Latency

Much has been made of the ‘race to zero’ for lower and lower latencies in High Frequency Trading, for example the hype generated by Michael Lewis’ infamous book ‘Flash Boys’. It is easy to see why latency matters in an Arbitrage strategy: given multiple firms competing for the same opportunity, the one with the best overall latency will win the profitable trade and keep the money.

Less obviously, but by the same reasoning, on any sudden move in the fair price, the Market Maker with the fastest latency will be able to place their order first on new, empty prices (obtaining time priority) as well as cancel orders with negative edge fastest (allowing them to otherwise quote more tightly, obtaining price priority). Have better latency is better in all (short-horizon) forms of trading.

Suppose a trading firm has a lower latency connection to an exchange 2 ms faster than all other participants. This is equivalent to the trading company seeing 2 ms further into the future than everybody else. Trading firms can make more informed trades given this extra 2 ms of information. Lower latency connections will increase the profitability of practically all trading strategies. Trading companies therefore make significant investments into technology which provides the lowest latency. Here are some of the techniques trading firms use to reduce latency.

Server Colocation: trading firms will often host their servers in close proximity to exchanges to increase the speed of trade execution and arrival of data from an exchange.

State-of-the-art Networks: significant investments are often made in the lowest-latency, highest-bandwidth data transmission technology to reduce latency further. For example, trading companies might invest in fibre optic cables or radio-frequency lines to optimise latency between servers and exchanges, particularly so if the transmission is across large distances such as between Europe and the United States.

Rapid Computation: companies where low latency is a high priority will invest in sophisticated computation technology such as field-programmable gate arrays (FPGAs) to execute standardised algorithms faster than any standard computer. FPGAs are more difficult to program and so are typically only used for a small number of algorithms in which rapid execution is a must.

Fast code: latency-critical aspects of trading, such as connectivity with exchanges, trade execution and price evaluation, is often implemented using a very low latency programming language, such as Java or C++.