Big data & AI pose challenges for quant investors

Big data & AI pose challenges for quant investors

03-07-2018 | インサイト

The advent of big data and artificial intelligence (AI) has emerged as a major game-changer for the financial industry. In the asset management sector, although the adoption of these innovations is still in its infancy, a growing number of players are examining how they can be used to design better investment strategies.

  • David Blitz
    Head of Quant Research
  • Rob van Bommel
    van Bommel
    Managing Director, Portfolio Manager


  • Big data and AI could enable investors to take better decisions
  • Data fitting and short histories raise robustness concerns
  • We apply our cautious pioneering philosophy to this field

But while it is true that improvements in data analytics may bring valuable insights for investors, there is little hard evidence that big data and artificial intelligence actually work in practice, at least for now. Concrete cases of investment managers able to deliver consistent outperformance in real life using these techniques are still absent.

A solid investment strategy requires extensive empirical testing and falsification on broad samples of data and over long periods of time, but the evidence for big data and artificial intelligence is largely anecdotal. Strategies based on big data and artificial intelligence may also lack the necessary backing by a clear economic rationale. In fact, most investment ideas solely rely on paper back-test results, which should always be considered with caution.

We always strive to identify factors that are rewarded with superior risk-adjusted performance over the longer term. We also look beyond mere statistical patterns and aim to understand the economic drivers of returns.


Technical challenges

As a result, managing money based on an untested AI algorithm that scrutinizes exotic datasets raises a lot of concerns, even though big data and artificial intelligence have become a popular discussion topic among investors. Moreover, these innovations pose a number of challenges for asset managers looking to incorporate them into their investment process.

The first of these technical challenges is that ‘big data datasets’ generally have a short or even very short history. The second issue has to do with the lack of breadth, or at least the very fragmentary nature, of most big data signals. Another concern is that many signals provided by big data and AI tend to be very short-term focused. Finally, high quality datasets are not easy to obtain and can be very expensive.

For now, we see more potential in analyzing data signals available from credit, option and lending markets, using proven and transparent techniques

Given all the caveats mentioned above, we consider the current trend around big data and AI as a very interesting development, but one that should be treated with caution. For now, we see more potential in analyzing data signals available from credit, option and lending markets, using proven and transparent techniques, than in analyzing exotic big data variables with complex algorithms.

At the same time, we acknowledge the potential disruption this kind of innovation may lead to in the future. In the past, the datasets that are currently widely used by quantitative asset managers were subject to similar issues as big data today. Over the years, the quality, breadth and history of these datasets improved, and they became usable. Now, as time passes and more data becomes available, big data will probably also become increasingly usable.

Not necessarily one or the other

Finally, the fundamental issue for investors may not necessarily be about choosing between one type of data or the other. There is a wide array of possibilities between sticking to traditional price and financial statement information, and solely relying on things like satellite imagery of parking slots. For example, big data and AI signals could be very useful to fundamental credit and equity analysts. This would feed through into our quantitative strategies, that take analyst revisions into account. In this case, we would be using big data and AI information in an indirect manner.


当資料は情報提供を目的として、Robeco Institutional Asset Management B.V.が作成した英文資料、もしくはその英文資料をロベコ・ジャパン株式会社が翻訳したものです。資料中の個別の金融商品の売買の勧誘や推奨等を目的とするものではありません。記載された情報は十分信頼できるものであると考えておりますが、その正確性、完全性を保証するものではありません。意見や見通しはあくまで作成日における弊社の判断に基づくものであり、今後予告なしに変更されることがあります。運用状況、市場動向、意見等は、過去の一時点あるいは過去の一定期間についてのものであり、過去の実績は将来の運用成果を保証または示唆するものではありません。また、記載された投資方針・戦略等は全ての投資家の皆様に適合するとは限りません。当資料は法律、税務、会計面での助言の提供を意図するものではありません。




商号等: ロベコ・ジャパン株式会社  金融商品取引業者 関東財務局長(金商)第2780号

加入協会: 一般社団法人 日本投資顧問業協会