High frequency trading it’s an arms race to create the best bot algorithms. Its main characteristic is winning small amounts from a very large number of orders placed at strategic times, based on clear defined strategies. Was profitable in the beginning but as time passed, it became a crowded space with lots of competition. In the end, the best algos will succeed dominating the market, while 99% of other competitors will be excluded by increasing expenses and drop in profitability.
The following practices are representing just a small fraction from the multitude of strategies applied in this space.
Front Running / Latency arbitrage
A major client places a large order to it’s broker. The order must be splitted amount different liquidity providers. Because exchanges are in different locations there is a latency between each order. A third party can detect the first order initiated on the closest location and create new orders on the remaining providers before the primary orders are registered, pushing the price higher. One solution to this problem is to make the orders in parallel by issuing the first one to the slowest provider.
Custom Order Types
Some brokers might create special order types for their clients, allowing the orders to take priority over others placed at the same price level. This gives an unfair advantage by allowing a privileged position in the queue.
Having access to the latest news, algorithms can analyze the information and make a trade decision much faster than any human trader could do. Such events are visible on low time frames as a result. During crashes the bots usually get out as fast as possible creating a further liquidity issue.
By monitoring all the exchanges at any moment, the price spread can be detected and orders instantly executed keeping the prices aligned at all time.
High frequency trading was both critiqued for its effect on the market but at the same time appreciated as a liquidity provider mechanism that makes the market more efficient.