plant-crack-resilence-asphalt-780.jpg

As June drew to a close, with equities selling off on fears of a second wave in China and increasing cases in the US and Germany, the new supply of investment-grade and high-yield corporate bonds was slowing down sharply, after the rush to issue into the great rally driven by central bank buying.

Analysts were sending notes of caution.

Matthew_James_Citi-160x186.png

Matthew James,
Citi

Matthew James, head of global spread products research at Citi, told investors: “We remain positively inclined across spread products as we move into summer, given the tail-end of the QE wave. However, markets could get choppy as liquidity drops into the holidays.”

Suggesting that the surge of central bank liquidity has crested, James said: “We believe that investors should take advantage over the next few weeks to tidy up portfolios – lighten up on less-confident positions, bolster liquidity buffers and tactically add to core views.”

It’s not just overall credit spreads that have been extraordinarily volatile during the past four months. So too have movements in relative value between pairs of bonds that typically trade within a range of each other.

When one looks relatively cheap and the other expensive, that’s a classic meal ticket for bond traders. Sell the expensive bond, buy the cheap and wait for mean reversion to close the linked trades.

However, amid a regime change, historic correlations can break down.

Some credits may start to look cheap because their balance sheets are less robust than those of competitors in the same sector and, if they continue to struggle, their bonds may carry on getting cheaper and never mean revert.

santiago-braje-katana_ing-160x186

Santiago Braje,
Katana

Santiago Braje, chief executive of Katana, a fintech that uses advanced machine learning to identify relative value in bond markets, tells Euromoney: “We have seen extraordinary dislocations in credit where spreads widened by a factor of four times in a couple of weeks in March and, while all spreads widened in sync, so too did the spread differences between many pairs.”

Since that initial sell off, markets have seen two further phases.

Braje says: “Next came discrimination, when some credits stabilized but others continued to widen. Then in the past six to eight weeks we have seen a clear reversion trend.

“Spreads, for example, on European investment grade have come down and spread differences between correlated pairs have also reduced. In emerging markets, overall spreads are still much higher than they were, but we have seen a reversion in spread differences and more normalized relative value.”

What follows?

It’s always so hard to make predictions, especially about the future.

Really big data

Incubated within ING, where it developed a prototype for the bank’s own credit traders, Katana provides artificial intelligence analysis of market price data across 10,000 emerging market and European investment grade bonds, scanning 40 million potential pairs most of which are not usually linked in buy and sell trades.

It combines this with unstructured data, for example from dealers’ records of filled and unfilled requests for quotes from investors.

Katana was spun out in November 2019. ING Ventures is just one minority shareholder.

Katana’s algorithms now provide trade ideas to some of the world’s largest asset managers including insurance companies and pension funds across the US, Europe and UK.

The firm has also opened its platform to subscribers from some of the large sell-side firms.

This is very much a human-in-the-loop application designed to enhance the expertise of the portfolio manager or trader 

 – Santiago Braje, Katana

Braje, who founded the business when he headed global credit trading for ING, explains why it made sense to scale it as an independent.

“As a single dealer, you are somewhat constrained in only taking unstructured data from what your own traders see and hear,” he says. “Outside, we have access to much wider pools of data.”

The firm is close to announcing a new data partnership, which Braje says will expand its coverage to 130,000 corporate, sovereign, government and agency bonds across investment grade and high yield.

That gives it correlations to map between 500 million potential pairs, now covering the US credit markets and Asia-Pacific, as well as Europe and emerging markets.

“This really is big data,” he says.

Braje adds: “We have also significantly improved the algorithm, which has had enormous volumes of new data to learn from that we have been interacting with in real time. At the core, it accurately identified dislocations and we are now releasing a major upgrade which is even more accurate.”

Human in the loop

Perhaps counter-intuitively for the chief executive of a machine-learning fintech, Braje stresses that algos are simply a tool to help portfolio managers, not a replacement for them.

“This is very much a human-in-the-loop application designed to enhance the expertise of the portfolio manager or trader,” he says.

“The algo is very good at processing huge amounts of data at high speed and spotting patterns, but that’s only one element to investment decisions. It can’t understand the wider market context, particular credit fundamentals or the background macro picture. The portfolio manager brings all that.”

Katana has upgraded the platform to allow for greater customization so that individual subscribers can load their own watchlists of bonds that are already in their portfolios or that fit their particular investment mandates and that they are interested in.

“They can do that securely without us or anyone else seeing their watchlists,” Braje says.

There’s always a next step. Beyond the new data-partnership now being finalized, what is Katana’s?

“Our next focus will be on liquidity, which is very important for our investor clients,” Braje says. “Recent events have brought the constraints on liquidity even more to the fore. The debt market has doubled in size in the last 10 years and is now growing fast, while dealer inventories have shrunk.”

Integrating real-time liquidity data into the discovery of trade ideas might transform the efficiency of the whole process.

Braje says: “We are in discussions with multiple other parties, both data providers and trading venues, over potential integrations and partnerships.”

This post was originally published on this site