Innovation in execution: Morgan Stanley – Risk.net

“Clients’ businesses were becoming increasingly electronified. The ticket sizes were getting a lot smaller. It was a lot more granular. Our need for a more effective way of responding to those tickets was very important, especially given the fact that the European credit market is incredibly competitive,” says Rehan Latif, head of Europe, the Middle East and Africa (Emea) credit sales and trading and global head of emerging market credit trading at Morgan Stanley.

Instead of hiring more traders to process the requests, the bank turned to a bot to price and execute odd-lot flow, trades in smaller and more irregular sizes than typical orders.

And it does the job well. The algorithm may initially have been introduced in 2017 to help the desk keep up with a rise in enquiries, but it has earned its place on the team, and now trades almost half of the credit desk’s tickets across investment-grade and high-yield globally.

Latif calls the machine “an incredible trader”.

Morgan Stanley is not the only bank to use algos to trade corporate credit, but over the last year, it has established itself as one of the leaders, with its algo propelling it into the top three from as low as the seventh spot in platform rankings by volume for investment-grade credit, says Latif.

As a result, the bank has been giving it more headroom – one-fifth of the available balance sheet for the US business and one-quarter for Europe. That’s a significant increase from over a year ago, and a testament to the algo’s performance, says Phil Allison, head of fixed income automated trading with the bank. Morgan Stanley has declined to disclose more precise measures on the record.

Today, the algo streams firm, two-way prices on over 12,000 US bonds and 4,500 European bonds, in addition to responding to all client requests for odd-lot pricing within its parameters.

The changes have not been lost on the buy side, where firms have been yearning for better electronic pricing and liquidity. A trader at a large asset manager calls the bank “one of the best algo-pricers in the fixed income market”.

“We’ve seen them noticeably pick up in terms of their ability to systematically respond to us and leverage that algo-pricer. It allows us to do our business much more efficiently now,” the trader adds.

My traders are becoming a lot more efficient. They’re focusing on the bigger tickets – the more interesting opportunities that they should be focusing on

Rehan Latif, Morgan Stanley

The firm prices are a key selling point in a market where many quotes are only indicative and can change once an investor wants to trade.

One trader at a European asset manager says Morgan Stanley’s executable prices are “very helpful: first, finding liquidity for smaller trades; and, secondly, finding out where it should price – seeing accurate pricing on your screen”: “A lot of others … are not that accurate when updating prices. There is a lot of noise.”

Morgan Stanley’s algo takes in pricing data from trade reporting venues to generate a mid-price, which it then uses to stream quotes to various trading platforms.

Whereas human traders may still be better at pricing large orders where data may be sparse, a computer can more efficiently digest small-ticket pricing data and previous quote requests, “assimilating all that into a model, and providing a better price in the smaller tickets than a human would be capable of – however many of them you had”, says Allison.

The bot is fully autonomous for trades in investment-grade credit up to roughly $5 million, and trades in high-yield up to roughly $1 million.

“We would generally be looking to run a reasonably well-hedged portfolio. But certainly within that, there will be a set of bonds the algo wants to hold because it thinks they are important to the portfolio in some shape or form. It could be as a hedge, or it could be because the algo explicitly thinks they’re trading cheap,” says Allison.

Capable of trading and managing risk without human intervention, the algo performs better during market-wide events – when volatility surged at the end of 2018, for example – than it does when news affects a specific credit, says Allison. But the algo is essentially never turned off during market hours, says the bank.

The unemotional nature of an algo often makes it a better trader than its human colleagues – particularly on news-filled days. It doesn’t get spooked, it never feels trapped and it doesn’t panic.

“When a name is moving up and down fairly rapidly, it’s able to capture the bid-offer much better than a human trader, who might be hit or lifted on a name and be wrong-sided, and then try to protect that position. The algo would square out its position and then trade with that bid-offer – and be able to capture it,” says Latif.

In additional to helping the bank gain market share, the algo has also allowed the trading desk to do more with its available resources.

“My traders are becoming a lot more efficient. They’re focusing on the bigger tickets – the more interesting opportunities that they should be focusing on,” says Latif.