As hedge funds continue to tap into alternative datasets to help manage risk and boost performance, the need for managers to strengthen their expertise in sourcing, testing and processing information is now critical.
As hedge funds continue to tap into alternative datasets to help manage risk and boost performance, the need for managers to strengthen their expertise in sourcing, testing and processing information is now critical.
In the past, hedge fund strategies using computer-based investment models historically leaned on rear-view structural and market data – such as GDP growth or current account data – for their inputs. But as the pace of markets has quickened, so technology has rapidly advanced, with systems moving with the momentum.
As a result, quant strategies now look to an ever-expanding set of alternative data – from shipping and flight patterns to geolocation data and weather mapping, as well as natural language processing – to get an edge on the competition. According to a poll of some 30 hedge fund firms conducted by Hedgeweek, more managers are planning to increase their usage of alt-data in the coming 12 months.
Reflecting the expansive volume of alternative data on the market, different hedge fund strategies – equity, macro, CTAs and so on – are utilising different forms of alternative data in an assortment of ways in their day-to-day processes.
Thomas McHugh, co-founder and CEO of Finbourne Technology, is keen to draw a distinction between the data available to managers and the usefulness of the information it can generate.
“What you tend to have are many platforms that will give you access to vast quantities of data, but very few of them that have the in-built sophistication to turn that data into valuable information,” McHugh explains.
Established in 2016 with the aim of reducing the cost of investing and strengthening transparency, Finbourne Technology has grown from seven founders to more than 140 people in under five years. Specifically, it combines a cloud-native, SaaS foundation with digital data management capabilities and PMS functionality, which Finbourne sees as key to achieving efficiency, scale and diversification across hedge fund workflows.
Sentiment
Underlining the value-add of Finbourne’s offering, McHugh says: “You need a platform accessible via APIs, that works alongside machine learning frameworks and all of the other the toolkits out there, but in a way where it knows what a trade is, what a portfolio is, what an instrument is, what a cost-basis is. That’s where we are trying to operate.
“We know hedge funds are primarily focused on generating alpha and reducing costs. As a result they no longer want to host and be responsible for maintaining tech environments. Our aim is to take that pain away and help derive value out of the data they hold today, and convert it into meaningful portfolio positions.”
A 2020 survey by the Alternative Investment Management Association and SS&C found that traditional long/equity hedge funds typically crunch sentiment data, online reviews and payment information as part of their research and investment approach. Certain quant hedge funds – like equity market neutral and systematic macro – will often tap into weather patterns, satellite data and logistics metrics, along with certain web-crawled data. CTAs and trend-following strategies, meanwhile, use logistics data and consumer spending and lifestyle information, according to the survey findings.
“We use a lot of sentiment-based data now – we believe that there is a lot of information from the wisdom of crowds,” says Razvan Remsing, director of investment solutions at Aspect Capital. “For example, option market volatility surface data can provide information about the way participants are positioned or where the risks are perceived, without needing to trade the options themselves.
“We infer sentiment from option market data, or we infer sentiment outright from using, for instance, natural language processing to scrub blogs and websites to give us a far more frequent update on the broad sentiment around, say, US stock markets or G10 currencies.”
Remsing observes how this approach proved particularly successful in navigating the coronavirus-driven market meltdown of March 2020.
“The models that did well were those driven by these alt datasets which looked at sentiment, rather than fundamental data, which was often quite backward-looking. It was obvious that a lot of economic activity had stopped – but the traditional data hadn’t caught up with that; some of those data updates were on a weekly, or even monthly basis.”
In contrast, many of the systematic macro programmes utilise alt-data and were correctly positioned for the sizable market swings, he explains. “In a very dislocated environment, they worked very well. Once sentiment and equity markets came back, and when there was more confidence in the rally, these alt datasets were able to confirm that sooner than traditional models and traditional data.”
But as the types of data available continue to widen, dramatically recasting how hedge funds look to develop trading themes and manage portfolios, a key emerging challenge for hedge funds is how to determine the quality and relevance of datasets.
Bin Ren, founder and CEO, SigTech, a quant-focused information evaluation and analysis platform spun out from Brevan Howard Asset Management in 2019, believes the market for alternative data remains very fragmented, making it tough for hedge funds to access, evaluate and onboard new data sets. “Hedge funds are in an arms race to unlock the true value of financial data,” he observes.
Against that backdrop, managers running quantitative strategies now acknowledge that firms need to develop new expertise in this field, specifically around data sourcing and data testing. This, says Laurent Laloux, chief product officer at CFM, will help managers narrow down the pool of potential data providers and, critically, ensure that datasets can provide relevant insights into economies and financial markets.
“The first step is really knowing the ecosystem of providers – who the big players and the niche players are; which ones are doing a great job at creating the data and not introducing any statistical bias. The next step is looking at the data, and assessing that it does what it should,” he explains.
Dangers
Julien Messias, co-founder and head of R&D at Quantology, explains how his firm collects, stores and manages data, both traditional and alternative, itself internally.
“The point is to avoid getting flooded by data, as too much data kills the data itself,” Messias tells Hedgeweek. “Keep in mind that relevant data is quite rare, and this tends to be very expensive. Data is the fuel, the algorithm is the engine. Which out of the two is worth more?”
Comparing the process with finding a needle in a haystack, Laloux says establishing data quality has added a degree of complexity to the quant asset management process.
“When you do your research, there are a lot of statistical issues about assessing, first, whether the data is of good quality and whether you have a sufficiently representative cohort without any future information bias,” he says, pointing to the risks that stem from systems leveraging sub-quality data and mistakenly identifying alpha where there is none.
“That’s the key point – many of the alternative data providers are not necessarily attuned to the issues of modeling prediction, because of the look-ahead bias. For us, being experienced in the field, we realise the dangers and risks that come from even tiny look-ahead biases.
“But that’s also what makes it interesting and challenging – if you don’t know the dangers of finding a spurious correlation in data, you will just take the data, do your blind machine learning, have something which works on paper but ends up going in the opposite direction once you go live and trade the market. So you have to be extremely careful.”
This need for a cautious approach towards data has been further highlighted as managers look to capitalise on investors’ search for yield.
“The underlying hope is that the alternative data is not already priced in by markets and might bring some information that no investor already knew about by other means,” says Nicolas Gaussel, founding partner and CEO of Metori Capital Management. He says his firm looks to better quantify the degrees to which press articles or other informed sources of information can impact market moves, and better understand what is shaped by fundamentals and what is related to endogenous mechanisms. “A better understanding of those issues would shed new light on market mechanisms and improve the allocation of R&D resource,” he adds.
Gaussel warns that the maturity and efficiency of developed markets makes the hope of efficiently predicting financial markets via alternative datasets a “very ambitious endeavour.” “Investors have to be prepared to be disappointed. Costs might be higher than gains,” he adds.
Hedgeweek’s poll found that some 43 per cent of managers use alternative data for performance and alpha generation; 30 per cent use it to uncover or strengthen new investments, while 13 per cent use it for risk management and regulatory compliance.
But McHugh sees each element ultimately as part of the same challenge for asset managers.
“I don’t really see those things as separate, they all interplay,” he remarks. “When looking at a new company to invest in, you need to ensure performance and strengthening of a new investment idea is aligned to your mandate. You also need make sure that you are investing in companies that are aligned to your strategies, for instance, on ESG.”
Finbourne’s offering uses an API first approach, making it easier to take data in and out of the system through integrations, while its platform provides support across mission critical workflows, supporting CFOs with a more accurate, real-time picture of NAVs. At the same time, its secure cloud infrastructure can also be permissioned safely into the extended ecosystem, such as to investors.
“If you are going to look at data for fund performance and alpha generation, you still have your compliance to think about, as you do that. Having the tools to manage data and these mission-critical workflows in one platform is critical,” McHugh adds.
Read the full A Tech Revolution: How machines are reshaping hedge fund investment Insight Report here.