Wheel rules · exact expectations

European vs American vs French roulette: what changes?

One extra zero doubles the usual house edge. One zero rule halves it for even-money bets. The payouts printed on the layout barely change; the result depends on which pockets and rules sit behind them.

2.70%
European edge
single zero, standard rules
5.26%
American edge
double zero, standard bets
1.35%
French la partage
even-money bets only
7.89%
American top line
five-number bet, 0/00/1/2/3

Compare the wheels and zero rules

All three use numbers 1–36 and the same conventional payouts: a single number pays 35:1, a dozen 2:1, and red or black 1:1. French roulette uses a single-zero wheel; its lower even-money edge comes from the zero rule, not a different set of winning numbers.

Scroll the comparison table sideways to see every column.

Roulette variant comparison: pockets, zeros, red probability, and house edge
Rule setPocketsZero pocketsP(red)Regular bet edgeEven-money edge
European37048.65%2.70%2.70%
American380, 0047.37%5.26%5.26%
French, with la partage37048.65%2.70%1.35%

“Regular” means a standard payout without a zero concession. American top line is the exception: five pockets pay only 6:1, giving it a 7.89% edge. See the five-number calculation.

01 / 37 pockets

European

Full stake lost on zero.

1 ÷ 37 = 2.70% edge on red
02 / 38 pockets

American

Full stake lost on 0 or 00.

2 ÷ 38 = 5.26% edge on red
03 / 37 pockets

French, with la partage

Half of an even-money stake returned on zero.

½ ÷ 37 = 1.35% edge on red

Why a second zero matters

A $1 bet on red wins $1 when any of 18 red pockets lands. European roulette has 18 black pockets and one green zero, so the net expectation is (18 − 19) / 37 = −$0.0270. American roulette adds 00 without changing the payout: (18 − 20) / 38 = −$0.0526. The same stake loses nearly twice as much in expectation.

The extra zero changes the odds for every standard bet. It does not make streaks more predictable. After ten reds, the next spin is still red with probability 18/37 on a European wheel or 18/38 on an American one. See the exact streak probabilities.

What la partage actually changes

Under la partage, a zero returns half of an even-money stake. A $10 red bet therefore wins $10 on red, loses $10 on black, and loses $5 on zero. Its expectation is (18 × $10 − 18 × $10 − $5) / 37 = −$0.1351 per spin: 1/74 of the $10 stake, or 1.35%. A dozen or straight-up bet still loses its full stake on zero and retains the 2.70% edge.

En prison is a different zero procedure: the even-money stake stays on the table for a later spin and can be returned if its side wins. What happens on repeated zeros or when play ends needs an explicit table rule. French regulations describe a choice between taking half back and imprisoning the stake; do not assume every table applies both in the same way.

Expected loss on red or black

Flat stakes on independent spins. These are averages over many sessions, not a forecast for one run.

European−$27.032.70% of $1,000 wagered
American−$52.635.26% of $1,000 wagered
French, with la partage−$13.511.35% of $1,000 wagered

What this simulator covers

0xRoulette models the European single-zero and American double-zero wheels with standard payouts. It does not implement la partage or en prison settlement. The French column and calculator above are exact expectations for the stated la partage rule, not a runnable third wheel. Use the simulator to compare standard European and American runs with the same seed, or inspect the paired wheel-risk study.

Sources and method

Pocket counts and standard payouts follow the Nevada Gaming Control Board roulette rules of play. The French zero choice follows Article 52 of France’s casino rules. European and American standard payouts are checked against the site’s payout rulebook; their probabilities and expected losses use those payouts and pocket counts. French la partage expectations are derived separately from the half-stake return on zero. Results assume a fair wheel and independent spins; physical wheel bias is a separate question.