The risks involved in algorithmic forex strategies
Algorithmic trading is a method that uses advanced mathematical calculations and automated software to enable the rapid execution of trades.
One such strategy, developed by Haim Bodek, former head of low-latency trading at UBS and founder of Decimus Capital Partners, hinges on using algorithms to exploit mispricing in very specific stocks.
It predicts when a stock will rise and fall based on its past movement and volume – essentially predicting where the price will be at a given time.
Nicknamed “The Knife” because it stands to make or lose money depending on whether you get your order filled before prices to move against you (or not at all), the strategy hinges on using algorithms to exploit mispricing in very specific stocks.
Trading is full of danger, even for professionals – One false move and an opportunity may be lost forever.
Bodek’s strategies are complex enough that they can’t meet their trading goals without adverse events happening occasionally. “An average human trader would not care about trading The Knife because you lose so much money doing it,” he said.
Once algorithms start making their own decisions based on pre-programmed information, there’s no telling what will happen next.
For instance, if there were a bug in the coding logic itself, or if data feeds went down unexpectedly, no human would be able to step in to stop the system.
Bodek is well known for his risky and controversial algorithmic strategies. He was also forced to step down as the head of low-latency trading at UBS because his ideas were too radical.
There are several reasons to use algorithms, and Bodek’s techniques are only the tip of the iceberg.
For example, not all algorithms can be implemented by humans – robots continue working whenever there is a market disruption; they don’t get paid overtime or complain about long hours. They don’t even need coffee breaks.
But the biggest advantage comes in terms of speed: no other method offers such quick execution. Robots can execute thousands of orders per second, making it possible to make vast amounts of money before anyone else notices – or even realizes what just happened.
There is another, less glamorous side to trading. “It’s about the engineers who are coding these things,” says Steven *(not his real name), a senior computer scientist at one of the largest financial firms in the world.
He works for over 5,000 people working on designing algorithms, but it takes dozens of programmers to code every individual strategy. “Sometimes you spend months… It’s very time-consuming.”
Not only is it time-consuming, but the task is inherently complicated. According to *, most strategies are built on top of one another.
For instance, a single strategy might be designed to execute ten orders per second. Another algorithm will then monitor this basic strategy and adjust its own settings if anything goes wrong.
If the first bot starts losing money too quickly, the second tier will automatically trigger another algorithm that takes control from the other bot, preventing possible losses. “I don’t think I’ve seen any other industry where you have so many layers,” he says.
The Problem with Algorithms
The systems are also prone to bugs – programmers are people, too, after all.
Because algorithms are just software applications running on computers, they can be affected by all the same issues as any other application: problems with the hardware, software faults and human error. “There’s always a chance of something going wrong”.
Although he would not go into details about the exact nature of some bugs that have been found in his company, he did explain that errors can come from anywhere.
In one case, a programmer miswrote code, and it simply didn’t work as intended for over a year.
In another case, some data fed into an algorithm came from several sources whose timestamps were slightly out-of-sync.
This is a common problem on trading floors where lots of people are working simultaneously.
Finally, flaws can appear in the algorithms themselves if their creators don’t take the time to think through every possible line of logic.
For an algorithm to make a sound decision, it must combine and process all available information and remember the steps it took to conclude successfully.
That’s why Bodek implemented his risk management system: it prevents algorithms from executing orders unless they’ve been thoroughly tested. “I’m not going to let this thing run if something is wrong with it,” he says