68 / 84
66
Bloomberg Businessweek
May 14, 2018
teams and getting his way. He never lost his temper, but his
mind, once set, was like granite. Benter was also unwilling to
budge. Their alliance was over. In a it of pique, Benter wrote
a line of code into the software that would stop it from func-
tioning after a given date—a digital time bomb—even though
he knew it would be trivial for Woods to ind and ix it later.
Woods would keep betting algorithmically on horses, Benter
was sure of that. He resolved that he would, too.
enter’s Las Vegas friends wouldn’t stake him at
horse racing, but they would at blackjack. He
took their money to Atlantic City and spent two
years managing a team of card counters, brood-
ing, and working on the racing model in his spare
time. In September 1988, having amassed a few hundred thou-
sand dollars, he returned to Hong Kong. Sure enough, Woods
was still there. The Australian had hired programmers and
mathematicians to develop Benter’s code and was making
money. He’d moved into a penthouse lat with a spectacular
view. Benter refused to speak to him.
Benter’s model required his undivided attention. It moni-
tored only about 20 inputs—just a fraction of the ininite fac-
tors that inluence a horse’s performance, from wind speed
to what it ate for breakfast. In pursuit of mathematical per-
fection, he became convinced that horses raced diferently
according to temperature, and when he learned that British
meteorologists kept an archive of Hong Kong weather data
in southwest England, he traveled there by plane and rail.
A bemused archivist led him to a dusty library basement,
where Benter copied years of igures into his notebook.
When he got back to Hong Kong, he entered the data into
his computers—and found it had no efect whatsoever on
race outcomes. Such was the scientiic process.
Other additions, such as the number of rest days since a
horse’s last race, were more successful, and in his irst year
after returning to Hong Kong, Benter won (as he recalls)
$600,000. The next racing season, ending in the summer
of 1990, he lost a little but was still up overall. He hired an
employee, Coladonato, who would stay with him for years,
and a rotating cast of consultants: independent gamblers,
journalists, analysts, coders, mathematicians. When the vol-
ume of bets rose, he recruited English-speaking Filipinos from
the ranks of the city’s housekeepers to relay his bets to the
Jockey Club’s Telebet phone lines, reading wagers at the rate
of eight a minute.
A breakthrough came when Benter hit on the idea of incor-
porating a data set hiding in plain sight: the Jockey Club’s
publicly available betting odds. Building his own set of odds
from scratch had been proitable, but he found that using
the public odds as a starting point and reining them with his
proprietary algorithm was dramatically more proitable. He
considered the move his single most important innovation,
and in the 1990-91 season, he said, he won about $3 million.
The following year the Hong Kong Jockey Club phoned
Benter at an oice he’d established in Happy Valley. He
winced, remembering the meaty hand of the Las Vegas pit
boss on his shoulder. But instead of threatening him, a Jockey
Club salesperson said, “You are one of our best customers.
What can we do to help you?” The club wasn’t a casino try-
ing to root out gamblers who regularly beat the house; its
incentive was to maximize betting activity so more revenue
was available for Hong Kong charities and the government.
Benter asked if it was possible to place his bets electronically
instead of over the phone. The Jockey Club agreed to install
what he called the “Big CIT”—a customer input terminal. He
ran a cable from his computers directly into the machine and
increased his betting.
Benter had achieved something without known precedent:
a kind of horse-racing hedge fund, and a quantitative one
at that, using probabilistic modeling to beat the market and
deliver returns to investors. Probably the only other one of
its kind was Woods’s operation, and Benter had written its
code base. Their returns kept growing. Woods made $10 mil-
lion in the 1994-95 season and bought a Rolls-Royce that he
never drove. Benter purchased a stake in a French vineyard.
It was impossible to keep their success secret, and they both
attracted employees and hangers-on, some of whom switched
back and forth between the Benter and Woods teams. One
was Bob Moore, a manic New Zealander whose passions were
cocaine and video analysis. He’d watch footage of past races
to identify horses that should have won but were bumped or
blocked and prevented from doing so. It worked as a kind of
bad-luck adjuster and made the algorithms more efective.
The computer-model crowd spent nights in a neighborhood
called Wan Chai—a honey pot of gaudy bars and topless danc-
ers that’s been described as “a wildly liberated Las Vegas.”
Benter’s Math (Radically Simplified)
Let’s say Seabiscuit is about to
run at Happy Valley Racecourse.
These numbers
suggest that if
the race were
conducted five
times, Seabiscuit would win once for
every four losses.
Benter’s
algorithms—
tracking inputs
like straight-line
speed, recuperation time, weather,
etc.—predict Seabiscuit will
actually win one of every
four
hypothetical races.
Public odds
Benter’s odds
4
to
1
3
to
1
A gambler who bet $1
five times could expect
to win $5—pointless.
That means Benter can
bet $1 four times to
win $5. That’s a profit
margin of 25 percent.
A small edge can
turn into big profit
when multiplied
across thousands
and thousands of
races. Benter says
his gambling systems
have made close to
$1 billion.