What is AI and how can it help my trading?

AI or artificial intelligence is defined loosely as machines demonstrating a level of intelligence. This is rather a broad term and can mean many different things depending on the context. It could range from a smart washing machine to a chess program that can beat a grand master.

Trained systems versus hard wired systems

There are two types of AI systems; those that learn by example and those that are hard wired. A hard wired system has a set of rules encoded into it. These rules may be complex enough that the system seems to be intelligent. These AI systems tend to be useful for very well defined tasks where the range of decisions is limited.

Machine learning systems are those which learn by example. With machine learning, the system is initially fed with training data, in what’s called a guided training phase. In this phase the system learns how different cases look and how they should be classified. From that it builds up a model that can be thought of as a memory. In the run phase, the system uses its memory to make decisions based on new data that it sees.

Most modern machine learning systems are statistical classifiers; these include neural networks, as well as a range of other classifier algorithms.

AI trading

Program trading, or trading by software, has been around for decades. Most trading software is designed to follow a predefined set of rules, written in by the programmer.

A set of rules might be simply written like this: Buy when indicator 1 has a value with a range and indicator 2 has a value within a range and indicator 3 has a value within a range.

This approach has several weaknesses. The first is that the programmer needs to anticipate the ranges in advance. The second is that the relative importance of each indicator is unknown and extremely difficult to quantify into written rules. This rule writing approach tends to be complex and can lead to brittle trading systems. The third weakness is that it can result in design bias because the programmer develops the code to work well on one or two particular historical data sets.

A machine learning system does not use rules but rather a statistical model. With a machine learning system, a human trader gives a few chart examples of good entry points for buy orders and for sell orders. This is the training phase. The system then analyzes all of the chart data to mathematically determine what characteristics good buy trades and good sell trades have in common. It uses this model to make new buy and sell decision as it sees new chart data.

To summarize then, a software writer looks at the chart and tries to figure out trading rules that would work in that given scenario. Machine learning works the other way around. It looks at the data and tries to figure out what essential differences exist so that it can use that to make its decisions.

GIGO

Machine learning systems need to be able to discriminate good trades from bad trades. That means a good trade entry point needs to have characteristics that allows the system to set it apart from any other entry points that are not ideal. This is true of human traders as well. A human trader might use a set of indicators, and look for specific chart patterns like candlestick arrangements to make that decision. An AI trading system needs to do the same. The machine system has some advantages because it is typically much better at data mining; that is finding relationships in big data sets.

That does mean however that the output or results will only be as good as the data that is fed in. This is the GIGO rule: garbage in, garbage out.

Quickly test your trading ideas

A big plus of this approach is that you can test trading ideas quickly and without writing software. Most expert advisors require at the very least a few hundred lines of scripting language and can take weeks to create.

If you can’t write the software yourself you’ll have to write down your idea and pass it to a developer. Then hope they understand it. Furthermore, if your trade idea proves successful, what’s to stop them using it themselves, or selling it on. Either of these would dilute its profitability.

We created Tradoso.com to overcome these problems. You can test the feasibility of a trade idea yourself, quickly and efficiently without writing a single line of code.

When to use AI

Training an AI trading algorithm is much faster and simpler than writing software. As our help video shows, you can create a trading system in no more than a few minutes. The advantages of creating your own system are several.

  1. its quick and easy
  2. nobody else has your code or your idea
  3. you can modify and retrain it yourself as you get results
  4. you can incorporate standard indicators or your own
  5. you can test and deploy ideas rapidly

Get Started

Risk Disclosure: Information given on this website is for general purposes only and should not be construed as investment advice. Certain instruments shown here are complex and may come with a high risk of losing money rapidly due to leverage. You should consider whether you understand how these instruments work and whether you can afford to take the high risk of losing your money. Market data displayed is indicative and is not a solicitation to buy or sell.