In this chapter I would like to focus on one of the biggest myths and misinformation about algorithmic trading that exists among novice traders. I am in contact with lots of beginners who are focused on discretionary trading and it is amazing how often they demonise automated trading systems (ATS). Yet I dare say that they are totally unjustified. If I paraphrase these discretionary traders, their distrust in algorithmic trading is based on the opinion that:
«ATS can never replace human judgement and they have no human feelings and therefore they cannot work.»
I, on the contrary, see the whole thing quite differently: Human judgements as such may be very a dangerous phenomenon in exchange trading in the long term! It is because the trader starts to feel that he can see formations, repeating situation, etc. in markets. In short, he is convinced that can already predict what will happen in the future, because he «got a feel» for markets. But how can you measure feeling or judgement? It is indisputable that one of the prerequisites for success of a company oriented on their product´s or service´s quality or customer satisfaction is the decision making based on facts. In order to be able to define the facts we must be able to quantify and measure them. This means that, for example, that it is not enough to conclude that we have a quality product, but there also must be some metrics (e.g. the correct length and width of a certain car component) that confirms or refutes the hypothesis about the product´s quality by testing certain criteria. And if we want to be systematic in our trading, we must use the same approach to our trading systems. We must be able to confirm or refute the ATS´s performance by the use of predefined testing criteria, whether via a robustness tests or other analyses. The problem of discretionary traders is that robustness tests and other computationally-demanding processes which are virtually unavailable to them, as we will explain later. It is because discretionary traders have a fundamental disadvantage in these crucial aspects:
1. They cannot perform relevant backtesting
Try to prepare an Excel spreadsheet and record individual trades in it (by manually going through the market candle after candle, or bar after bar, for the given timeframe). Firstly, it will be an incredibly large amount of work and, secondly, you will never obtain a relevant backtest. It is scientifically proven that people tend to distort backtest results. Many local and foreign lecturers who are fierce advocates of discretionary trading recommend to worsen discretionary backtest results by 30-40%. In my opinion this is totally incomprehensible and ungraspable argument. Where did those 30 to 40% come from? Every person is so individual that everybody can perform the same backtest in a completely different way. In contrast, when you have a programmed algorithm that defines the precise entry and exit conditions of the trading system, you do not have any room for deviations. You can quickly and accurately see if your trading system may potentially have some statistical edge and if it is worth your time and effort to further analyse this system´s behaviour by various robustness tests.
2. They cannot perform robustness tests
Robustness tests are so computationally demanding processes that they are virtually beyond possibilities of discretionary traders. A shining example of the computational complexity is the Cluster Walk Forward Analysis that is available in TradeStation. Yet robustness tests are the essential prerequisite for verification of a trading system´s potential profitability. This concept is often misunderstood and fatally underestimated by discretionary or novice traders. Thanks to my profession I have the opportunity to meet with owners and employees of various hedge funds. These professionals quite strictly advocate the view that it is actually not possible to neglect algorithmic trading these days. And be sure that one of the main reasons is the possibility to perform a tremendous amount of robustness tests of individual ATS. Therefore, my advice to you is: Don´t be foolish and do not think that a manual backtest on a couple of months of historical data can be of any relevance. Remember against whom you trade. Everyone on the exchange who speculates on the underlying asset´s price increase or decrease are your opponents and if they consistently use better weapons than you, it is only a matter of time when you will be defeated in this relentless battle. Do not believe illusions and search what specific weapons in the form of trading systems´ robustness testing do your opponents use.
3. They have a huge disadvantage regarding proper execution of trades which relates to the negative psychology of trading
If you have never manually executed any exchange trade, you may not consider this argument important or justified enough. But try to manually open a position in any market and watch how you are getting into attractive profits and then suddenly end up in even higher open loss. Markets are very dynamic and they tend to unexpectedly and rapidly change. And believe me, it can be very frustrating to emotionally suffer during individual trades. In algorithmic trading and ATS trading this problem is completely irrelevant. A programmed algorithm executes trades for you and you actually do not have to even sit at your computer. As for me, I rented a virtual server close to the exchange so that my trades were executed as precisely and quickly as possible and at the desired price. Yet do not think that when you trade via ATS you will not deal with any negative psychological aspects. Within your long-term trading you will have to go through periods when your ATS will stagnate or have series of losing trades. The key is to have precisely defined and quantified these aspects:
- How long this losing period may be and
- what is the theoretically largest decline (maximum % drawdown).
Needless to say that without precise calculations, robustness tests, and Monte Carlo analyses you will not be able to obtain these important statistics about the nature of your ATS. And as I mentioned, the only way how to obtain these complex and challenging calculations is a defined and programmed algorithm that substantially accelerates historical backtests and the related analyses and gives us relevant results.
I tried to rebut the argument of many discretionary traders that the algorithmic trading cannot work because it is not capable of human judgement. I believe that this argument is unfounded and can be very dangerous for novice traders who cannot orient themselves enough in this issue. I am convinced that the trading should be approached as any other scientific field. In science, if you want to confirm or refute a hypothesis, you must be able to quantify and measure its subject-matter. In trading, our primary aim is to confirm or refute the hypothesis that our ATS has or has not a statistical edge. A mere manual backtest of several months of historical data is unfortunately not enough. We need to take many years of quality historical data and verify whether the ATS could have a statistical edge. If we see the profit potential, our work begins – we have to perform a lot of analyses and robustness tests of the trading system. The fact is that this area is beyond the capabilities of discretionary traders who backtest their trading systems manually. And maybe that’s why they tend to demonise algorithmic trading and claim that it does not work. One gets mostly scared by what one does not understand or cannot do. I perceive algorithmic trading as a great tool for passive capital appreciation by tens of percent. It can help you to regularly appreciate your savings regardless of the economic cycle.