[00:02] Vegas, February 1979. A 26-year-old man in a brown corduroy jacket is standing at a roulette table watching the croupier spin the wheel. He is not betting yet. He is not drinking. He is not talking. By every measure any
[00:19] security camera in the building can identify, [music] he is just standing there. The croupier releases the ball. For approximately 3 seconds, the man does nothing. Then, he steps forward and places three chips on a specific section
[00:33] of the table. The croupier calls no more bets. The ball drops. It lands in the section he bet on. It will land there on this wheel and on every other wheel he plays on for the next 18 months at a frequency the laws of probability say
[00:49] should be physically impossible. This is Doyne Farmer, a physics PhD student at the University of California, Santa Cruz. He has never been arrested by any casino in Las Vegas. He has never been caught by any security team. He has
[01:05] caught by any security team. He has never broken a single Nevada state law. Across 3 years of methodical operation across Las Vegas and Reno, he has won tens of thousands of 1979 dollars from casino roulette wheels. All of it is
[01:20] legal and all of it depends [music] on a piece of technology that in 1979 nobody on Earth knows is possible to build. So, how does a 26-year-old PhD student build
[01:32] a device that lets him beat casino roulette wheels across Las Vegas without breaking a single American law? And what was inside that device that the Nevada legislature would, 6 years later, be forced to write an entirely new state
[01:47] be forced to write an entirely new state law to outlaw? do at the Golden Nugget, you have to understand what the roulette wheel was The American wheel has 38 numbered pockets. The croupier spins a small
[02:03] white ball against the rotation of the wheel. 15 seconds later, the ball loses momentum and drops into a pocket. A single number bet pays 35 to 1. The actual probability is 1 in 38. The difference is the casino's edge, 5.26%.
[02:21] It is mathematical. It is permanent. It is built into the geometry of the wheel is built into the geometry of the wheel itself. For 300 years, every casino in the world treated this edge as inviolable. You could not count cards.
[02:34] Roulette had none. You could not track patterns. Every spin was statistically independent. You could not mark the ball. It was changed every shift. Bribing the croupier was pointless. The croupier did not know which pocket would
[02:49] catch the ball [music] any more than the player did. The only way to win at roulette was to be lucky. The casinos counted on people not being lucky. What Farmer realized in spring 1977, sitting in a physics seminar at UC Santa
[03:04] Cruz, was that this was not, strictly speaking, true. Roulette is not a random speaking, true. Roulette is not a random game. Roulette is a Newtonian game. The ball has a measurable initial velocity. The wheel has a measurable initial
[03:17] velocity in the opposite direction. The friction, given a particular wheel and >> [music] >> If you knew the initial conditions, you could, with enough computing power, predict where the ball would land. You
[03:31] did not need [music] the exact pocket. You needed only which eighth of the wheel. Even imperfectly, that would give you a 15% [music] edge over the casino. Doyne Farmer was born in 1952 in Silver
[03:45] City, New Mexico. His stepfather was an astronomer named Tom Ingerson, who ran the local Boy Scout troop and openly dreamed of one day building a colony on He took Farmer and the other boys camping in the New Mexico desert, taught
[04:00] under the stars, and told them repeatedly that anything they could imagine, they could build. Farmer believed him. By his late teens, he had been accepted to Stanford. In 1973, [music] he
[04:14] transferred to UC Santa Cruz to pursue a PhD under a small group nobody at any paying attention to. The group was studying what would, 15
[04:26] theory. In 1973, it had no name, six members. The senior researcher was Robert Shaw, a soft-spoken theoretical physicist who
[04:38] had refused several major university appointments to stay at Santa Cruz. The graduate students were Farmer, his lifelong friend Norman Packard, James Crutchfield, and three others. They met informally in Shaw's office, in graduate
[04:53] student apartments, in the back of the physics building at hours when it was officially closed. They called themselves the chaos cabal. Their proposition, most systems in nature that appear random, weather, fluid
[05:07] turbulence, pendulums, a small ball on a spinning wheel, are not random. They are deterministic. They look random only because the initial conditions are too complex, [music] and the sensitivity to them too extreme for any human to
[05:21] predict without computational help. In 1973, that meant a university mainframe. 1973, that meant a university mainframe. By 1977, it meant something else. In June 1975, an engineer at a small semiconductor
[05:36] company [music] called MOS Technology released a microprocessor he had designed in his garage in Pennsylvania, the 6502. It cost $25. It fit in the palm of a hand.
[05:50] It fit in the palm of a hand. By autumn 1977, the same chip sat inside two new consumer products called the Commodore PET and the Apple II. It also sat on the workbench of a 24-year-old physics PhD student in Santa Cruz who
[06:04] had just realized that predicting where a roulette ball would land was no longer a theoretical problem. It was an engineering problem and he was an engineer. In April 1977, Farmer walked into the
[06:18] office his friend Norman Packard shared on the third floor of the Santa Cruz physics building. He closed the door. He explained what he wanted to do. Packard listened for approximately 15 minutes without interrupting. Then he said one
[06:32] sentence, "Let's build it." By the summer of 1977, [music] 11 people had joined the chaos cabal in what would become one of the strangest engineering projects in the history of American physics. They called
[06:45] themselves Eudaemons from Aristotle's Nicomachean Ethics, eudaimonia, the flourishing that comes from the exercise of human reason against the indifference of the universe. The first thing the Eudaemons needed was
[07:00] The first thing the Eudaemons needed was a roulette wheel. In autumn 1977, Farmer and Packard drove from Santa Cruz to a casino supply warehouse in Reno. They paid approximately [music] $1,400 in cash, most of two graduate students'
[07:14] stipends, for a regulation American roulette wheel. They drove it back to a rented house at 707 Riverside Street in Santa Cruz where together, pooling graduate [music] stipends into a common fund, sharing
[07:29] rent, and eating dinner around one long wooden table. [music] And installed the wheel in the basement. For the next 18 months, that basement was open 24 hours a day to any cabal member who wanted to spin the wheel and time the ball. The
[07:44] wheel was spun approximately 20,000 times. The device had three components. The input, two micro switches hidden in a shoe. The wearer pressed one with the big toe when the croupier released the
[07:58] ball. The second when the ball passed a fixed reference mark. The third when the wheel completed one rotation. Three button presses, the fundamental physics button presses, the fundamental physics of the system. The computer, an MOS 6502
[08:13] microprocessor on a circuit board 5 cm square. Two memory chips, a timing crystal, analog circuitry, four D cell batteries. The whole assembly was the size of a packet of cigarettes. It hid in the hollow heel of a custom platform
[08:29] in the hollow heel of a custom platform shoe. A fashion item that in 1978 drew no second glance on a Vegas floor. The output, eight electromagnetic solenoids wired to a leather harness
[08:41] worn under the shooter's shirt around the waist. When the computer calculated which eighth of the wheel would catch the ball, it sent a coded vibration to the corresponding solenoid. Three short pulses on the third solenoid meant bet
[08:56] the third octant. There were two roles. The kicker stood at the wheel timing the ball, not betting, not speaking. The shooter stood at the betting table 3 m away receiving the vibration signal placing the chips. The two did not
[09:11] acknowledge each other inside the casino under any circumstance. If something supposed to be caught. By spring 1978, [music] the system worked. On a real regulation wheel, the computer predicted the
[09:26] correct octant 44% of the time. The casino edge on a single number was 5.26% [music] against the player. The Eudaemon edge on an octant [music] bet was 44% for the player. By any measurement of any casino
[09:41] game ever played, [music] it was the largest mathematical advantage anyone had ever found in a regulated gambling operation.
[09:54] Farmer walked onto the casino floor of the Stardust in a brown corduroy jacket, bell-bottom jeans, and custom platform shoes that weighed 3 lb each. He a
[10:06] roulette table in the center of the floor and stood at the rail in the kicker position. The croupier was a man in his 50s. He did not look up. He spun the wheel counterclockwise. He released the ball clockwise.
[10:20] Farmer pressed the switch in his right shoe at release. The second when the ball passed a metal post. The third when the wheel completed one rotation. One and a half seconds later, three
[10:32] short vibrations triggered against the left side of his abdomen, the third octant. Farmer did not bet. He was there to verify the system worked. The ball dropped. It bounced twice. It came to rest in pocket 23.
[10:48] Pocket 23 was in the third octant. Farmer stood at the rail for 2 hours and recorded 63 additional spins. The computer predicted the correct octant on 38% of them, meaningfully above the statistical baseline of 12 and
[11:03] a half percent. The team had calibrated for 44. In the heat, vibration, and electromagnetic interference of an active casino floor, the device was running at 38. 38 was enough.
[11:16] Farmer flew back to Santa Cruz the next morning, handed the data to Packard, and spent 4 weeks recalibrating. By April 1978, the 11-person [music] team was organized into four rotating two-person crews.
[11:30] Farmer worked the Strip, Caesars, the Sahara, the Stardust, the Marina, the Golden Nugget. Packard worked Reno. The protocol, no crew entered the same casino more than once a week. No member won more than $700 per visit. No crew
[11:48] spoke to any other crew member inside any casino. By autumn 1978, the any casino. By autumn 1978, the Eudaemons had won approximately $12,000. By spring 1979, 25. By summer 1979, 40.
[12:04] The money paid rent on the Riverside house. It paid for rental vans to drive to Vegas and Reno, for cheap motel rooms two blocks off the strip, for soldering irons and oscilloscopes in the basement, for the long [music] communal dinners,
[12:19] pasta, salad, cheap California wine, where the team debriefed each run at the long wooden table. Nobody bought a car. Nobody bought a house. Nobody quit Nobody bought a house. Nobody quit graduate school.
[12:36] 1979, every Eudaemon who had worn the harness in an active casino reported the same thing. After 90 minutes under casino lighting, humidity, and the thermal load of nervous adrenaline, the wires from the
[12:50] solenoid harness to the shoe computer began to slip. The insulation chafed. The copper came into intermittent contact with sweat. Sweat conducts electricity. The current was small, 4 and 1/2 volts, a fraction of an amp, but
[13:05] applied to bare skin through a 1 mm contact point for 90 minutes, it is enough to produce a burn. By December 1979, three members of the team had been northern [music] California for minor electrical burns sustained during casino
[13:22] operations. The fourth incident in early 1980 was not minor. The team member, whose identity was kept out of every published account, was wearing the harness on a casino floor in Reno when a movement aligned the chafing point with
[13:36] a sweat bead the wearer had not noticed. The current arced. The wearer froze at the betting table. The shooter at the next table noticed, walked calmly to the bar, then to the lobby, then signaled to the kicker by prearranged code. The
[13:51] operation was over. They met outside the casino 7 minutes later. Farmer drove the team member to a small clinic in Reno. The physician was told hair dryer. The team member required
[14:05] approximately 3 weeks to fully heal. Farmer drove through the night back to Santa Cruz. Leticia his girlfriend, an attorney who had joined the project as a shooter, was waiting at the Riverside house when he arrived at 4:00 in the
[14:18] morning. They sat at the long wooden table in the kitchen, neither of them turning on a light. >> It's not worth it anymore. >> No, it isn't. >> The Eudemons met in the basement on
[14:31] Riverside on February 12th, 1980. The conclusion was unanimous. The system worked. The system would continue to work. But the system was putting the lives of the people who operated it at meaningful physical risk. To continue
[14:46] would require substantial redesign. The team members were making real progress on the dissertations that had brought them to Santa Cruz. Robert Shaw's chaos attention from physics departments across the country. The Eudemons voted.
[15:01] The project was over. The harnesses were destroyed. The five versions of the device were boxed and stored. The roulette wheel was sold back to the warehouse in Reno. They had won by their own final accounting between 40 and
[15:15] own final accounting between 40 and 60,000 1980 dollars across 26 months of casino operations. They had not been caught, identified by any state gaming commission, or even seriously suspected. The casinos would not learn anything had
[15:29] The casinos would not learn anything had happened until 1985. journalist named Thomas Bass heard a rumor that physics graduate students
[15:41] that UC Santa Cruz had done something unusual with a microprocessor and a pair unusual with a microprocessor and a pair of platform shoes in a Las Vegas casino. specifics. He found Farmer first. By 1981, Farmer
[15:56] had finished his PhD and accepted a research position at Los Alamos National Laboratory. Farmer agreed to speak for one reason. A 1980 internal report from the Nevada Gaming Control Board leaked [music] in late 1981 had used the phrase
[16:13] "the most coordinated assault on roulette in Nevada history." The casinos were starting to suspect. Farmer worried other groups would try to copy what the U Demons had done. And without the discipline, safety protocols, and
[16:27] voluntary exit ethos of the original team, would get themselves arrested, team, would get themselves arrested, injured, or killed. He shared 90% the injured, or killed. He shared 90% the remaining 10, source code, wire routing,
[16:39] calibration formulas, stayed locked in the boxes with the shoes. The Eudemonic Pie was published by Houghton Mifflin in October 1985. The New York Times, Wall Street Journal, and Time reviewed it. It sold 200,000
[16:54] copies in eight languages. The casino industry response was immediate. Within 48 hours, the Nevada Resort Association, Vegas operator, wrote to the Nevada
[17:07] revision of the state's gambling cheating statutes. The legislature, 11 weeks later, complied. Nevada Senate Bill 467 passed on June 4th, 1985 added
[17:21] one paragraph to the Nevada Revised Statutes. Under title 463, it criminalized the possession or use of any device which assist in projecting the outcome of any game on the premises of any licensed gaming establishment.
[17:36] The penalty, a felony. Maximum sentence, The penalty, a felony. Maximum sentence, 10 years. The law has, in 2026, never been amended. It has been adopted in similar form by every US state that has
[17:49] since licensed casino gambling and by Macau, Singapore, every Australian state, the UK, and the European online gambling regulators. By 2026, what the U
[18:01] Demons did in 1978 is illegal in every commercial casino is illegal in every commercial casino jurisdiction on Earth. It was, in 1978, jurisdiction on Earth. It was, in 1978, illegal in none of them.
[18:16] published, the U Demons had moved on. At Los Alamos, Farmer had become one of the interdisciplinary field called complexity science. Packard had returned to academic physics at the University of Illinois.
[18:30] They had no further interest in casinos. They had a very specific interest in one other place. A place that ran on the same mathematics. A place whose participants, by 1985, were beginning to use exactly the kind of computational
[18:44] prediction the U Demons had pioneered. A place called Wall Street. In March 1991, Farmer and Packard, with a Stanford-trained physicist named James
[18:56] McGill, incorporated a Delaware corporation called Prediction Company. Headquarters, an Adobe office at 236 Montezuma Avenue in Santa Fe. Five blocks from the Santa Fe Institute. Three founders, no clients, no
[19:12] marketing. What they had was a hypothesis. The same mathematics that had led two physics graduate [music] students predict a roulette ball in 1978 would, they believed, predict short-term price [music] movements of liquid
[19:25] financial securities. The standard academic position in 1991 was the efficient market hypothesis. Markets were random, prediction impossible by Farmer had [music] spent 14 years proving that things a economist called
[19:41] random were, in fact, [music] deterministic. He had a roulette wheel in a basement to show for it. The first client was a Chicago options firm called O'Connor and Associates. In November 1991, O'Connor agreed to pay a flat fee
[19:56] plus a share of trading profits for exclusive access to Prediction Company's models. By the end of the first year, the models were returning [music] roughly three times the average of comparable Chicago options operations.
[20:09] In 1995, O'Connor was acquired by what would, after two mergers, become UBSAG. By 2005, Prediction Company was a wholly-owned [music]
[20:21] wholly-owned [music] UBS subsidiary. Between 1991 and [music] its quiet dissolution in 2013, the company generated, by Financial Times and Bloomberg estimates, between 100 and 200 million [music] dollars in
[20:35] cumulative profit. Farmer was an early seven-figure equity holder. The proceeds another paying corporate position for the rest of his career. He did not. He bought no yacht. He bought no private plane. He funded
[20:51] research positions at the Santa Fe Institute. He returned, repeatedly across the next four decades, to the New Mexico desert where Tom Ingerson had once taught him to navigate by the stars. He married Letty.
[21:05] Farmer left Prediction [music] Company in 1999. He spent 8 years as a professor at the Santa Fe Institute. In 2007, he took a Santa Fe Institute. In 2007, he took a tenured chair at Oxford, where by 2026,
[21:18] he was director of the Complexity Economics program and a senior advisor to the Bank of England, the ECB, and the British Treasury. His official Oxford biography does not mention roulette. The shoes remain in a small wooden box on
[21:33] the upper shelf of a closet in his Oxford home. Oxford home. In 2026, there are approximately 3,500 roulette wheels in commercial service on the casino floors of Las Vegas.
[21:46] They are the same wheels manufactured by the same suppliers that Doyne Farmer was the same suppliers that Doyne Farmer was timing in February 1979. The casino edge is the same. The mathematics is the same.
[22:00] The ball on every spin still obeys [music] Newton's laws. [music] Newton's laws. What has changed in 2026 is the law. Every casino floor in Las Vegas now displays a small printed notice required
[22:13] displays a small printed notice required since 1985 by the Nevada Gaming Control Board informing patrons that possession of any device which assists in felony. That paragraph was written specifically
[22:28] because of what Farmer did in 1978. He did not break Las Vegas. He did not break American gambling law. He did not break a single rule that existed at the time he was playing. He just timed the ball in a specific
[22:43] just timed the ball in a specific sequence until Newton's laws paid.