Overbond has won the award for Most Innovative AI in Trading Initiative in A-Team Group’s Innovation Awards 2023. The awards celebrate innovative projects and teams across vendor and practitioner communities that make use of new and emerging technologies to deliver high-value solutions for financial institutions in capital markets.
Overbond’s trading solution was selected as a winner by A-Team’s independent, highly experienced advisory board in collaboration with the company’s editorial team.
Vuk Magdelinic, CEO at Overbond, explains how the company’s innovative solution addresses the challenges faced by fixed income trading desks, its use cases, inclusion of AI technologies, and the benefits it can deliver. He also considers next-steps development.
A-Team: Vuk, please tell us about Overbond and the types of capital markets clients the company works with?
Vuk: Overbond provides AI-driven quantitative fixed income analytics for institutional trading desks. We provide data aggregation solutions and a suite of AI algorithms for bond pricing, trade workflow automation, market monitoring, and pre-trade signal generation.
We were founded in Toronto about seven years ago with just a handful of employees. Today, we have 30 employees, an international presence with an office in London and a research lab in Montreal. Our customers range from large sell-side liquidity providers to smaller buy-side institutions looking for niche investment opportunities.
A-Team: What challenges does your award-winning trading solution address?
Vuk: Traders must respond to hundreds of RFQs daily and are expected to provide quotes faster than ever. We help them automate a portion of these trades so they can respond to more RFQs in a day. This frees them up to concentrate on larger or more complex trades, which can lead them to be more profitable.
Overbond helps to fill the information gap left by the lack of a real-time consolidated tape for fixed income markets by aggregating data from multiple sources to create near-real-time desk-level consolidated tapes of bond transaction data.
The structure of the bond market also means traders are faced with pricing securities that they trade rarely. Overbond can help traders determine the best executable price and offer insight into the bond’s liquidity.
A-Team: What does Overbond offer to users and how does it work?
Vuk: We offer traders the ability to automate their bond trading partially or fully through our execution management system interoperability. This application displays real-time market data and analytics, allowing traders to execute with any counterparty or venue on the street quickly and seamlessly. It supports low-latency electronic and algorithmic trading, and provides pre-trade tools to maximize automation in various market risk environments. Using our EMS interoperability, traders can work the order using a variety of suitable automation protocols and order types, including smart order routing, auto-RFQ, and one-touch adjustment.
Our portfolio execution management system interoperability allows traders to simultaneously monitor their portfolio holdings in real time on an individual security level, and on the portfolio level. One interface allows seamless analysis of issuer curve, liquidity scoring and best-executable pricing so that portfolio managers and traders can easily identify rebalancing needs and execution strategy.
Our AI algorithms incorporate client trading style by training on client-specific execution history.
A-Team: What are the use cases of the Overbond solution?
Vuk: Our suite generates liquidity confidence scoring per bond CUSIP or ISIN that can segregate which RFQs are eligible for auto execution. For those that are ineligible, a price adjustment model called margin optimization, which is trained on the desk trading style, can help traders capture more trades and more profit.
Buy-side desks can use Overbond models to parse through historical data to identify those dealer counterparties, venues and protocols where the most efficient execution has been achieved and train on that historical record to recommend routing the trade optimally. In credit trading terms, these services are called transaction cost analysis and smart order routing.
A-Team: What makes your proposition particularly innovative?
Vuk: Fully automated trading requires calculating trade margin according to the desk’s trading style. To model margin effectively, we first need to model for changes in the amount of a bond issue available to trade and for market risk conditions at the time of the trade. These are complicated mathematical and statistical problems, and by developing innovative solutions for them, we have significantly enhanced the precision of automated trading.
To model the amount of a bond available, we developed a way to determine the total market capacity of the bond being traded, which is the portion of the outstanding amount of the bond not tied up by buy-and-hold accounts unwilling to sell. When a trade exceeds this amount, there won’t be enough of the bond readily available to fill the order, and traders will likely increase the margin.
To train the model to adjust the margin to total market capacity during a trade, and according to the desk practices, we apply AI learning and prediction to six or more months of historical data to discern trade size and volume patterns.
To train the margin model to adapt to market risk conditions at the time of the trade, we use a technique used in medical research called case-mix adjusted cluster analysis, which groups similar observations into a group or cluster. This technique allows us to separate the impact of market conditions on historical margin levels. It is applied to a historical series of post-trade data points from the desk to determine if trades were made in a normal, heightened, or low-risk environment.
This allows the model to discern the margin patterns attributable to these markets and adjust according to current market conditions. Margin in the automated trading system becomes dynamic and adjusts to the current market conditions as the trader would.
A-Team: What are the benefits of implementing Overbond?
Vuk: Sell-side traders can increase their hit rate or P&L, and buy-side traders can reduce transaction costs.
A-Team: Please talk us through a brief user case study
Vuk: We recently worked with a large European bank. It trades in EUR and USD, and wanted to increase confidence in its pricing, improve its execution speed, and increase its level of trade automation.
We integrated our suite in two phases. First, we tested interoperability and generated a sample of COBI prices on which we performed an error and difference analysis against other pricing sources. Then, when we were comfortable that the system was performing as expected, we trained the model to the desk trading style, including margin optimization, and back-tested the model to compare the trading performance of the Overbond model with the record of the trading desk.
A-Team: How will you develop and innovate Overbond over the next year?
Vuk: Through our research in quantitative analytics, we are continuously generating new product ideas and enhancing existing models. This year is no different. We are working on several exciting innovations that we plan to introduce in Q3. Clients should look forward to enhanced pricing and analytics coverage of the fixed income universe, and tools for swifter and smarter trade execution.
A-Team: Finally, what does winning the award for most innovative AI in trading initiative mean to Overbond?
Vuk: It is always fantastic to receive recognition. Receiving this award confirms that Overbond continues to gain momentum, and it will motivate everyone at the firm to continue developing market-leading fixed income trading innovations.
Subscribe to our newsletter