Overbond Taps AI for Primary Bond Market
Although machine learning has made incredible inroads in the secondary securities markets, Canadian startup Overbond has brought the discipline to the primary corporate bond market.
The vendor’s platform uses supervised machine learning to price potential issuances, score the likelihood of issuances, and identify potential institutional investors who might be interested in the issuance, Vuk Magdelinic, CEO of Overbond, told Markets Media.
“We do this for the quality of results, which are bulletproof,” he said. “The algos also learn faster when there is an expert in the loop.”
Overbond’s Corporate Bond Intelligence (COBI) algorithm uses data sourced from Thomson Reuters’ big data suite to price bond offerings a day or two before to issuance. “This spring, we had COBI price 30 deals, and it came within two basis points of the final price for each deal,” said Magdelinic.
The algorithm also provides an “R-Score” that represents the likelihood that a corporation will issue a new bond based on the company’s balance sheets, corporate actions, and regulatory filings.
Finally, the platform examines institutional investor’s current and previous bond holdings to identify parties who might want to take part in the issuance.
Overbond cut its teeth on the Canadian corporate bond market earlier this year and plans to add functionality for the US markets when it officially launches in September.
“The same macro trends that affect the US corporate bond market also affect the Canadian corporate bond market,” said Magdelinic. “The Canadian market is smaller, which is good for a growing company.”
Overbond has completed 90% of the necessary changes that the vendor needs to make to have the platform support the US primary market, he added. “The rest should be completed by our hard launch.”
Since its soft launch, there approximately 150 corporate issuers and institutional investors on the platform. “We like to keep it 50-50 because it helps balance the supply with the demand,” he said.
From a technological perspective, Overbond developed its platform using Amazon Web Services and SalesForce.com’s Heroku cloud application platform.
Magdelinic expects institutional investors to use the platform as a dedicated application on their desktops. However, the vendor has developed its offering using OpenFin architecture, which could ease its integration into other OpenFin-based platforms. “We like to be agnostic,” he said.
Electronification of the municipal bond market also presents a large opportunity.
The success of Northbound trading showed electronic execution is way forward for the bond market.
IRS trading volumes have fragmented without an equivalence agreement.
Increased electronification has created useable and accessible real-time and historic trade data.
Members are evaluating payment-versus-payment for currencies not yet eligible for CLSSettlement.