It is true that as humans we tend to simplify the things around us and risk management and trading are certainly not the exception. When developing trading strategies we first come up with some worst case trading scenarios based on formal statistical methodologies (such as Monte Carlo simulations) and these first levels form the basis of our evaluation of a trading strategy’s expected statistical characteristics. However most traders never look again at these statistical evaluations, even after some time has passed and the system has already shown some live trading evolution. On today’s post I am going to talk about the issue of dynamic worst case scenarios and how values derived from a first statistical evaluation cannot be assumed to be the “never changing” worst case attributed on any trading strategy.

When you develop your trading systems – if you do so correctly – there is an in-depth statistical evaluation of strategies that helps you understand how the trading system works and what you might and might not expect going forward into live trading territory. The most important part of this first statistical evaluation is without a doubt the worst case scenarios which help us determine when a trading strategy has gone outside of what is considered “normal” under the historical analysis we’ve made. Usually these worst case scenarios are determined through some standardized Monte Carlo criteria which can include worst case draw downs, consecutive losses, months, years, weeks, chi-square tests and other statistical properties such as deviations from logarithmic growth curves.

You may think that once you get these worst case scenarios all you have to do is trade until you reach one one of them but this is a clear oversimplification of the actual nature of trading system evolution. When you consider how a trading system develops into the future it becomes obvious that trading statistics are not static and the worst case scenario could change to become better or worse than your previously deduced statistical values. What this implies is that after every trade all expected worst case scenario values change (although in a very small way) due to the inherent distortion these implies over the original statistics.

–

Although recalculating the worst case scenarios after every trade is not very convenient it is also true that after a considerable number of trades have passed the statistical character of the strategy might have changed enough as to consider a recalculation appropriate. This is most important whenever you have especially bad trading periods as these can lead to important modifications of a strategy which can lead to substantially worst cases than those that had been first deduces. For example if your strategy had an initial worst case scenario of 30% but you faced a 21% draw down period which lasted for 6 months, this distortion in statistics could now imply that thew new worst case scenario is superior to the previously defined one. It is foolish to trade a strategy based on a static worst case scenario neglecting how it changes with time as the trader must be aware and willing to accept all changes in worst case scenario that come as the strategy evolves.

Many of you may be thinking right now that this would inevitably end with strategies having worst case scenarios that would be close to total account loss and this is exactly the case since as strategies deteriorate so do their worst case scenario values. When a user runs a strategy he or she must then we aware of where the worst case scenarios fall as the strategy evolves so that the trader can decide whether or not he or she wishes to trade the strategy despite the deterioration in its statistical characteristics. So if a trader starts using a system with a worst case scenario of 30% and after a year of trading this scenario has evolved to 35% (without ever touching it) the user would then need to make a decision of whether or not he or she wishes to trade the system even towards this new draw down low threshold.

Note that this sheds light on a very interesting fact which is that strategies that have the least tendency to fail (strongest statistical characteristics) can go into draw down without increasing the worst case scenario significantly. This further reinforces the notion that the worst case MC derived draw down scenario to historical maximum draw down ratio is a very good proxy for system robustness as when strategies with maximum draw downs that are close to Monte Carlo derived ones are traded moves towards low draw downs cause very little deterioration in worst case scenarios. In the case of these strategies we could think of the historical profit and draw down values as very good proxies for future results while in the case of other strategies where moves into draw down deteriorate worst case scenarios it becomes important to do judicious and continuous evaluation of where they are going.

After reading the above you should be aware of the fact that trading system statistics are permanently evolving into the future and diligent evaluation of them through time is fundamental to the long term success of your trading efforts. Evaluating strategies and portfolios two or three times a year to determine new Monte Carlo targets according to live trading evolution is a very important aspect of quitting systems which you might not be willing to trade anymore early on. People who neglect to do this will generally be able to avoid deeper draw downs as if a trading strategy has a worst case scenario of 20% which evolves to 40% the user can decide to either half risk (to keep the previous scenario) or eliminate the strategy, even before it reaches the previous 20% worst case. As I have said it is very important to keep an eye on your systems and evaluate their risk levels and worst case scenarios as time evolves.

If you would like to learn more about automated trading and how you too can perform Monte Carlo simulations to learn more about your trading strategies please consider joining Asirikuy.com, a website filled with educational videos, trading systems, development and a sound, honest and transparent approach towards automated trading in general . I hope you enjoyed this article ! :o)

Interesting, Daniel the good thing is in 5 years time we will be able to build better more robust strategies because of 50% more historical data, the Monte Carlo worst case should stay constant for a longer period the more history your strategy is build/tested on if I understand this correctly…

Daniel,

A really interesting note about dunamic nature of the worst case.

Can you please tell how can I know that re-evaluating time came based on “enough trades” from statistical point of view? Evaluating according to time period does not actually mean that there were “enough trades”.

Maxim