A Comparison of a Neural Network and a Conventional System for Trading FTSE Futures
To conquer fear is the beginning of wisdom.
The Financial Times Stock Exchange FTSE 100 index is the last of the trilogy of our equity index trading systems.
FTSE futures are traded on the Euronext LIFFE CONNECT electronic trading system from 9:00 am to 10:00 pm Central European Time under the symbol Z (IB platform). The Z futures market has a daily trading volume of approximately 100 000 contracts and a daily price range of approximately 100 points. The contract value is £10 per FTSE point and the overnight maintenance margin was £3000 at the time of writing. Either each broker or the exchange can change margins periodically according to market volatility.
Neural network-based systems have generated a great deal of interest lately but are they as effective as the software vendors claim? I decided to investigate. In the first part of this chapter I will present a conventional rule-based system using the multiple regression model which produced the best results in trading gold (see Chapter 11). I used NeuroShell Trader Professional to construct a neural network model using, for comparison purposes, the same intermarket relationships with the conventional multiple regression system. The software’s built-in genetic algorithm was then used to select the best inputs and optimize the parameters. The neural system is presented in the second part of this chapter and the code and ...