Overbond, the API-based credit trading automation and execution service, has entered into a data sharing and redistribution agreement with MTS BondsPro, the all-to-all fixed income trading platform. By partnering with MTS BondsPro, Overbond will enhance its AI-generated fixed income data feeds and automated trading solutions by integrating an additional 10 million price updates daily, across 25,000 investment grade, high yield and emerging market bonds. In turn, MTS BondsPro clients will gain access to Overbond’s fixed income liquidity confidence scores and best executable pricing.
Earlier in the summer Overbond secured additional funding from Fitch Ventures, stating that the company planned to integrate new data sources to expand the coverage of its AI models, and provide clients with enhanced AI trade automation solutions.
“To date, we’ve been able to aggregate data from real-time streaming pre-trade composites, but although those data feeds have quotes and ticks, size typically doesn’t get disclosed,” says Vuk Magdelinic, CEO of Overbond. “We’ve worked really hard over the last couple of years to also include post trade data, by aggregating Euroclear and Clearstream settlement data, as well as data from APAs. By adding data from all-to-all venues like MTS BondsPro, which provides firm quotations including size, and is an uncorrelated data set to an RFQ mini auction, or to settlement post trade data, we improve the precision and coverage of our outputs.”
The company has also this week released a new an AI-powered margin optimisation function into its existing automated trading system. By training the automated system to optimise their hit ratio according to their desired parameters, sell-side traders can increase the number of trade inquiries that they can respond to without manual intervention, thereby avoiding workflow bottlenecks.
The Overbond margin optimisation model incorporates variables that give insight into security, issuer and macro-level market risk and ensure that the automated margin is sensitive to intra-day risk movements. This data is collected from data vendors such as TRACE and includes bond-specific data such as coupon and amount outstanding, issuer-specific data such as quote counts and the volatility of the mid-price for the issuer, and sector-specific data such as the volatility of the bid-ask spread.
“This new model is geared for the category of RFQs that are eligible for auto-response, where you really don’t want to introduce a trader to manually margin the price,” says Magdelinic. “You want to auto margin based on a combination of the market parameters and your internal firm or desk-specific parameters. And you want to be able to optimise that by looking at things like the execution record over the last two years for similar bonds, from a similar client, on a similar venue, in a similar size, with a similar risk situation, and so on. So we’ve engineered a model that is end-to-end, zero touch, automated margining, where you can include a desired hit ratio within your auto response, which is key for maximising P&L.”
He adds: “Because this is an AI self-learning model, it trains itself on your historical executions from the last two years. Then you can control it further by adjusting thresholds and parameters.”
Subscribe to our newsletter
