High Frequency Trading Algorithms

What are High Frequency Trading Algorithms?

65% of all trades now a days are handles by algorithms. These algorithms reside on servers that are located as close as possible to the exchanges so they can get access to information before anyone else. This is a world where the speed of light is a factor. Companies have spent as much as 300M to shave milliseconds in time so the algorithms can be more competitive. The flash crash occurred on May 10 at 2:45. The algorithms made a decision without anything happening around the world and essentially took a trillion dollars out of the financial markets. Today we still don’t know what happened during that half hour at 2:45. Although there has been one arrest, there still is no 100% conclusive evidence to say that one single trader caused the crash. The limit to human decision making is 1000ms. The algorithmic ecosystem is dependent on the speed of light.

The speed of light takes 65ms to travel from London to New York. A trade can be done on the Nasdaq in 1×10 micro seconds. Recent chip developments had produced results in the 740 Nano seconds range. A subset of algorithmic trading is HFT, which can be broke into market makers and statistical arbitrage, which is trying to guess what will happen in the future. One example of statistical arbitrage is an algorithm that does pairs trading or analysis of one stock or index to a similar one. It determines the correlation among these and when there is divergence it executes.

For institutions that trade large volumes of stock they may want to liquidate their positions. Rather than sell as one order they may use an algorithm to breakup into smaller units (“iceberging”).  Algorithms are getting smarter. Rather than just processing numbers they are also now able to read words and act on the information.

For example, if there were a tweet that came out from Bloomberg about an oilfield in the middle east that shutdown it would process this information in nanoseconds and sell oil futures before your brain could even process the information. As technology advances and more and more firms develop these ultra-fast algorithms we could see more and more crashes. They may not be large ones, but if you think about an algorithm as something designed to move in and out quickly than if there are too many algorithms all trying to get out or get in at the same time it will cause a massive spike.