It’s possible for algorithmic trading systems to make errors indeed. Although rare, mistakes can still occur at the most unexpected times. Some serious mishaps can wipe out the entire fund in a matter of seconds. It’s required to have quality risk management for a sustainable system. There are 3 main risks we found.
Buy and Sell Loop
This is the top risk that can occur in algorithmic trading systems. When a loop of buying high and selling low occurs, 99% of an account in terms of asset value can be lost in just a minute.
Knight Capital, a market maker, lost $440 million in just 45 minutes on August 1, 2012. A new trading algorithm at Knight mistakenly trades 150 stocks, buying higher and selling lower than the market price. Knight was on the brink of bankruptcy when the issue had been flagged. A loop error resulted in a disastrous loss to Knight just in 45 minutes.
Data Error
When data errors occur, like delayed data, the trading system may start to make arbitrary decisions. It is a state in which the system makes trades not based on real-time data. It leads to serious losses and inconsistent performance system-wide. At ALGOTRADE, we use at least 3 different data sources to cross-check in real time and to minimize data errors.
Unexpectedly Low Liquidity
Investors can suffer big losses if they only focus on the latest price and disregard liquidity with market orders. This may not impact small individual accounts; however, the effect is much larger for institutional accounts. In particular, this effect will have a huge impact on multi-algorithm trading with strategies using market orders to open and close positions in the same direction.
On May 6, 2010, the Flash Crash happened in the U.S. market. It was a stock sell-off that led to a sharp price fall in a matter of minutes. Without taking into account liquidity issues, trading orders were executed with a 60% price difference from before the Flash Crash event. It almost wiped out the entire account balances of the parties involved.
Serious risk management to avoid the instant collapse of an algorithmic trading system is extremely important and should be of the highest priority. Other risks related to computer programs such as API and connection errors may occur frequently. Yet these errors are much less likely to crash the entire system.
Iterative improvement is pivotal to enhance a system's stability. However, investors shouldn’t stress too much over small, unavoidable bugs. Even though it may open an arbitrary position, it may not always lead to loss. In fact, erroneous trading positions may still unexpectedly make profits.