changes could have generated an early warning sign and helped surviving during the change in a relatively safe condition.
As with risk management, robustness assessment can be done in a number of ways. The simple examples considered above are only discretionary estimates that may work as the top level signs of the most major environmental changes. However in many cases it is possible to suggest certain quantitative metrics and therefore quantify the whole process of robustness assessment. This in turn allows to predict possible malfunctioning of a trading system depending on the environmental changes even on the research and development stage.
Picture 2 shows a parameter stability analysis of a very robust strategy made back in 2012: it’s clearly seen how slowly and linearly decreases performance with changes of the parameters, which is a sign that the strategy is indeed robust, and there’s no surprise that it continues to work in the changed recent environment. The key to the correct robustness assessment and therefore to adequate reactions to changes in the market environment is deep understanding of all factors that contribute to the environmental changes, their importance and mutual influence in various situations. In simpler words, it gives us answers to two crucial questions: why this particular strategy may work and under which conditions it will stop working.
Alex consults in automated trading systems design and proprietary robustness assessment methods. He can be reached at email@example.com. More information at http://edgesense.net