EDDIE Beats The Bookies

At the University of Essex, during the mid 1990s, I performed research in machine learning for financial decision making. My work spawned the Centre for Computational Finance and Economic Agents, an advert for which can be seen to the right.

I recommend the courses at the CCFEA to those looking to study higher educational qualifications for a career in finance or who just have an interest in mathematical and computational applications in finance.

My research centred around genetic programming, which I introduced to the computing department at the university, although I was not the originator of the technique. Genetic programming is a field in evolutionary computation whereby a population of decision making trees are evolved using (rather like biological evolution) mutation and competition in a constrained environment to optimise a solution for a decision making problem.

To this day, every now and again, my supervising professor at Essex University contacts me about yet another member of the public asking him for information on the research that led to the paper EDDIE Beats the Bookies, which used horse race betting markets as a proxy for financial markets. Judging by the continued interest, over twenty years later, the paper has gained a cult status amongst those using machine learning applications for horse race betting.

EDDIE (Evolutionary Dynamic Data Investment Evaluator - I like inventing catchy acronyms) came out of the third-year project of my bachelor's degree. I scraped through the degree with a Desmond (Tutu - 2:2) with all my focus on that one project. I had no interest in most of the courses.

First year philosophy I enjoyed for the chance to debate with other students about machine intelligence. Computability I enjoyed because it all referenced back to the work of Alan Turing. The bottom-up AI I understood before I entered university (and is why I taught genetic programming to the university and not vice-versa). The rest of the courses were of no interest to me, including the top-down AI course, which I never liked as nature didn't create human intelligence in such a God-like manner.

After graduating I stayed in Essex to continue my research in a PhD. A colleague mentioned my work to a friend of his working for a subsidiary (Equisoft) of Reuters that happened to be in Colchester, where the university is located. I was invited, along with my professor, to talk to Equisoft about our research. Unknown to me, Equisoft was working on a project for Reuters called Reuters 3000 Xtra and wanted to add machine learning to the trading software for optimising trading systems.

Rather than offering a collaboration to Essex University, I was offered a job in a blue-skies research group within Reuters instead and so I left the PhD after a year of study. I did a little more work on an EDDIE clone for Reuters but then moved on to machine translation of financial news stories. I had the foresight to see that computers could automatically trade events without human interaction through machine translation of the news.

My thoughts went further in that I thought that the best way to translate news stories was to write them in a simplified manner that machines could understand more easily and then apply a template to the story to make it look more flowery for human consumption. Unfortunately, the people I reported to at Reuters didn't have my foresight and nothing came of it. Today, all of the big trading firms will have trading systems based around machine translation of financial news.

I also had ideas about RFID technology for tracking goods and their interactions with people in shops. Also shops that had no frontage, just a dark storage facility with robots picking items and shipping them to customers. For the 1990s I was in the wrong company for all these ideas!

Now, I find myself writing about my research. Teaching, writing and research is what I enjoy doing the most. To all those who have asked for more details about EDDIE then please wait a little longer as it will be in my forthcoming book, which will detail machine learning applications for sports trading. The book will include EDDIE and other machine learning methodologies applied to system building, trading and money management.