Visualisation techniques are essential to gaining insight into data, processes and systems, and could also answer the question of whether high frequency trading is good or bad for financial markets, said Kerr Hatrick, head of Deutsche Bank’s Quantitative Products One team for Europe and Asia at MathWorks’ Matlab Computational Finance Conference in London this week.
Hatrick described the use of data visualisation at Deutsche Bank, supported by MathWorks’ Matlab programming language and interactive environment for quantitative applications, as important to information discovery and a means to allow management to understand data quickly and easily.
“There is a transition on the quantitative side from model to data-driven work,” he said. “Quant research has become data-driven and Matlab can help. We use Matlab for tasks such as constructing portfolios, its optimisation is more flexible than that of other solutions. It can also “denoise” data, has a number of toolboxes for specific tasks and a compiler that allows Matlab functionality to be integrated with other systems. It is not Excel, so that is a big tick in the box.”
On the visualisation aspect of data management, Kerr noted other available tools such as GlyphSea, Internet Maps, Mathematica and Cascade from NYT Research, and described the 3D representation provided by Matlab before running a series of Matlab visualisations of high-frequency trading data. But he did caution: “Fabulous visualisation is all very well, but if it detracts from seeing data patterns you may as well use line charts.”
On a more positive note, Kerr’s visualisations showed data changes over the past few years, including the 2008 financial crisis. The visualisations included an animated bar graph showing the split of trading as new venues opened – the images over time are strung together with an open source tool, scatter plots of traded volume and value data, spread data over time, the order book over time and the effect of high frequency trading in terms of the number of price quote changes occurring every day.
Commenting on the output of the visualisations, Kerr said: “As time passes, it is clear that volatility comes and goes, but this is caused by macro events in the market and not particularly by automation. Looking at the data visualisations you can see that high frequency trading doesn’t deserve to be vilified, but whether it is a good thing is more difficult to prove.”
Beyond data visualisation, the Matlab product family is designed to provide a single environment for quantitative data analysis and forecasting; risk measurement and management; developing, testing and implementing trading strategies; price calculations; portfolio valuations, optimisation and construction; and the implementation of quantitative applications into front-office systems.
Commenting on MathWorks’ drive to develop new products and processes that deliver great processing speed and efficiencies, Roy Lurie, vice president of engineering for Matlab at MathWorks, said: “Matlab can now handle a reasonable size problem with, say, 400 data points in 0.2 milliseconds.” A problem with 1 million data points takes 183 milliseconds.
In terms of enhancing the company’s technologies for users, recent additions include the ability to allow Matlab components to run by an applications server, database or wherever is appropriate, and support for multicore and graphical processing units in the Matlab parallel computing toolbox that can significantly increase computation power and speed up fine grained aspects of finance systems such as Monte Carlo simulations.
As well as developments in MathWorks’ computational capacity and product set, the company has been building resources around Matlab both internally and externally. Lurie said that since his arrival at the company in 1994 it has grown from 220 employees to 2,400, including 1,000 developers, and hit revenue of $700 million in 2011. He noted 1 million Matlab users in 175 countries and a large community of developers accessing the free-to-use Matlab Central website to exchange files, ideas and information.
Subscribe to our newsletter
