High Frequency Trading

CODEX Entry 2450: High Frequency Trading


High Frequency Trading or HFT averages 40% of trading volume in equities, and 15% of volume in foreign exchange and commodities. HFT firms do not hold their portfolios overnight, and therefore their risks are an order of magnitude lower than traditional buy-and-hold strategies. All of their strategies involve leveraging latency, or their speed advantage, to ‘beat’ orders to a stock exchange server and buy or sell shares in advance of the market, and to then sell them once the market’s orders arrive at a slightly higher or lower price. This is achieved in two ways. The first relates to speed of market access. HFTs may own cable or microwave links between the main exchanges that are faster or straighter than those used by traditional stock orders. Or they pay a premium to place their server inside the exchange server to reduce latency. With each or both of these techniques, the HFT fund will now leave a very small bid open on the Chicago exchange for a particular stock. This is simply a market finder. When a large order is received for that stock, and the bid is triggered, a signal is sent swiftly ahead on the HFT link to buy the shares in New Jersey, before the offer arrives at that exchange. The HFT offer arrives a couple of microseconds before the original offer, and buys the stocks. The original offer will have a spread, so once the first offer has been rejected, the offer will automatically be raised by small increments within that spread. The HFT has now bid the shares it just bought back on the exchange for 0.1% more than its purchase price. The original offer, with the incrementally higher offer, now arrives and buys the HFT shares. By dealing in massive volume, these 0.1% premiums add up to significant profits at almost no risks.

A latency advantage can also be achieved by faster computers or faster algorithms. Examples of this are ‘Tickertape trading’, when computers read data announcements faster than humans, or through access to the data microseconds before the traders. This microsecond advantage can then be used to place orders ahead of the traders. This is linked to ‘Event arbitrage’, where the computer is familiar with stock reactions in the past to different events, and therefore reacts quicker to any announced events. ‘Statistical arbitrage’ is achieved by rapid algorithms that look for any inconsistencies between interest rates, bond yields, and exchange rates around the world and trades those spreads before the market has naturally corrected them. With ‘Index arbitrage’ the algorithms know the weighting of the Index Funds. When prices rise disproportionately in one sector, the Index funds have to recalibrate their portfolios to maintain their weighting. As the HFT have faster access to the market they can use algorithms to predict this buying and selling of the index funds, microseconds ahead of their processors. There are many other arbitrage strategies based on whatever latency advantage an HFT firm can secure.

There are examples of this advantage being used at under 20 nanoseconds. This behaviour is not legal for humans but it’s legality with a computer is a judicial grey area which HFT players exploit. HFT represents the largest proportion of exchanges’ revenues, discouraging reform from within. Partners in the biggest firms like Tudor, Citadel, Renaissance, and DE Shaw, pay themselves individual bonuses in excess of US$1 Billion each year, for skimming the market with this range of strategies, largely because the ‘revolving door’ structure of SEC membership in combination with politicians who do not understand the details of these trades, ensures that reform is always several years behind the markets HFTs are generating.