67 / 84
65
Nevada at Las Vegas, which kept a special collection on gam-
ing. Buried in stacks of periodicals and manuscripts, he found
what he was looking for—an academic paper titled “Searching
for Positive Returns at the Track: A Multinomial Logit Model
for Handicapping Horse Races.” Benter sat down to read it,
and when he was done he read it again.
The paper argued that a horse’s success or failure was
the result of factors that could be quantiied probabilistically.
Take variables—straight-line speed, size, winning record, the
skill of the jockey—weight them, and presto! Out comes a pre-
diction of the horse’s chances. More variables, better vari-
ables, and iner weightings improve the predictions. The
authors weren’t sure it was possible to make money using
the strategy and, being mostly interested in statistical mod-
els, didn’t try hard to ind out. “There appears to be room
for some optimism,” they concluded.
Benter taught himself advanced statistics and learned to
write software on an early PC with a green-and-black screen.
Meanwhile, in the fall of 1984, Woods lew to Hong Kong and
sent back a stack of yearbooks containing the results of thou-
sands of races. Benter hired two women to key the results
into a database by hand so he could spend more time study-
ing regressions and developing code. It took nine months. In
September 1985 he lew to Hong Kong with three bulky IBM
computers in his checked luggage.
he Hong Kong that greeted Benter was a booming
inancial center, with some of the most densely
populated spaces on the planet. The crowded
skyline that had recently inspired Ridley Scott’s
dystopian megacity in
Blade Runner
seemed to
sprout towers weekly.
Benter and Woods rented a microscopic apartment in
a dilapidated high-rise. Warbling Cantonese music drifted
through stained walls, and the neighbors spent all night
shouting in the hall. Their oice was an old desk and a
wooden table piled high with racing newspapers. If they went
out at all, it was to the McDonald’s down the street.
Twice a week, on race days, Benter would sit at the com-
puter and Woods would study the racing form. Early on, the
betting program Benter had written spat out bizarre predic-
tions, and Woods, with his yearlong head start studying the
Hong Kong tracks, would correct them. They used a tele-
phone account at the Jockey Club to call in their bets and
watched the races on TV. When they won, there were satisied
smiles only. They were professionals; cheering and hooting
were for rubes.
Between races, Benter struggled to make his algorithms
stay ahead of a statistical phenomenon called gambler’s
ruin. It holds that if a player with limited funds keeps betting
against an opponent with unlimited funds (that is, a casino,
or the betting population of Hong Kong), he will eventually
go broke, even if the game is fair. All lucky streaks come to
an end, and losing runs are fatal.
One approach—familiar to Benter from his blackjack days—
was to adapt the work of a gunslinging Texas physicist named
John Kelly Jr., who’d studied the problem in the 1950s. Kelly
imagined a scenario in which a horse-racing gambler has an
edge: a “private wire” of fairly reliable tips. How should he
bet? Wager too little, and the advantage is squandered. Too
much, and ruin beckons. (Remember, the tips are good but
not perfect.) Kelly’s solution was to wager an amount in line
with the gambler’s conidence in the tip.
Benter was struck by the similarities between Kelly’s hypo-
thetical tip wire and his own prediction-generating software.
They amounted to the same thing: a private system of odds
that was slightly more accurate than the public odds. If the
public odds rate a given horse at 5 to 1, a gambler would have
to bet $1 ive times on average to win once. He spends $5 to
get back $5—his winnings and stake, minus the racetrack’s
$1 commission. Pointless. But if Benter calculated the
true
likelihood to be 4 to 1, he could bet a dollar four times to win
$5, a 25 percent proit. And he could diminish the impact of
bad luck by betting thousands and thousands of times. Kelly’s
equations, applied to the scale of betting made possible by
computer modeling, seemed to guarantee success.
If, that is, the model were accurate. By the end of Benter’s
irst season in Hong Kong, in the summer of 1986, he and
Woods had lost $120,000 of their $150,000 stake. Benter
lew back to Vegas to beg for investment, unsuccessfully,
and Woods went to South Korea to gamble. They met back
in Hong Kong in September. Woods had more money than
Benter and was willing to recapitalize their partnership—if it
was renegotiated.
“I want a larger share,” Woods said, in Benter’s
recollection.
“How much larger?” Benter asked.
“Ninety percent,” Woods said.
“That’s unacceptable,” Benter said.
Woods was used to being the senior partner in gambling
COURTESY BILL BENTER (3)
Benter (pictured here around 1982) got his start
playing blackjack professionally in Las Vegas.
Woods had the idea to bet on horses in
Hong Kong—but he and Benter quickly fell
out over money.
A sta Christmas party in 2000. Benter hired anyone—
coders, academics, journalists—who could improve
his algorithms.