Energy Trading Robots

03 September 2007

I’ve just finished a really interesting project for a big player in the UK Energy sector, working with the guys from Planlogic .

Hamish is an expert in trading strategy and in particular, Trayport GlobalVision and the RORC trading exchange from Reuters – he came up with the clever algorithms, which I then helped to codify. We’ve now got some nice abstractions and what seems to be a good, reusable framework implemented in .NET.

Basically we built an automated trading system where a trader can set up a price book with a variety of limit orders and associated trading rules. These prices are hidden from the market, but the software we wrote constantly monitors the market changes and hits prices if the rules match. Obviously, time-to-trade is a key measure so there is a real challenge in balancing the needs of readable / maintainable / bugfree code versus high performance and fast execution. The code had to be good too – mistakes could be costly, so we’re using some automated testing based on NUnit.

Energy trading is still relatively young in the UK (complete liberalisation didn’t occur until the late nineties) and so there are still lots of opportunities for decision support technology. I think we’ll see lots more “robots” in the coming years and I wonder how that will play out. Will energy trading just become a competition to see who can build the fastest or cleverest robot?

Somehow, I dont think we’ll ever replace human beings – there are still many things that we cant capture in code – common sense and gut-feel being good examples. Also, I think that humans are still probably the best market makers.