Technical Trading in Financial Markets

December 2017

After attacking the problem of fully automated technical trading using numerous AI approaches, we finally made some headway. We had previously tried systems built using Genetic Programming, regressions of various types (logistic to symbolic), and Neural Networks, some of which provided interesting and promising results. We even set about searching huge parameter spaces using our modest High Performance Computing facility. In the end though, our best results came down to some well defined mathematically based machine learning, and a steady disciplined approach.

Over our period of work in this area, we have noted many groups' attempts that have provided good short term gains. However, so far they have invariably failed the very important test of time. This is why we always work with a good 20 years' worth of data.

What we have found, is a substantial first step in the right direction. Ironically, however, this project is now being put on hold due to a need for further funding to take it forward; if you are interested in funding such research projects, then please register your interest with us!

Testing on historical data, with our best (of 6) performing FX pair...


Days:                     4871
(Years:                    13)
Profitable Trades:         229
Losing Trades:             167
Difference:                 62
Profit:Loss Odds are   1.371:1
Total Profits:          223550*
Total Losses:          -135102*
Maximum Drawdown:         9510*
Maximum Drawdown (%):     4.25
Sharpe Ratio:         2.717779

* ticks

ER Relocation

October 2015

We have now moved to our new office in Staffordshire, where we have far more room. As we finish refurbishing the property, we look forward to continuing our work and using our new space.

3D Printing (more accurately, Additive Manufacturing)

July 2014

Over the last 18 months we have been working on a company's new product to perform the necessary pre-processing for 3D designs that are to be sent for 3D printing. As this work has targetted additive manufacturing, this has involved preparation of 3D models for (primarily) metal printing systems. Without such software, an enormous amount of expertise is required to both ensure that printed parts are generated successfully and optimally, to ensure the efficient use of resources.

No other software provided a solution to perform the two major functions needed, never mind to achieve this using a fully automated (intelligent) system.

A working prototype was created, and this was well received by the internal consultants. In addition, as the second flagship product of the business, it was demonstrated to a Blue Chip company as part of the main business proposition, for potential acquisition. In May, the acquisition was successful, with the company selling for $88M. Of course, we are very happy to have played a significant part of this success for the business.

Augmented Reality

January 2013

ER had been tasked with reviewing the current software of a mobile phone software house, and investigating the latest methods of computer vision.

Although the company already have the most efficient augmented reality software in the market, their software relies on fiducial tracking, which is becoming an older technology. The plan has been to design their next software platform.

The work has gone extremely well, and we have enjoyed learning a huge amount in this area while poring over the many scientific publications. We are very pleased that after completing the review we were able to present the strengths and weaknesses of their current system, showing how it could be very successfuly leveraged beyond the competition. In addition, we produced a new design that will see them way into the future. ER was applauded for its work by several (independent) stakeholders, and was invited to help develop the new system.

The computer as my teacher

April 2012

After a month of research and design, and three months of development work, the software is deployed. A new system has been created for automatically learning what a student is struggling with, and adapting their course accordingly. This work has been part of a European Commission project ("TARGET") that focusses on how modern technology can greatly impact education, and had the specific aim of targeting business men and women.

The software takes unlabelled time-series data from user lessons in a virtual teaching environment, and determines what the norm (of that data) should be, given how an individual student has performed relative to others. Not only does it take into account the experiences of others, but it also uses information about how they have performed in other tests, tailoring each lesson as a student progresses through the system, so that it is never too hard nor too difficult.


We are asked: can YOU make money?

September 2011

After being approached by an independent investor, we have been impelled to develop a new system to trade the financial markets. This has proved to be a fascinating problem for us to work on. Unsurprisingly, for reasons of privacy, the investor must remain anonymous.

Developing software to carry out the live (and demo) trades has been the easy part. We have since developed an advanced (symbolic regression) system that uses GPUs to search the algorithmic space of trading strategies, using genetic programming. This software continues to be developed, but is already performing as intended, searching for successful trading strategies throughout each night.

So far we have seen some potential in our results, however, we have also met with specialists in artificial intelligence, and in finance and machine learning, who have related their own fruitless experiences. We remain conservatively optimistic.