Back to the Future

Imagine a scenario where a concept you dreamt up nine years ago finally becomes a reality? That’s the remarkable situation I find myself in now, as here at LFX we begin to implement a design and concept I built when I first started this project many moons ago.

In fact the true story is I designed the system whilst travelling on a train from Edinburgh to London after a work trip and I was bored. An advert for a trading product talked about risks in trading and it sparked an idea. Always travelling with a notebook, (in fact I even sleep with one as the amount of times a solution has come to me whilst asleep is startling), I began to write page after page of notes…

Three months later and I realised that to build it I needed a lot of infrastructure and a lot more knowledge in trading, but I had a working example and it was profitable.

Let’s fast forward through nine years of learning to trade, working in a bank and building Littlefish FX.

Something pretty unusual and weird happened last week which actually meant I got some time to think and clear my head. It resulted in me digging out a few old ideas.

I sat at my computer and started to play with these ideas in our new infrastructure and it turns out I now have everything I originally needed to build what I wanted to.

The concept is simple in theory and a lot of people have tried it but I know my design was approached in a slightly different manner.

Essentially we build trading systems for other people to use in their trading. We encapsulate the logic into a ‘black box’ (they are literally black computer servers) and plug it in and it ponders away. Sometimes we are asked to update or tweak them, other times they are just left to run.

We build all of these systems off of our original underlying logic that I designed years ago; the trading principals haven’t changed and they are unlikely to as they were built on pretty sound principals used over tens if not hundreds of years in trading. Some of our data feeds are pretty unique and ways we can manipulate the data is a bit unusual; which gives us our USP. They are pretty good systems but only the very beginning of a big journey.

This comes to what I wanted to do, and excitingly now we can do it…and are now doing it.

They key here is we have absolutely no idea how well this may or may not work but nonetheless the possibility of finally seeing my ideas realised is incredibly exciting.

The concept is pretty much artificial intelligence, that is to say a computer system that can teach itself.

Now, for those programmers out there, this typically takes the form of a Neural Network, and although we use these actually this is a slightly different way of looking at an artificially intelligent system (and as such I have no intention of sharing any more!).

But suffice to say the idea is to actually have a system that is perfectly unique but can learn from all current and previous information and therefore amend the decisions it makes as it progresses.

Will we be the first people to do this? No I very much doubt it.

But if we can get this to work (which I honestly have no idea if we can, so this is a bit of a brutal blog post if we can’t and you all come back to me in 12 months) then it could be a pretty beautiful thing that many people could benefit from.

Want more? Read CEO Sam Barry’s series on building quant models here

Retail versions of similar components to what we use in our systems feature as part of our Forex Trading course – find out more about it here.