The Eudaemons
A Santa Cruz group turned roulette physics into a wearable-computer experiment.

The Eudaemons built wearable computers to predict roulette outcomes. Participant accounts and a museum description document the method and its hardware problems. Reported advantages are historical estimates, not guaranteed returns or an audited profit.
From physics to a casino experiment
Doyne Farmer recalls forming Eudaemonic Enterprises with Norman Packard and friends while in graduate school. They hoped roulette winnings would fund a science commune. Their idea depended on measuring the physical game, rather than changing stakes after wins or losses.
A wheel in the physics laboratory
Farmer describes buying a roulette wheel and studying its physics experimentally and mathematically. That work provided a setting for testing predictions before casino play. A result from one test wheel does not establish the same accuracy on another: the equipment, conditions and measurement errors all matter.

A shared scientific project
Farmer connects the group’s name with eudaemonia, a life guided by reason. The project brought together physics, programming and electronics. Its scientific ambition helps explain why a successful calculation and a successful gambling venture were separate goals.
A computer the player could wear
Farmer says the first equipment was worn under the arm and a later version fitted inside a shoe. Toe switches supplied timing information, and vibrating signals communicated the prediction. One person measured the ballequipmentBallThe small ball that settles in a numbered pocket to determine a physical roulette result.Read full definition → and rotorequipmentRotorThe rotating part of a physical roulette wheel, carrying the numbered pockets.Read full definition →; another placed bets. The method needed enough time between measuring the spin and the close of betting.

Testing the complete system
A useful prediction still has to reach the bettor in time. The computer, switches, power supply and communication link all form part of that system. A failure at any stage can make an otherwise useful calculation impossible to act on. This is why a model’s estimated advantage should not be treated as the return of a working casino device.
Wheel measurements and a wearable computer, described by Farmer.
The museum describes radio interference disrupting the Sundance trial.
Thomas Bass’s The Eudaemonic Pie is published.
The shoe computer at the Sundance, 1981
The Heinz Nixdorf MuseumsForum, which displays the shoe computer, describes a November 1981 trial at the Sundance in Las Vegas. Farmer used a toe switch to time the ball. A transmitter sent the predicted wheel sectorwheelSectorA continuous arc of neighboring pockets around the wheel, rather than a group defined by the betting layout.Read full definition → to Bass, who received vibrating signals in his shoe and placed the bets.
The museum reports that casino electronics disrupted the radio link and made the signals unreadable. The team decided against another lengthy rebuild. Its account supports a hardware problem during that trial, not a general claim that electronic prediction always succeeds or always fails.

What 44% means
On a uniform American wheel, a fixed straight-up bet wins with probabilitymathProbabilityA measure of how likely an event is, from zero to one. Percentages express the same scale from 0% to 100%.Read full definition → 1/38. At a 35:1 net payoutmoneyPayoutThe amount paid on a winning wager. Check whether the quoted figure is net profit or total return including the stake.Read full definition →, its expected return is 36/38 − 1, or −5.26%. If a prediction method truly raises the hit probability to 1/25 on the bets selected, the same calculation gives +44%.
This is conditional arithmetic, not a measurement of the Eudaemons’ device. A real estimate would need the bets placed, the outcomes and uncertainty. Predicting a sector also differs from making a single-number bet; the return must be calculated across the full stake.
Why a big edge still felt like gambling
A positive expected return does not make a short session predictable. In an illustrative model with independent spinsmathIndependent spinsSpins for which learning an earlier outcome does not change the probabilities assigned to a later one.Read full definition →, one unitmoneyUnitA base amount used to express stakes or results.Read full definition → on a single number each spin, a true hit probability of 1/25 and a 35:1 payout, the player is still behind after 100 spins about 23.2% of the time. That is the probability of zerowheelZeroThe 0 outcome, conventionally shown in green. Ordinary color, parity, high/low, dozen and column bets do not include it.Read full definition →, one or two wins.
This example is not a reconstruction of the group’s sector bets. Smaller stakes reduce the money exposed, but they do not eliminate losses or establish that a particular bankrollmoneyBankrollThe amount set aside for play. Bankroll management concerns stakes and limits within that amount.Read full definition → is safe.
Estimates, winnings and profit
Bass’s preview gives a 44% advantage estimate; Farmer’s retrospective account reports 20%. These figures should stay attached to their sources. Without matching test conditions and betting records, they are not interchangeable measurements.
A percentage advantage is also different from money earned. Net profit would need to account for stakes, payouts, equipment and other costs. The sources used here do not establish a complete audited profit-and-loss account.
Reading the book today
Thomas Bass describes the experiment in The Eudaemonic Pie (1985). His prologue explains the wearable equipment from a participant’s perspective. Farmer’s account and the museum’s description offer useful points of comparison. Keep the distinction between an estimated advantage, a reported test and a complete record of casino results.
Questions
Farmer says the group did not realize the large sums it had hoped for. The accounts used here do not establish an audited net profit. A promising estimate is not the same as earnings after costs.
As checked on 15 November 2026, NRS 465.075 prohibits specified use, assistance and possession with intent to use advantage devices, software or hardware. It includes exceptions for approved games or Commission permission. That is a Nevada rule, not a worldwide legal summary.
The account cannot establish that. Prediction accuracy, equipment, the betting cutoff and legal restrictions all matter. A video stream or a familiar-looking wheel does not establish an exploitable opportunity.
Fun with Math
Try these yourself first. Each solution is folded; open it when you are ready.
Suppose a prediction gives a straight-up bet paying 35:1 an expected return of +44% of its stake. What hit probabilitymathProbabilityA measure of how likely an event is, from zero to one. Percentages express the same scale from 0% to 100%.Read full definition → does that require? On a 38-pocketequipmentPocketA compartment on the rotor in which the ball can settle. The number of pockets depends on the wheel.Read full definition → wheel, how many pockets would need to be excluded in a model that assigns equal probability to every remaining pocket?
Show the solution
Answer Once in 25 spins. That is the same as ruling out 13 pockets and leaving 25.
A winning chipmoneyChipA token used to place a wager, with a printed or assigned value. Also called a check in some casino terminology.Read full definition → comes back with 35 more: 36 chips. If the number wins once in n spins, n chips staked bring 36 back, so the bet is worth 36/n − 1.
Set that to 0.44: 36/n = 1.44, so n = 25. With no prediction n is 38 and the bet is worth 36/38 − 1 = −5.26%.
Now suppose a prediction rules out k pockets and the ballequipmentBallThe small ball that settles in a numbered pocket to determine a physical roulette result.Read full definition → is equally likely to land in each of the other 38 − k. A chip on one of those wins once in 38 − k spins. For 44% we need 38 − k = 25, so k = 13: a little more than a third of the wheel.
This equivalence depends on equal probabilities across the remaining pockets. Excluding eight of 38 pockets gives 36/30 − 1 = +20% in this model. Different probability distributions require their own calculation; this exercise does not measure a historical team’s performance.
pockets ruled outpockets lefta straight up is worth038−5.26%2360%830+20%1325+44%1919+89%Suppose your number really does win once in 25 spins at 35 to 1, a 44% edge. You bet one chip a spin. How likely are you to be behind after 100 spins? After 500?
Show the solution
Answer About 23% after 100 spins and still about 6% after 500.
After 100 spins you have staked 100 chips, and each win brings back 36. Two wins bring 72, a loss of 28; three bring 108, a gain of 8. So you are behind if you win twice or less.
The chance of no win in 100 spins is (24/25)¹⁰⁰ = 1.7%. Exactly one win: 100 × (1/25) × (24/25)⁹⁹ = 7.0%. Exactly two: 4,950 × (1/25)² × (24/25)⁹⁸ = 14.5%. Together 23.2%.
After 500 spins you need 14 wins to be ahead (14 × 36 = 504), and the average is 20. The chance of 13 or fewer is about 6.2%.
A positive expected return can coexist with a substantial chance of finishing behind. These binomial results assume independent spinsmathIndependent spinsSpins for which learning an earlier outcome does not change the probabilities assigned to a later one.Read full definition →, a stable hit probability of 1/25, and enough funds to place every planned bet. They do not establish that a real prediction has that accuracy.
spinswins needed to be aheadchance of being behind50240.0%100323.2%200618.6%500146.2%Suppose every bet has a known +44% expected return and you make 50 bets an hour. What is the expected hourly result at $100 per bet, at $0.10 per bet, and at four $0.10 bets per spin if each retains that same edge?
Show the solution
Answer +$2,200 at $100 per bet; +$2.20 at $0.10 per bet; +$8.80 with four $0.10 bets per spin. These are expectations, not hourly guarantees.
Expected profit is the amount staked multiplied by the player edge. At $100 per spin and 50 spins an hour, turnovermoneyTurnoverThe sum of amounts wagered over a session or period, including money wagered again after a return. Also called handle or action.Read full definition → is $5,000 and expected profit is $5,000 × 0.44 = $2,200.
At $0.10 per spin, hourly turnover is $5 and the expectation is $2.20. Four $0.10 bets per spin make turnover $20 and expectation $8.80, provided each bet really retains the assumed 44% edge.
The arithmetic scales with stakes. It does not establish a real-world edge or describe the stakes, earnings or motives of a historical group.
Comic relief

Reported laboratory promise met hardware problems in casino tests.
Sources
- Bass, T. A. The Eudaemonic Pie.
- Thorp, E. O. “Optimal gambling systems for favorable games.”
- Thorp, E. O. “The invention of the first wearable computer.”
- Nevada Revised Statutes, chapter 465.
- Nevada NRS 465.075 — current text checked 15 November 2026
- Heinz Nixdorf MuseumsForum, Shoe computer — the Sundance trial
- J. Doyne Farmer, Roulette — participant’s account
- Thomas Bass, The Eudaemonic Pie — public preview