ING is in advanced trials for a prototype technology tool called Katana Lens, aimed at asset manager clients that invest in emerging market (EM) bonds.
It may prove important in two ways: as a potential proof of the usefulness of artificial intelligence (AI) and predictive analytics in generating alpha in bond markets, and as a marker for how far banks can evolve into technology providers to their clients.
In December 2017, ING announced – to almost no fanfare – the first version of Katana, a tool developed within ING wholesale banking advanced analytics to help its own bond traders make sharper prices in response to institutional investors’ request for quotes (RFQs).
Santiago Braje, global head of credit trading at ING, says: “We aimed to bring in real time all the information a trader might want before responding to an RFQ.
“Obviously, it’s important to have the recent history of prices in liquid bonds, but also it would be useful to combine that with internal data, for example what prices our traders may have previously quoted on the same bonds when that quote didn’t lead to a trade and by how much we may have lost out to the order that eventually went through.”
He adds: “It would be good then to overlay all that with information on present market conditions and with analytics that help predict what might be the winning quote.”
Santiago Braje, global head of credit trading at ING
The outcome is a predicted price at which the client will trade that also takes account of where competitors will likely quote in that specific bond, given prevailing market conditions.
Katana has now proved itself internally at ING.
Braje says: “We have analyzed data for traders using the tool and not using it, and have found that it improves both our hit rates – how often traders win the deal – and their trading cost efficiency – the difference between our price and the second-best quote – by around 20% in each case.”
ING is now making a different version, called Katana Lens, available to asset management clients, after partnering with PGGM – the Dutch pension administration cooperative – to focus on a relative value discovery tool.
Bond investors typically grow used to thinking about associated pairs of bonds and look for periods when they diverge from the traditional value relationship and one trades cheaper than usual to the other. That may be a good moment to sell the expensive bond, buy the cheap one and wait for the mean reversion to realise a gain.
For years in EM credit, investors monitored Argentina versus Brazil, for example, but if an asset manager’s investment team is surveying a universe of 2,000 bonds in EM credit globally, that’s actually two million pairs potentially to trade. An asset manager covering 3,000 bonds in investment grade credit globally has potentially 4.5 million pairs to consider.
The first challenge is to work out which of those possible trade ideas are worth analyzing in depth not just by looking at prices but also by combining changing market data with a lot of unstructured data, for example from conversations between portfolio managers and issuers as well as the sell side.
A lot of banks are making big investments in technology, but from our understanding the list of those coming up with tools like this is not a long one- Santiago Braje, ING
When an asset manager is seeing a lot of suggestions from dealers to trade in a certain bond, then it may well be one to look at, but perhaps in combination with a non-traditional paired buy or sell.
Braje says: “There may be a lot of news around Brazil one day or Russia the next, but there is limited value in looking at one bond in isolation or just against the overall market index.
“One of the valuable things our Katana Lens algorithm does is throw up for asset managers trade opportunities in pairs of bonds they don’t normally consider, either because they have never thought of them as a pair, none of their providers of trade ideas has suggested them or they don’t have a tool that spots deviations from the trend quickly enough to take advantage of.”
This is what AI can bring to financial markets. It doesn’t just analyze the price history of Argentina versus Brazil. From an addressable universe of 2,000 bonds, it can analyze everything against everything else and throw out suggestions for particular trades within a framework of how they look against every other possible idea.
“Our clients give us feedback that is very encouraging, that it throws up trade ideas to put on positions at interesting levels that are profitable and which have improved their performance,” Braje reports.
“Just as importantly, it does so in different market conditions. Its success is not correlated to overall market direction. And this alpha capture accords with our own back-testing, though all back-testing is theoretical, of course.”
The big question it seems to Euromoney is how does ING get paid for providing this technology to investors?
Braje tells Euromoney: “Although Katana Lens is fully functional, we still consider it a prototype for now. How we may eventually get paid for providing this is not yet defined.
“We are exploring a range of methods, from integrating it into our current client offering all the way to offering it as a stand-alone product with its own fee income, and various combinations of those options.”
Does he not worry that other banks will seize this new business opportunity?
“A lot of banks are making big investments in technology, but from our understanding the list of those coming up with tools like this is not a long one,” says Braje.
“Of the 30 or so investors we have shown Katana Lens to, 80% have nothing like it and none has told us that any other banks are offering anything similar. The other 20% are using tools they have built internally themselves.”