By Jeff Kum | 21 November 2018
Recently I had attended few seminars promoting investment strategy using artificial intelligence. Due to curiosity, I have done some research about Algo trading vs AI and would like to share it here. For readers info, I’m not a computer genius nor super trader so the reader can give feedback or correct me if I am wrong in whatever info provided here.
This method of trading is referred to as a method of preset action of execution by a trader or a computer repeatedly over a period set. It starts with a set of rules and condition, which developed by experienced traders where it will be tested first in simulation to gauge the effectiveness of the program to make a profit in the market.
The experiment can be started as simple as below:
- Buy when the stock price is close above 50 moving average
- Sell when the stock price is close below 50 moving average
This simple instruction can code into a computer program using the past data movement of the stock price and put in the test simulation that we called as ‘backtest’. The test shall generate more than 100 trade or even 10,000 trades provided the stock price data history is long enough from the past to the present. With the generated test simulation, the traders will know how much profit/lost, risk/reward ratio, recovery factor and much other info that generate using this trading method in the past. The trader will assume that the program will create a similar performance in future and will use it to trade in real account if it is profitable and within the comfortable risk.
So far until now, I haven’t seen anyone using artificial intelligence to trade. Yes, I know some of the investment gurus, investment fund house or brokers claim that they already use it to trade the market, but I had a doubt on that. Let’s understand what A.I. is and how it works.
The AI starts with a programming system which mimics the human intellectual and they called it ‘neural network’. By the layman terms, everything starts with few points and the computer will start the first step to learn to identify ‘correct’ and ‘not correct’.
Example like a human child, if you want to teach an AI to learn how to differentiate a cat picture with rest of the image, show it a set few random pictures then you select the one with the cat to tell the computer that is a cat. The process can run many times until the machine can identify more accurately which image has a cat in it. The learning process can be unlimited hence the more complex the situation, the more it needs the support of super fast multi servers to store and to run it. That is how an AI learn, from experience.
The Different between AI & Algo Trading
|Artificial Intelligent||Algo Trading|
|It will set own rules and conditions based on learning experience.||Rules and conditions need to set by a human|
|Rules and conditions will change by the machine based on objective||Human will need to change the rules and requirement from time to time, to tune it to current market behaviour|
|Outcome shall be better in future due to the machine will learn and correct itself each time.||The result is often predictable and similar to past. But it will change when market behaviour change.|
To trade market using artificial intelligence, I think it will need some massive tech company and funding to support it like Goldman Sachs already start the AI project.
You will need an AI that can learn from multi and diverse info like what US President will speak about, when he will be talking, How his tone will impact the market, when is the storm coming and how it will change the market, etc..etc.. The cost will be huge; it will be like combining many AI into one super AI to trade the market successfully.