Low latency: the Formula 1 arms race
John Barr, The 451 Group
At a Thomson Reuters event in London, a few months ago, the focus was on low latency. Fast cars are often used as a metaphor for low latency, and this trend was continued as Adam Parr, the CEO of the AT&T Williams Formula 1 racing team, gave the keynote speech.
Parr explained that the difference between first place and last in qualifying for a Formula 1 Grand Prix is about one second, and that the performance of most cars had improved by that margin during the first half of the season. So a car that was sitting in pole position for the first race of the season would now be at the back of the grid unless the team was continually bringing new developments into play. The team's goal is performance, which is delivered through a focus on speed, precision and reliability in order to optimize the car's weight, power and aerodynamics. And Parr noted that some teams really have gone from the front of the grid to the back. He didn't mention any names, but he may have been alluding to world champion Lewis Hamilton at McLaren.
When margins are paper thin, the difference between winning and losing is the blink of an eye. Beyond the technical issues, Parr touched on the team's operational methods, citing data collection and analysis, measured risk-taking, knowing when to say no and teamwork as contributors to success. These issues – narrow margins, the need for continual development and the non-technical issues – all mirror the requirements of low-latency developments in the financial markets.
The remainder of the event was filled with a balanced panel discussion – comprising representatives from Chi-X and Turquoise (two of the leading European MTFs), one investment bank (Goldman Sachs) and a technology vendor (Intel) – called 'Achieving Low Latency For The New Market Landscape.' The growth in market data volumes, the increased number of execution venues and algorithmic trading all mean that low latency is an important issue, but it is one of several tools used by financial firms, and it is difficult to be both the fastest and the best – so reality is often a compromise.
As the number of liquidity venues has increased, the opportunities for both cost reduction and latency-fueled arbitrage have grown. Intel described this as a 'latency arms race,' positioning itself as the arms dealer.
European markets are tracking the changes observed in the US, but at a more rapid pace. Drives toward standardization of tick sizes and market data will increase the fungibility of trading opportunity across multiple venues, although the move to a consolidated tape is driven more by regulatory pressure and the need to be able to demonstrate compliance with MiFID 's best-execution policy. Also, interoperability between two pan-European central counterparties for clearing offers choice and reduces costs.
Many firms have paused for breath following the credit crunch, and are now getting back up to racing speed. So the next wave of activity surrounding multilateral trading facilities (MTFs) is likely to be even more competitive. One or two venues was described by the panel as a monopoly or duopoly; three was viewed as healthy competition; and four or more was seen as 'too much fragmentation.' There are perhaps double that number of European MTFs today, so consolidation or failures will follow. Roughly 35% of London-based trades are now carried out away from the London Stock Exchange, and that number is expected to grow.
While there is competition around latency, the need for consistent latency is as important as low latency. The cost of installing a 16-core system at a colocation facility has dropped significantly in recent years, reducing the barriers to entry for statistical arbitrage traders. But it was also pointed out by the panel that you can't colocate with everyone (unless all of your target venues happen to be in the same datacenter), so the relative location to many venues may be more important than colocation with a single venue.
Having said that, the MTFs asserted that colocation, fast networks and high-performance systems were important to their clients. Latency monitoring and analysis is sporadic, with some components coming under the spotlight, but no one is really exploiting the advertised capabilities of latency tools by analyzing the full scope of a financial transaction's lifecycle. There were hints that new capabilities would soon be available for latency measurement, but no one was able to comment on the record – so watch this space.
The market opportunity to exploit low latency continues to be confusing, since there is more interest in this space than can be justified by the amount of trading that is truly driven by latency. Perhaps some of the players need to decide if they really are in the 'Formula 1' business or not. But for those that are in the latency race, colocation is one of the vital components that can help to tune performance.
