Tue, 05/04/2016 - 12:20
Piquant Technologies' Pegasus Fund launched in 2013 and has delivered stable and diversifying returns annualising at just over 10 per cent. Piquant runs a fully automated portfolio, gaining trading insights by applying machine-learning techniques to huge datasets.
The model uses multiple signals and as Piquant's Chief Investment Officer James Holloway (pictured) comments: "It's hard to find new signals and it's even harder to aggregate all that information and know what to do with it; not only to try and determine which way the markets are going to go, but also to make predictions about risks, costs, and liquidity.
"We have an engine at the core of our strategy that absorbs hundreds of thousands of market statistics every day, allowing the portfolio to adapt to changing market conditions. The engine understands how to use this information and this makes us good at making allocation decisions."
Holloway was a Senior Researcher at Winton Capital and Piquant's two other senior partners, CEO, Roddy Orr and CTO, Iain Buchanan were formerly with Goldman Sachs and Aspect Capital respectively.
Last year, Piquant's assets grew fourfold on the back of a successful 2014, which saw Pegasus return 19.87 per cent.
At the centre of the strategy lies one core allocation/decision-making system that controls everything. The system does not simply look at markets in isolation and try to predict their directionality, it will also look at their peers to determine how they are doing on a relative basis.
"We might look at the Hang Seng Index to see how well it is doing relative to the Nikkei 225, or to the S&P 500. We're not just taking an outright view. At least part of our portfolio is based on making relative judgments based on various common features of the markets," explains Holloway.
Where the machine learning component comes into play is through marshalling all of Piquant's systems, with their multiple signal processing algorithms designed to predict correlations, market movements etc, by pulling in massive data streams from which the different pieces of information are extracted. The more the system understands these signals over time, the better it is able to make decisions in the portfolio.
"A signal is basically a statistic that tells us how a particular market is behaving. What our engine does is take all the different signals and determine if they have any use in telling us what is about to happen. Overall, they do. Markets are pretty efficient, but not completely. Each signal gives us a small hint about what is going to happen. We use that information not only to decide what trades we want to do but also to position our trades and make our execution more efficient," says Holloway.
In essence, the Pegasus engine can be thought of as a massive brain, working through and processing data, which it uses to build a view of the market environment. Fundamental data is not used, however, for two reasons.
"Firstly, it's incredibly noisy and it's subject to constant revision. Secondly, it's hard to demonstrate that the information contained within a fundamental data set is different to what is already encoded in the market price. We need to know that we are extracting different, unique pieces of information from the data sets we are pulling in," says Holloway
Mon 27/02/2017 - 12:06
Mon 06/02/2017 - 09:43
Thu 05/01/2017 - 12:18
Fri 28/10/2016 - 10:14
Tue 25/04/2017 - 09:03
Tue 25/04/2017 - 08:50
Tue, 25/Apr/2017 - 13:26
Tue, 25/Apr/2017 - 13:21
Tue, 25/Apr/2017 - 13:14
Tue, 25/Apr/2017 - 12:39
Tue, 25/Apr/2017 - 12:36
Tue, 25/Apr/2017 - 12:34