Automated Trading

About a year after I got into the world of trading, like most new traders I stumbled into the world of automated trading. I thought it might be worth sharing some of my experiences here on my blog.

About a year after I got into the world of trading, like most new traders I stumbled into the world of automated trading. I thought it might be worth sharing some of my experiences here on my blog.

The Lure of Automated Trading
Many beginning traders are lured into the world of automated trading like moths to a flame. After spending many long hours glued to the screen failing at day trading, the idea of handing it all over to an auto pilot sounds very compelling. For one, a computer is more patient and is not likely to miss “the one that got away” because you needed to take a toilet break. Similarly, a computer does not suffer from the emotional problems us weak humans suffer from, they just execute the rules.

My Story
I tried my hand at automated trading for over a year. It was much more work than I ever thought.

I started my journey in to the world of automated trading by looking for a system. I started by downloading every system I could get my hands on and back tested it. From a cast of a thousand plus systems I let my poor old PC run for days on end testing every system on all manner of data sets with all manner of settings and optimizations. I reduced the list down to a handful of systems that looked promising and I dissected them, rebuilt them, combined ideas from different systems, testing new ideas. The net result of my first pass at trying to find a system was an abject failure. Not one of the systems seemed to by appropriate.

Given that many of these systems seemed to be for the equities market, I realised that may be I needed a forex specific system. I worked my way through every forex book I could get on trading systems and through every system discussed online. Not one of the systems seemed to by appropriate either, but it did give me some ideas.

Eventually I got tired of drifting from one system to the next and settled on just trying to get one or two concepts to work. I focused my energies on a mean reversion system and a trend trading system. None of these systems did brilliantly, but the important point was that with the application of appropriate money management I could turn a mediocre system into something of value. At that point things got a bit easier. I had a couple of systems and I was ready to face the market.

My first few attempts at this with a demo account was a disaster. I fiddled endlessly with different home brew adapters between my charting software, my real time price feed and my brokers order entry API. I cannot tell you the hours I spent tearing my hair out working with very buggy FXCM APIs and learning how to compensate for a very dodgy broadband connection and learning to trap events from my UPS to deal with power failures. I never wanted to become an expert in any of these things, but out of necessity I had to.
After I climbed that mountain, I set my software in motion on a demo account. The system seemed to work fine for a while. But I was impatient. I had spent all this time on trading and I had no dosh to show for it, so I then unleashed it on a small mini trading account to see where it would go (hint hint – even greed affects automated traders).

After leaving my system run for a few days on my small account, I woke up one morning to find out that I had increased my account by 50%. Yahoooooo. Then I inspected the logs – oh crap – my system actually built up a series of over leveraged positions and then failed to close a series of trades due to a comedy of errors in my software. I just got lucky – really lucky – my buggy positions went in the right direction.

I fixed the bug in the system and then left it run for a while. The system slowly ground my account into a pulp over the coming weeks. I then fiddled around with my systems for a few more months believing I could fix it – but things never really improved.

Eventually after fiddling with these systems for a while another penny dropped for me. To trade well you need a system that was ideally suited to the market conditions. In other words – to state the blindingly obvious you should range trade a ranging market and trend trade a trending market. But here comes the rub – to get an automated system to trade adaptively is extremely difficult because the naive logic encoded in computer can easily be tricked by the market. It is this kind of thing that meant my systems were slowly eating my account away. The only thing that slowed the decline into oblivion was my money management strategy. I had effectively reduced myself into becoming a noise trader.

Can it be done?
Looking back at the experience it was very useful – but I was really kidding myself. I wasn’t ready yet. Learning stuff from back testing other people’s ideas in books is one thing, but I really lacked the experience to know what is important to build into systems. A good automated trader is usually a trader who has paid his dues through years of discretionary trading and is able to reduce his thought processes down to a set of repeatable rules that a computer can execute. It can be done, but it is not easy.

There are examples of successful automated traders out there. One of my favorites is Malcolm Morley. His trading results can be found here. Another beautiful example is the team of guys who thought they had found the holy grail of trading systems and endeavored to trade their way to a million pounds in a thousand days. They made it a reasonable distance towards their ultimate goal but gave a large chunk of their profits back to the market when a regime shift in the market meant that they where not trading the right system for the market conditions.

Automated Trading by the Professionals
A number of market commentators often post that the nature of the market has changed since the late nineties because the prevalence of automated trading has increased. They have blamed increased volatility and non-explainable market movements on automated traders and that “most” of the volume on the market is done by automated systems these days. They point to decreasing rates of employment by the finance industry as a sign of increasing levels of automation.

No doubt about it, markets do evolve and I am willing to believe that “some” of the volume on the market is connected to automated trading, but I find it difficult to accept that “most” of the volume is due to automated trading.

If you google the term “automated trading” it is interesting to note that the first page of results is stuff for retail traders or stuff written by academics. It is a good sign of who is using it. It is not the big guys. Most of the big money in the markets these days is still money from central bank reserves, sovereign funds looking after a countries coffers, wholesale funds looking after your mum and dad’s pension and multinational companies who sell hamburgers, soft drinks, drugs, oil, steel, credit, etc. The majority of players still use old fashioned position trading methods that date back to Grahame and Dodd and about the only high tech thing that has improved is they use a bit of asset allocation theory / portfolio management theory and perhaps they protect themselves with some options or some other structured product. Such funds are usually bound by the simple one directional plays based and asset allocation rules described in their prospectus. Very few prospectus for the really big funds with really deep pockets rarely say anything like “we use automated trading” in their prospectus.

Automated trading is usually the area of boutique hedge funds or quant trading desks in large houses and is sold to the market as an alternative form of investment to complement the boring old fashioned single directional play portfolios. Many large firms operate quant trading desks because their customers ask for it and they think they can collect the management fees.

Automated trading tends to operate in specific niches of the market where the speed, patience and objectivity of a computer can out pace a human. For example I would be willing to believe the days of the manual scalper is numbered and trading desks are letting their scalpers go in favour of the Dilbert types who can program.

In terms of explaining the reducing numbers of staff in finance area, in addition to the reduced number of seats for scalpers, the big funds are hiring the Dilbert types to help them balance their portfolios and either sacking half their analyst staff or outsourcing the analysis function because they have realized that the overhead of having large analyst services is not giving them enough of an edge for the overheads that they bring.