Where the Edge Comes From
Start with the hardware. A European single-zero wheel has 37 pockets numbered 0 through 36. A straight-up bet on one number pays 35 to 1. If the game were fair, it would pay 36 to 1, because there are 36 other possible outcomes for every number you choose. The casino pays 35, not 36: the difference is exactly 1/37 of every pound wagered on that bet. Expressed as a percentage: 1 divided by 37 equals 0.02703, or 2.703%. The decimal never terminates. In practice you see it rounded to 2.70%.
That fraction, 1/37, is embedded identically into every payout on the European table. A split covers two numbers at 17 to 1; fair odds would be 35/2 = 17.5. The shortfall is 0.5/37, which reduces to 1/37 per unit staked. A corner covers four numbers at 8 to 1; fair odds would be 33/4 = 8.25. Shortfall per unit: 1/37. Even-money bets cover 18 numbers at 1 to 1; fair odds would be 19/18 = 1.056. The shortfall comes through zero: when zero lands, the casino takes the even-money bet. Zero lands 1 time in 37, costing you 1/37 of every even-money wager. All roads lead to the same fraction. This is the algebraic identity the spoken lesson references: it's not a coincidence built into the payouts, it's a structural consequence of having one more pocket than the payout schedule acknowledges.
American Double-Zero and the Basket Exception
An American wheel adds a 00 pocket to the 0 through 36 arrangement, giving 38 pockets total. A straight-up bet still pays 35 to 1. Fair payout on 1 in 38 odds would be 37 to 1. The shortfall is 2/38 = 5.263%. Every standard American bet carries this edge, almost twice the European figure for the same physical game with the same dealer and the same playing experience. The sole reason for the difference is the extra pocket.
The five-number basket bet on the American wheel covers 0, 00, 1, 2, and 3. It pays 6 to 1. The fair payout for covering 5 of 38 pockets would be 33/5 = 6.6 to 1. The casino rounds to 6, not 6.6, because 33 isn't divisible by 5. That rounding creates an edge of 7.895%, per Wizard of Odds. It's the only standard roulette bet with a house edge higher than the wheel's own headline figure. Every other American bet is 5.26%. The basket is 7.89%. There's no reason to place it.
Empire Casino at Leicester Square runs American double-zero roulette exclusively, per the venue's published table information. A player sitting at the Empire pays 5.26% on every bet regardless of whether they've chosen a straight-up, a dozen, or red. The postcode is Zone 1 Central London. The mathematics is Nevada.
La Partage and the Annual Cost of Getting It Wrong
La Partage is a French rule that changes what happens when zero lands on even-money bets. Instead of losing your entire stake, you recover half of it. The mathematics is straightforward: zero lands 1 time in 37 on average. Without La Partage, you lose 1 unit when zero lands, so the expected loss from zero on an even-money bet is 1/37 of your stake. With La Partage, you lose only half a unit when zero lands, so the expected loss from zero is 0.5/37 = 1.351%. Exactly half the standard European edge.
The practical annual cost of this difference is not abstract. At £100 per spin, 50 spins per hour, playing even-money bets: £5,000 wagered per hour. At 2.70%, expected cost per hour is £135. At 1.35%, expected cost per hour is £67.50. The difference is £67.50 per hour. Over 100 hours of annual play, that's £6,750 per year, simply from sitting at a European table rather than a French one for your even-money bets. The venues are different, the table minimum and maximum may differ, but the calculation is the same: £100 x 50 x 100 x (2.70% - 1.35%) = £6,750.
Aspers Westfield Stratford confirms La Partage on its European tables, making it the best standard edge available at a UK land-based casino floor on even-money bets. The Hippodrome in Leicester Square and Les Ambassadeurs in Hamilton Place run European tables; players should confirm La Partage availability directly with each venue, as it isn't universally offered even on single-zero tables. Evolution's French Roulette Gold and Auto Roulette La Partage apply the rule online, as confirmed by Evolution's product catalogue.
Why the House Edge Is a Per-Spin Tax, Not a Per-Session Budget
A common misreading of the house edge treats it as a percentage of starting bankroll. It isn't. It's a percentage of total amount wagered. If you bring £1,000 and make 100 bets of £10, you've wagered £1,000 total. Expected loss: £27. If you bring £1,000 and make 100 bets of £100 on an up-and-pull system where you sometimes win and reinvest, you may have wagered £7,000 total before the session ends. Expected loss: £189, even though you started with only £1,000. The edge accumulates on action, not on bankroll. Progression systems that inflate bet sizes after wins increase expected loss in direct proportion to the additional action they generate.
The Zero-Free Interlude and What It Proves
There's a brief historical moment that illustrates the edge calculation with unusual clarity. For roughly a year in late 1967 and into 1968, several London casino rooms removed the zero from their roulette wheels following Scotland Yard's warning of 30 December 1967 that a zero-bearing wheel might be prosecuted under the Betting and Gaming Act 1960. A wheel with 36 pockets and no zero has no house edge on even-money bets: there's no 37th pocket for the casino to withhold. The even-money bet becomes a genuine 50/50 proposition. The Gaming Act 1968 resolved the situation by explicitly authorising the modest house edge as a licensing term, restoring the zero and with it the 2.703% fraction that governs every London table today. The historical episode confirms the edge's source precisely: remove the zero and the even-money edge disappears. The zero is the entire mechanism.
Key numbers
| Wheel / Rule | Pockets | House edge | Annual cost (£100/spin, 50 spins/hr, 100 hrs, even-money) |
|---|---|---|---|
| European single-zero | 37 | 2.703% | £13,515 |
| European + La Partage | 37 | 1.351% | £6,755 |
| American double-zero | 38 | 5.263% | £26,315 |
| American basket bet | 38 | 7.895% | N/A (inside bet; different variance) |
| Zero-free (London 1967-68) | 36 | 0% | £0 (historical only) |
