The Eudaemons

A hunch in a Las Vegas bar, 1975

In the summer of 1975 Norman Packard, a physics student, spent two months counting cards at blackjack in Las Vegas with a friend, Jack Biles, and barely broke even. Over a last drink Biles said he would bet that roulette could be beaten with physics, and Packard remembered a cryptic line in Edward Thorp’s Beat the Dealer about a method for roulette. The two borrowed a tape recorder, stood beside wheels in several casinos and tapped the microphone each time the ballequipmentBallThe small ball launched along the track against the rotor. Modern balls are ivorine or Teflon, 18–21 mm.Read full → passed a fixed point. Plotted on graph paper, the taps showed the ball slowing and falling in a regular way. Packard’s school friend Doyne Farmer, a graduate student at the University of California, Santa Cruz, called it a waste of time until he did the sum himself: a person who can time the ball to about a tenth of a second should, on paper, be able to say where it will land to within a few numbers.

A wheel in the physics laboratory

In the spring of 1976 Farmer and Dan Browne paid $1,500 in cash for a reconditioned casino wheel at a workshop in Reno, saying it was for a fraternity party. For months they carried it at night to the university’s teaching laboratory and measured it with infrared sensors, a home-made electronic clock and a camera set off by the sound of the ball striking metal. What they found is still a good short list of what matters in a spin. The air does most of the slowing, more than the track. A wheel that is off level by as little as a tenth of an inch changes where the ball leaves the track, and on a tilted wheel the ball tends to come off at the high side, which makes the prediction easier. Balls differ: a Teflon ball slowed twice as fast as one of billiard-ball composite. And once the ball has left the track it still bounces, on average no more than a quarter to a third of the way round. That bounce was the question the whole plan hung on, and it was not wide enough to wipe out the prediction. By the end of the summer they put the possible advantage at 44%, the figure Thorp had given for his own device.

Roulette Wheel Physics

The company and the pie

They called the venture Eudaemonic Enterprises, from a Greek word for the happiness of a life lived by reason. Everyone who worked on it was to be paid from future winnings in proportion to the hours put in: a slice of what they called the Eudaemonic Pie. The winnings were meant to pay for a community of friends doing science on their own terms.

A computer under the arm, 1976–1978

The first machine grew from a kit computer that came by mail in September 1976 for $250, built on the 6502 processor that Apple’s computers also used. Farmer, who had never programmed before, wrote the equations for the ball and the rotorequipmentRotorThe spinning part of the wheel that carries the pockets.Read full → into it. The casino version was worn strapped to the body, under the arm or, for the women, in a bra. The player at the wheel clicked switches with the big toes: one toe marked the zerowheelZeroThe single green pocket that no even-money, dozen or column bet covers; the source of the house edge. Also: single zero, 0.Read full → and then the ball each time they passed a fixed point, the other moved the program between its settings. Fitting the program to a particular wheel took about twenty minutes of quiet clicking. Then, a few seconds into each spin, the computer named one of eight sectorswheelSectorA contiguous arc of pockets on the wheel, as distinct from a group on the layout.Read full → of the wheel, or signaled “no bet”, by buzzing small plungers against the skin. A second player farther down the table, linked by a weak radio signal, felt the same buzz and covered three or four numbers of that sector, sometimes swapping in a number next to it on the wheel so that the pattern was harder to read.

The first trip, to Lake Tahoe and Reno in December 1977, was one long equipment failure. A few weeks later, at the Golden Gate in downtown Las Vegas, Farmer played alone for four and a half hours at ten cents a chipmoneyChipA token representing a fixed amount, placed on the layout to make a bet. Also: check.Read full →. The profit was small, but over several hundred spins he and Browne put the edge, cautiously, at 25%.

Las Vegas Wheels

The summer of 1978

That summer the group played in shifts for weeks, downtown and on the Strip, two sessions a day and never the same casino on consecutive nights. They kept a notebook on each casino: its wheels and how far they were off level, its staff, its shift changes. Las Vegas wheels turned out to be obligingly tilted and the late-night games slow, which is what the computer needed, since a slow rotor and a fast ball leave more time to measure and to bet. There were good nights. At Circus Circus a change of croupierpeopleCroupierThe dealer at a roulette table.Read full → at two in the morning turned a losing session into several hundred dollars at 25 cents a chip. And there was a steady run of failures: loose wires, radio links that drifted with the temperature, electric shocks from the antenna wires sewn into their shirts. One night the plungers jammed, the wiring overheated, and Ingrid Hoermann played on through a burn sooner than leave a winning table. Casino staff noticed the winning and never the method; once a croupier, upset at a growing stack of chips, grabbed the turning wheel and wrenched it round.

In thirteen months they made eight trips to Nevada and, betting mostly dimes and quarters, won a few thousand dollars. A session at the California Club in December 1978 brought more than $1,000 in three hours and persuaded them that the method deserved better hardware.

The computer in a shoe, 1980–1981

To get the wires off their bodies they decided to put everything into shoes. Mark Truitt, a new physics graduate who had walked out of a job interview for military radar work, spent a year and a half on it. The idea that made it fit came from a university technician: switch the computer off between predictions, and it needs far fewer batteries. Truitt put the chips on two circuit boards he etched himself, turned one over on top of the other and filled the gap with hard wax, which made a block about two inches by four and half an inch thick that could be walked on. Batteries, antenna and two of the three buzzers went into the heel. Three buzzers at three speeds gave nine signals: eight sectors and “no bet”. A cobbler hollowed out ordinary leather shoes and asked no questions.

In November 1981 Farmer and the writer Thomas Bass took the shoes and $2,500 to Las Vegas. At the first table the first prediction hit, and then the buzzers began to fire at random. Casinos are full of electrical noise, from surveillance cameras, neon signs and slot machines, and it swamped the faint radio link between the shoes; a television set in the motel room did the same thing. Days of repairs by four people did not cure it, and the project stopped there. The equipment went into a trunk in a basement in Santa Cruz.

What 44% means

On an American wheel a straight upbetStraight upA single number, pays 35:1.Read full → wins once in 38 spins on average and pays 35 to 1, which costs the player 5.26%. A 44% advantage is what the same bet is worth when it wins once in 25 spins: 36 chips come back for every 25 staked. So the computer never had to know the number. It had to make the numbers it chose come up about one and a half times as often as chance allows.

American Roulette

Why a big edge still felt like gambling

An edge at 35 to 1 pays slowly and unevenly. One of the group’s advisers warned that a player staking $100 a number could be $10,000 or $20,000 down before the advantage showed, and the arithmetic agrees: with a 44% edge on single numbers a player is still behind after 100 spins nearly one time in four. That is why the Eudaemons started at dimes and built their bankrollmoneyBankrollThe total amount set aside for play. Bankroll management is deciding how much of it to risk per bet and per session.Read full → from winnings, and why Alan Lewis, who joined them in 1978, spent his time on how much to bet. Farmer later thought they had been too cautious. Small stakes kept them safe, and they also meant that years of work earned very little.

What it cost and what it paid

About $15,000 went into parts, tools and Truitt’s wages, most of it from Farmer and from Letty Belin, a lawyer who later married him. Farmer was credited with 3,500 hours of work, Truitt with 3,000 and Packard with 2,000. Against that stand the few thousand dollars won at small stakes. In 1983 Farmer advertised in a gambling magazine for investors, describing a prototype with a proven casino advantage of 20 to 40%. When the Eudaemons played, Belin’s search of Nevada law had found nothing that forbade predicting a game; in May 1985 the state made it a crime to use a device for that. What lasted was the science. Building the computers trained Farmer, Packard and their friends for the study of chaos, which is where they made their names.

Reading the book today

Thomas Bass, who placed the bets on the last trip, told the story in The Eudaemonic Pie (1985), built on long interviews with the group, and it is the source for this page. Its short history of roulette is less safe than its account of the project. It repeats the tale of Jagger’s six hired clerks, which goes back to a book of 1901 and no further (see Joseph Jagger), and it puts Thorp and Shannon’s casino test in 1962, where Thorp’s own accounts say the summer of 1961.

Questions