Algorithmic trading benefits: Trade like a Cyborg
In the United States, high-frequency trading firms represented 2% of the approximately 20,000 firms operating in 2010, but accounted for 73% of all equity orders volume, according to Aite Group. The Bank of England estimates similar percentages for the UK. The rise of the machines in trading often gives people associations like the one in the clip from Terminator Salvation below. Be sure to watch the one-ish minute clip by pressing the play triangle button:
We imagine the artificially intelligent motorcycle robot, representing the swarms of trading bots in the market, attacks us. It is intent on destroying us, our livelyhood. We think we need to put a symbolic knife to the processing capabilities of the aggressive robots, before we are capable of using them in any fashion.
While the above makes for a thrilling Hollywood plotline template, it’s the wrong metaphor for how we can use software. As computers, and more importantly their software, grow in capabilities, we are quickly becoming more adept at using them. In fact, they are becoming part of us. If you’ve ever had a mobile phone lost or stolen, you quickly realize how much this machine and its software has become a part of your life.
Trading with software has a number of benefits compared to manual discretionary trading. For one, programming your computer forces you to explain to a computer exactly how you want to trade. As a result, you define your system. Once you have an algorithm in production, it will keep running as long as you want-even over a 24 period. Although it may be obvious, you can have software which uses various mathematical tools, in order to calculate any part of your trading system: entry, position sizing, exit. In the end, the trick is to have the programs complement and enhance your trading where possible, instead of being a drag on your performance. This “cyborg” approach can add alpha to your returns, if you do it right.
Software forces you to trade scientifically
Software trading-in my mind-means you need to create or at least use existing software. Computers are essentially simple, rather stupid, calculation engines. You must be very precise when specifying e.g. your entry signals, so that the software knows when to enter into a position. For example, if you make the entry signal more sensitive, you will enter into more trades. This means higher transaction costs. If your edge is large enough, you expect to make this back. Alternatively, you may want to make your entry signal less sensitive, e.g. by introducing more entry criteria. Then you take less trades, but assuming your criteria are well-chosen, your returns per trade are higher. In short, your software forces you to think through your strategy much more, because you have to describe it to such a simpleton: your computer.
Rishi Narang, in his book Inside the Black Box: The Simple Truth about Quantitative Trading, says:
Computers are obviously powerful tools, but without absolutely precise instruction, the can achieve nothing. So, to make a computer implement a “black box trading strategy” requires an enormous amount of effort on the part of the developer. You can’t tell a computer to “find cheap stocks”. You have to specify what “find” means, what “cheap” means, and what “stocks” are.
This level of precision means that the process can be replicated systematically, and even more importantly, it can be tested. It is not hard to construct hypothesis tests using basic statistics, in order to determine if your strategy’s results are better than pure randomness. If you do this, you are getting deeper into using the scientific method, in order to prove that a particular strategy works. The computer forces you to trade scientifically.
Moreover, defining all of the components of your system precisely means that you take your emotional bias out of the equation. There are a number of trading biases we are all subject to, including the following, according to Van Tharp, in his introduction to trading system development with a cheesy title Trade Your Way to Financial Freedom:
- Representation Bias: over-realiance on assumptions that something is what it seems to be, like only using price bars instead of probabilities
- Reliability Bias: data quality can significant distort the results of your tests, i.e. don’t assume your data is perfect.
- Lotto Bias: the illusion of control you feel when analyzing the data can give you false confidence, similar to when you pick a number in the lottery
- Law of Small Numbers Bias: Giving too much weight to the outliers, since they draw more attention to themselves, particularly visually
- Conservatism Bias: we fail to recognize contradictory evidence
- Randomness Bias: Markets have much fatter tails than normal random distributions
- Need-to-Understand Bias: this results in ignoring randomness, which can always occur, by trying to force elaborate theories on the market
The above list doesn’t even go into the psychological biases you can fall victim to when testing or actually implementing a system in a live environment. If you create software, it will help you become aware of these biases. Describing your system to a computer forces you to be more aware of your market, and to think critically about exactly what you want to do.
Once you do have a system, you can continue using the scientific method to improve it continuously. Your initial focus is on creating an effective system. Afterwards, you can improve it, by using hypothesis testing, to determine if a particular change will give you a statistically significant improvement, over what you already have. It is easier to separate yourself from the idea, if you have an expression of the idea in software.
Trading Software: Always On
Once you have actually implemented a system, it can be watching the markets constantly for entry and exit signals continuously. Some markets, like FX markets that run from Sunday night to Friday night London time, are much better served by algos than human beings. As long as you don’t have a catastrophic failure on your production platform, your software will be watching the relevant price streams indefinitely, and executing mercilessly.
This is really a major benefit of using software to trade. Not only does it work better in global markets, they can work whenever markets are open. If you trust your algos, your portfolio never goes on vacation, even though you might be sipping a Bahama Mama cocktail in the Caribbean.
Your algos won’t call in sick. They won’t sleep. They won’t ever unionize. They will silently execute all of the trades you said you wanted, exactly how you said you wanted them.All of this works-assuming you have a good system in the first place. If not, the algos will be extremely efficient at losing money for you.
That’s the challenge really. You need to get over the initial cost of creating a system, and then coding it. You have to make hard decisions, about how you want to react to changing market conditions, which will inevitably come. All of this does require some research; however, once you have it working, you’re golden, until the market shifts.
Optional high tech analysis
The essence of your trading system is…not surprisingly, a trading SYSTEM. You can use the magic of computers to generate any number of indicators, analyses or studies pretty much on the fly, but ultimately the system’s only purpose is to make money. This is simple.
In Computer Analysis of the Futures Markets, Chuck Le Beau says:
The computer allows us to take any set of price data and manipulate it in infinite ways. We can smooth it, accelerate it, magnify it, compact it, color it, transport it, overlay it, store it, and erase it. Our possibilities are almost endless, and therein lies the problem as well as the opportunity. What exactly are we looking for and how will we know it when find it?
Some trading systems are very simple, and don’t require a computer at all. They can make money in any conditions, because they capture an essential characteristic of a particular market, taking advantage of particular anomalies. For example, a systematic mis-pricing can occur in the markets, identified with very simple criteria, and can be easily have software responding to it.
The January effect comes to mind…that old chestnut. Putting on a trade on the US stock market index for the month of January-for the duration of the month typically yields a net positive return. Academics have argued this is because institutional investors are re-entering the market after realizing gains and losses from the previous year for tax reasons. Paying attention to the markets is enough; you don’t really need a computer to discover this. A computer can help you get a feel for the expected return profile, but actually this can be traded manually or automatically. The computer is almost irrelevant in such a case, as it’s a very simple criterion. The “story” or thesis behind the system is more important than using a specific approach. This January (2012) happened to be a case in point. In and of itself, trading the january effect is not a system, but it can be one of the criteria used in an equity trading system.
By using software, though, you can test any mathematical or computational tool you want against your time series data. From regressions through more sophisticated mathematical and statistical tools, you can try your hand(s) at forecasting. With a machine, it’s only a matter of creating or using a program that will precisely identify what you want to do in the market. Once you have the right software architecture, it becomes easy to pull in various libraries using complex numerical methods, to leverage the computer’s capabilities fully.
The Benefits of Trading Like a Cyborg
The single biggest benefit of using software when trading, is that you can build on the strengths of using computers as a tool. Both you and your computer have relative strengths which complement one another, similar to the comparative advantage underlying specialization.
Computers force you to think like a mad scientist, obsessed with using technology to become a cyborg: half-human, half-machine. In fact, even professional discretionary traders are increasingly reaching for off-the-shelf execution algorithms provided by the brokers, in order to get the best possible price on a particular stock. This of course clouds the picture, and makes it hard to differentiate between old school “manual” trading and black box programmed algorithmic trading, because it all looks the same from the exchange’s point of view.
The scientific method, summarized above in the picture, confers rigor to your trading. Even though the approach can be used to deal with randomness, you can gain considerable insight into patterns of market behavior, with this approach.
And as a final parting thought and a reward for actually reading this far, I wanted to leave you with a good example of a proper cyborg. Although RoboCop is currently employed in the law enforcement sector, imagine what becoming a trading RoboCop can mean for your trading performance.