For many years Michael Platt, the former JPMorgan swap trader who founded BlueCrest Capital Management, was a relatively low-key investor who was known in the industry as a hugely profitable dealer but was otherwise able to go about his business unburdened by public attention.
He is now developing an alarming propensity for headline generation, however, even though he converted BlueCrest from a hedge fund to a family office in late 2015 and could in theory have retreated even further from the limelight.
In December 2019, an odd video emerged of Platt having a conversation with a New York cab driver in which he repeatedly described himself as “the highest-earning person in the world in finance”.
This generated intense debate over whether Platt was in on a joke with the cab driver – who posts videos of conversations with celebrities – or was simply boasting after a long evening spending some of his well-earned money.
For all Platt’s outward sophistication – as a financial engineer and art patron who made a cameo on the US television show Billions – he is also a trader who grew up in the UK, so the latter possibility cannot be ruled out.
BlueCrest made year-end headlines again in 2020, with its December agreement to pay $170 million to settle US Securities and Exchange Commission (SEC) charges that it misled investors by using an underperforming algorithm to pay them less than principals generated from their own proprietary fund.
The SEC on December 8 laid out the details of what it said was conduct from October 2011 to December 2015 that harmed investors in BlueCrest’s flagship fund BlueCrest Capital International (BCI).
In 2011, BlueCrest created a leveraged proprietary fund – BSMA – to trade the personal capital of its principals. As with BCI, the largest capital allocations for BSMA involved rates and relative value dealing – the trading strategies where Platt made his name, first at JPMorgan, then with his own fund.
Did this artificial intelligence tool simply duplicate the performance of the experienced traders? It did not, according to the SEC
The twist was that BlueCrest first transferred existing dealers from BCI to BSMA, then began to replace their capital allocations with what the SEC described as “a semi-systematic trading system – at its core, a replication algorithm – called Rates Management Trading (“RMT”).”
Did this artificial intelligence tool simply duplicate the performance of the experienced traders – 21 overall – who had left to concentrate on increasing their own earnings?
It did not, according to the SEC.
Problems
“RMT had lower P&L and greater volatility than the corresponding live traders and, therefore, underperformed them,” the regulator said.
A practice of simply duplicating deals by live traders a day later via the algorithm was a particular problem.
“RMT was not configured or intended to capture all of the trading activity of all of BlueCrest’s live rates and RV traders because certain strategies and asset classes were unsuited to algorithmic strategies and next-day execution,” the SEC continued.
And how did the algo cope with the unusual trading pattern in June 2013 now known as the “taper tantrum”, when Treasury yields spiked in reaction to Federal Reserve messaging about a retreat from quantitative easing?
Again, poorly.
“In June 2013, RMT’s loss was the largest of any capital unit in both BCI and BSMA by several orders of magnitude,” the SEC said.
Many of the charges that the SEC levelled against BlueCrest were related to failures of disclosure to investors in BCI, and there was evidence that managers at the fund were surprised by some details of the failure of the algorithm.
The relevant dealing was also more than five years ago, before Platt decided to convert the fund to a family office.
But there is a lesson for the future from the details of the SEC charges: any artificial intelligence applications in the more complex corners of the markets will only work if they are fully aligned with the interests of their human overlords.
