The initiative could simplify the collection and aggregation of fixed income data to enhance trading.
Bloomberg has launched Early Alerts, which predicts changes in US dollar corporate investment grade and high yield securities, and aims to expand the machine learning model to other currencies and securities.
Early Alerts, which was launched this month, uses Bloomberg’s proprietary library of fixed income data with machine learning models to develop predictive insights for more than 16,000 US dollar denominated investment grade and high-yield corporate bonds. The model generates scores over 1-day, 5-day, and 20-day horizons. The higher the score, the greater the probability that a corporate bond will have a significant credit spread tightening or widening.
Brad Foster, global head of enterprise content at Bloomberg, told Markets Media: “We would like to launch one or two more currencies before the end of the year.”
The model could also be extended to other securities such as government bonds or municipal bonds.
Foster continued that Bloomberg has deep and unique datasets, including pricing service BVAL, that enable a robust back-testing process to test the accuracy of the signals produced.
“As an example of Early Alerts’ demonstrated out-of-sample historical accuracy, bonds and dates with scores of 0.4 and above have seen 80% or more accuracy in identifying spread widening vs. spread tightening over the following five days from the beginning of 2015 through second quarter of 2019,” he added.
Early Alerts is available to Bloomberg Enterprise Data clients. Foster said the model is easy for both buy-side and sell-side clients to use and enables more efficient trade execution, warehousing of risk and rebalancing of portfolios.
Foster added that the model feeds into Bloomberg’s “one data” proposition.
“We want to be the one-data provide for our clients,” he said. “We want to offer simplified distribution with data that is ready to use by humans or machines.”
Transforming fixed income data
Foster added: “Fixed income markets, including corporate credit, are primed for the type of quantitative modeling, tooling and predictive power already available in other asset classes.”
Defining Fixed-Income Data, a report from consultancy Greenwich Associate, said a complete view of any fixed-income market is not likely soon but more can be done to simplify the collection and aggregation of the required data.
Kevin McPartland, head of research for market structure and technology at Greenwich Associates, said in the report aggregating both pre- and post-trade data will increase transparency.
“Trading venues such as Bloomberg, MarketAxess and Tradeweb have proven to be particularly adept at providing their clients with unique pre- and post-trade information based on that client’s trading activity and anonymized activity from the entire venue,” he added. “However, one is unlikely to share data with the other, and as such, a complete view remains elusive.”
Bloomberg Enterprise, which produced Early Alerts, is separate from the firm’s trading platform.
McPartland continued that over the last year some platforms have opened up to allow third parties to ingest their market data, allowing those end users to both aggregate and analyze multiple data sources in one environment.
“This enables portfolio managers and traders on the buy and sell sides to run more robust transaction cost analysis calculations, counterparty analysis and risk assessments of their portfolios,” he said. “The result of such analysis is often more trading, which is then directed back to the underlying platforms, leaving everyone better off. This process, of course, is much easier said than done.”