Amongst the people who develop mechanical trading strategies there are some who will utterly refuse to provide any simulation results with the excuse that the system’s trading will not be executed in the past and that only results in the live market will matter. Usually the people who put up this argument work with systems that cannot be simulated accurately or systems which have dramatically bad simulation results. It is very common for system developers to fall into this “denial state” and justify it by saying that the past is irrelevant since only the future is important for a system as that is the market that will be traded and not the market in 2005 or 2000. On today’s post I will explain why this view is absolutely wrong and why it leads to the development of strategies that are ultimately condemned to fail in the longer term. I will go through several aspects that severely complicate trading when using such strategies ultimately dooming them to long term extinction.
Yes, I think we have all experienced this. You are within a forum thread or you bought a commercial system and you ask the person in charge if long term simulations can be provided. Then you get answers that range from “backtesting is inaccurate and therefore meaningless” to “the past doesn’t matter”. It definitely sounds logical since the system won’t be trading in the past but on live market conditions so as well as it performs profitably under current market conditions the past can be ignored, right ?
There are many reasons why strategies are back-tested that go far beyond showing if the system was or was not profitable in the past. When you back test the objective is not only to develop a system that was able to succeed under past market conditions but to allow you to understand the system’s characteristics. Of course, the main characteristic you want to discover is the statistical edge of a strategy over the market. When you carry out simulations – carried out in a correct way – you can accurately tell if the system has a positive mathematical expectancy and an actual long term edge over the market. You can also tell if the market is able to adapt to changes in market conditions and what the expected long term profit targets are.
Another extremely important set of data you get when you carry out long term simulations is the draw down profile of the strategy, you can know how deep and long draw down periods have been in the past, something which allows you to effectively prepare to similar or even worse draw down periods in the future. Even though past conditions will not be repeated in the exact same way in the future, the fact that you know how your system behaves under extremely varied market conditions allows you to have some idea of what might happen in the future when the market starts to go through a long term volatility cycle.
–
–
Running a system with no long term simulations only focusing on how it performs under current live market conditions is an approach doomed to failure because you are absolutely ignorant about all the characteristics of your trading system. First – and most importantly – you don’t know if your system has any kind of edge under the current market and what you are doing (which involves some sort of past data analysis since you cannot know the future) only focuses on a very small amount of data which fails to tell you if you have or if you don’t have a long term statistical edge. Another point is that you completely ignore the draw down characteristics of the strategy, so you will not be able to know what draw down is predictable and what draw down is not.
Every time I have observed a system with the same excuse regarding the running of long term simulations it ends up with the exact same results. People run the strategy under live conditions and whenever a draw down periods starts they begin an endless chain of system modifications that eventually lead to heavy losses or even account wipe outs. The problem here is that there is absolutely no knowledge about the strategy and its long term standing against the market so modifications are implemented to avoid any draw down period without any notion about their long term effects or their adaptive qualities. In the many years I have been involved with mechanical trading strategies I have never seen such an approach used successfully since it is a mindless pursuit of an optimum system that performs perfectly under current market conditions with an absolute ignorance about the true statistical edge of the strategy and its long term characteristics.
Many times the creators of such strategies will be in denial until the strategy – through time – proves to be nonviable, something which could have been proved with long term simulations from the very beginning. So next time you hear such an excuse bear in mind that it is just that, an excuse. Anyone who wants to develop a reliable trading system that can handle changes in market conditions, has a high probability of succeeding in the future and a proved statistical edge (under extensive past market conditions) will run simulations for as long as they can since simulations of systems under past conditions are the way in which we can discover – for systems that can be simulated accurately – if success is a likely outcome in the future.
In the end I would rather trade a system which I know was profitable within the last ten years, a system which has clear long term draw down and profit characteristics I can evaluate and understand. If you would like to learn about automated trading and gain a true education in system design and evaluation please consider joining Asirikuy.com, a website filled with educational videos, trading systems, development and a sound, honest and transparent approach automated trading in general . I hope you enjoyed this article ! :o)
