The sinking of tech billionaire Mike Lynch’s yacht in a freak storm off the Sicilian coast last week certainly has to rank among the most bizarre fatal celebrity accidents in years. There was the weird coincidence that Lynch had just gotten acquitted after a yearslong legal battle over a multibillion-dollar fraud; the eerie synchrony of the same-day death of his co-defendant after being struck by car while jogging; the fact that the $40 million vessel had been described as virtually unsinkable; the fact no vessel that size had been sunk by a waterspout in centuries; and the fact that the area where it struck is not known for waterspouts. But perhaps the wildest thing about the whole saga is the yacht’s name, Bayesian.
It refers to a method of statistical calculation that was originally devised by an 18th-century Presbyterian minister, Thomas Bayes. Lynch named his yacht after Bayes’s method in recognition of its role in building his fortune. In short, he had honored a method of calculating probabilities — only to be killed aboard its namesake by an accumulation of wildly off-the-chart improbabilities.
“The irony is tragic,” says British science journalist Tom Chivers, who writes for Semafor and published a book this year on Bayesian statistics called Everything Is Predictable. “It’s not a subtle irony. Bayes is the maths of prediction. This sequence of events is just spectacularly unlikely.”
It’s as if Charles Kane had been crushed to death under a giant rosebud.
How did we get here? In statistics, total probability is calculated by multiplying together all the component probabilities. So let’s look at the individual parts of the case.
The name
Mike Lynch was something of a black-swan event in himself. Until he came along, the U.K. had never before had a software billionaire. The son of a fireman and a nurse who had immigrated from Ireland, Lynch grew up in England, then rose above his humble roots by attending Cambridge University. After graduating, he returned to Cambridge to earn a Ph.D. in signal processing, then stayed on to conduct postdoctoral research in neural networks, the data-processing technology that underlies machine learning. In 1990, while still a student, he founded his first company at the age of 25 with a loan of £2,000 from an eccentric acquaintance he made in a bar.
His research had led him to understand that by using neural networks computers could sift through unstructured data to find information that a user wanted. To do this, a program needed to assemble a vast number of interrelated probabilities and then update these probabilities as new information came in using Bayes’s formula. In 1996, he founded a separate company, Autonomy, to do just that.
Lynch was way ahead of the curve in seeing the technology’s commercial potential. At the time, the field was struggling through a so-called AI winter: Research had stalled, and the processors hadn’t yet been developed that would allow for the creation of the large language models of today. But Lynch’s foresight served him well. By 2000, the company was worth more than a billion dollars and Lynch was being lauded as the U.K.’s answer to Bill Gates. In 2011, he sold Autonomy to Hewlett Packard for $11 billion. It was a tremendous coup for a former scholarship boy.
He used some of the money to buy a 184-foot long yacht named Salute from a Dutch property developer named John Groenewoud, who had bought it new for nearly $40 million in 2008. At the time it was built, the yacht’s 237-foot-high aluminum mast was the tallest in the world. Lynch renamed it after the kind of statistics that had made his fortune: Bayesian.
The rap
Lynch’s luck soon turned. After the deal closed, HP realized that Autonomy’s revenues were largely fictitious and that the deal it had staked so much on was in fact a dog — what one analyst later dubbed “the worst, most value-destroying deal in the history of corporate America.” HP wrote down the investment by $8.8 billion. The company fired its CEO, and the company broke up into two parts.
A lot of people were pissed. HP blamed “willful effort on behalf of certain former Autonomy employees to inflate the underlying financial metrics of the company in order to mislead investors.” These suspicions were the basis for a series of legal actions, and in 2019 U.S. prosecutors indicted Lynch and Stephen Chamberlain, Autonomy’s former vice-president of finance, on fraud charges that carried a maximum penalty of 25 years in prison. Lynch’s own lawyers called it “one of Silicon Valley’s biggest-ever fraud cases.”
The trial took a long time to unfold in part because Lynch waged a yearslong battle against extradition from the U.K. to the United States, a battle he finally lost in 2023. This past March, his trial began in San Francisco. The New York Times noted that “Mr. Lynch’s odds do not look good.” (Former Autonomy CFO Sushovan Hussain had already been convicted on similar charges and sentenced to five years in prison.) But Lynch and Chamberlain beat the odds. On June 6, a jury found both men not guilty on all counts. “Lynch’s win is extremely unusual in federal criminal cases,” as the New York Post reported. “In fact, only less than 1% of federal cases ended in acquittal in 2022, according to the Pew Research Center.”
The sinking
Lynch decided to celebrate with a Mediterranean cruise aboard Bayesian. The yacht was to carry its full complement of ten crew and 12 guests. Invited along with Lynch and his 18-year-old daughter, Hannah, were several people who had helped secure his freedom, including lawyer Christopher Morvillo of the firm Clifford Chance and Jonathan Bloomer, the chairman of Morgan Stanley International, who had testified to Lynch’s good character at his trial. Not on the guest list was his co-defendant. Three days after Bayesian set sail, Chamberlain was back in England and jogging near Cambridge when he was struck by a car. Severely injured, he was taken to the hospital and put on life support.
Bayesian was off on a meandering course through the Tyrrhenian Sea. It wandered among the Aeolian Islands of Pecorini, Malfa, and Lipari, then stopped at the port of Milazzo on August 14. From there, it sailed westward along the northern coast of Sicily toward Palermo. On the evening of Sunday, August 18, it anchored about 300 yards from the harbor of Porticello, where severe thunderstorms had been forecast. “We all knew that a storm was coming and that during that night it was better to keep the boats inside the port,” a local fisherman later told the BBC.
Under certain very specific conditions, thunderstorms can yield tornadoes. When these occur over water, they are called waterspouts, but in every other regard the phenomena are identical, says Wade Szilagi, director of the International Centre for Waterspout Research. Tornadoes are measured according to the Enhanced Fujita Scale, with EF5 being the strongest with winds of up to 300 miles per hour.
Like tornadoes, waterspouts are more likely to be found in areas where the meteorological conditions are favorable for their development, such as the Florida Keys and Lake Erie. As for the region where the Bayesian sank, “I wouldn’t say that this is a hot spot” for waterspouts, says Szilagi.
Shortly before dawn the vessel was apparently struck by a waterspout of unknown size and intensity while it was riding at anchor. Violent gusts slammed the yacht from side to side, forward and back, before ripping it from its anchorage. Bayesian tipped on its side, filled with water, and sank below the waves just 15 minutes after the storm began. While nine of the crew and four of the passengers escaped, six of the passengers remained trapped in the hull, including Mike and Hannah Lynch, Jonathan Bloomer and his wife, Judy, and Chris Morvillo and his wife, Neda. Their bodies were later recovered by divers. The body of the yacht’s cook, Recaldo Thomas, was found floating on the surface.
Even given the intensity of the storm, the speed with which Bayesian sank was shocking, nearly incomprehensible, to the yachting community. “I have never seen a vessel of this size go down so quickly,” Karsten Borner, the captain of a sailing ship that came to the survivors’ aid, told the Guardian. “Within a few minutes, there was nothing left.”
In interviews with Italian media, Giovanni Costantino, chairman of the company that built the Bayesian, said that his firm’s sailing yachts are “unsinkable” and among “the safest boats in the world.”
It won’t be entirely clear what happened until Italian authorities are able to complete their inquiry, but many, including Costantino, have speculated that the crew failed to prepare the yacht for the storm, for example by remaining at anchor and leaving hatches open to the weather. Italian prosecutors have opened a manslaughter investigation into the captain, James Cutfield, who survived.
Even assuming gross negligence on the part of the crew, the sinking of a vessel as large and robust as the Bayesian by a waterspout is vanishingly improbable. Such events are simply exceedingly rare, and none has been so deadly in centuries. In 2004, a waterspout in Greece picked up a boat and flung it onto a 10-year-old boy, killing him. And last year, a waterspout struck a 50-foot boat on Italy’s Lake Maggiore, killing four. The last reported time a waterspout claimed as many victims as died on the Bayesian was back in the 16th century, when one struck a crowded harbor in Malta, killing hundreds.
The same day that the Bayesian sank, U.K. police reported that Chamberlain had died in hospital of his injuries. He became one of the 165 pedestrians between the age of 25 and 59, out of a U.K. population of 42 million, to be killed each year by vehicles.
The bizarreness of this string of unlikelihoods has been lost on no one. “This sequence of events is just spectacularly unlikely,” Chivers notes. “There is this Bayesian-probability aspect to it, this building of unlikeliness upon unlikeliness.”