Neural Networks? You have probably seen the term over the years. We have assembled marketing hype from purveyors of this nonsense:

• By analyzing global market relationships, we are able to forecast a market’s move in advance, so our clients can trade with confidence.
• Neural trading systems are poised to change the nature of analysis performed on the financial markets. With the means to develop dynamic trading systems that can adapt themselves to changing market conditions, without the necessity of relying upon preconceived trading rules, this new sixth-generation trading technology has the potential to affect the way computerized traders apply technical and fundamental analyses to global markets of the 1990s.
• Artificial neural systems are information processing models which mimic how the human brain processes information. They are modeled after the structure and function of the brain. Because they can generalize from past experience, neural systems represent a significant advancement over rule-based trading systems, which require a knowledgeable expert to define if-then trading rules to represent market dynamics.
• Neural networks are perhaps the most significant forecasting tool to be applied to the financial markets in recent years.

Sounds great. Just like Elliott Wave, it sounds as if we can predict the market using new brain-trained computers. Give us a break!

It doesn’t matter the year or month, there is always some guru that promises they can time the market. First, they called it market timing, then there was Gann and Elliott and now it’s neural networks and expert systems. Stop trying to predict the market and maybe you will make some money.

The Template Never Changes

Every generation of trading hype follows the same template. The technology changes. The marketing language changes. The underlying promise is identical: we have discovered a way to forecast market direction in advance, and you can trade with confidence because our system knows what will happen next.

In the 1920s it was W.D. Gann’s magic squares and astrological angles. In the 1940s it was Elliott Wave counts. In the 1980s and 1990s it was neural networks and expert systems. In subsequent decades it has been machine learning, deep learning, and artificial intelligence. The technology becomes more sophisticated with each generation. The promise remains the same. The results remain the same.

The third bullet point above is the most instructive because it attempts to argue that neural networks are superior to rule-based trading systems specifically because they do not require preconceived rules. This is presented as a feature. It is actually the core problem. A system that does not have explicit, testable rules cannot be evaluated, validated, or understood. It is a black box whose outputs cannot be audited against a defined logic. When it fails, there is no framework for diagnosing why. When it appears to succeed, there is no way to determine whether the success reflects genuine market structure or overfitting to historical data.

Rule-based trend following is the explicit, testable, auditable alternative. The rules can be stated. They can be applied to historical data. The results can be verified. When the system enters a position, there is a defined reason. When it exits, there is a defined reason. The logic is disclosed rather than hidden inside a black box that “mimics the human brain.” The human brain is exactly what systematic trading exists to bypass.

The closing observation stands across every decade: stop trying to predict the market and maybe you will make some money. Neural networks are prediction tools. Expert systems are prediction tools. Gann analysis is a prediction tool. Elliott Wave is a prediction tool. All of them dress up the same underlying claim in different technological language. The claim is that market direction can be known in advance. The claim is false. Trend following makes no such claim. It responds to what price is doing rather than predicting what it will do. That distinction is the entire difference between the hype on this page and the approach that actually produces documented long-run returns.

Frequently Asked Questions

Why are neural networks for market prediction unreliable?

Because they are pattern-matching tools trained on historical data, and financial markets are nonlinear dynamic systems where historical patterns have limited predictive validity for future price movements. Neural networks trained on financial data are particularly susceptible to overfitting, finding patterns that appear predictive in the training data but fail on new data. They also lack transparency: when a neural network fails, there is no explicit rule structure to diagnose why.

What is the difference between a neural network trading system and a rule-based trend following system?

A neural network system makes decisions based on learned patterns in historical data without explicit rules that can be audited or tested. A rule-based trend following system makes decisions based on defined, stated criteria that can be applied consistently, tested on historical data, and evaluated for robustness. The trend following system’s logic is fully disclosed. The neural network’s logic is a black box. When the trend following system fails, the failure can be diagnosed. When the neural network fails, there is no framework for understanding why.

Why does the prediction promise keep appearing in new technological forms?

Because the desire to know what markets will do in advance is permanent, and each new technology provides a new vehicle for the same promise. The marketing template is simple and effective: our new technology has solved the prediction problem. It has not. The markets are nonlinear dynamic systems that resist prediction. The appropriate response is not to build a better prediction tool but to stop predicting and build a reactive system instead.

Trend Following Systems
Want to learn more and start trading trend following systems? Start here.