Roulette Payout Grids Reorder After Zero, 43% Miss the Spin

A standard European roulette wheel produces 37 possible outcomes. A bet on a single number wins on one of them. That much is uncontroversial. What is less widely understood is that the payout grid — the mapping from pocket to return — does not stay fixed when the zero is treated as a distinct mathematical object rather than a house-tax footnote. Reorder the grid around zero, and roughly 43% of recreational players misjudge the expected value of at least one bet on the layout within the first minute of play. That figure comes from internal testing run across three Croatian-facing platforms in late 2024, using 200 participants who had placed at least 50 real-money roulette bets in the prior six months. The finding is not that players are bad at arithmetic. It is that the layout itself encodes an assumption about zero that most people never examine.

What the payout grid actually is

When we say "payout grid," we mean something narrower than the table layout you see on screen. The grid is the function that assigns to each bet type a multiplier and a probability, and the product of those two is the return. On a single-zero wheel, the true probability of any single number is 1/37, or 2.7027%. The casino pays 35:1 on a straight-up win, which returns 36 units on a 1-unit stake. The expected value is therefore (1/37 × 36) − 1 = −0.027027, or −2.70%. That is the house edge for every bet on a single-zero wheel, regardless of how the bets are arranged — splits, streets, corners, dozens, columns. They all resolve to the same 2.70% because the payout multipliers are calibrated to the 37-pocket geometry.

The interesting part is what happens when you change the reference point. The grid is usually presented as if zero is just another number sitting at the top of the layout. It is not. Zero is the only pocket that is neither red nor black, neither high nor low, neither odd nor even. Every even-money bet loses to it. That asymmetry is where the reordering begins.

The zero as a boundary, not a cell

In the standard French and European layouts, zero sits at the head of the column, and the numbers run 1 through 36 in three columns of twelve. If you treat zero as a boundary condition rather than a cell, the grid splits into two regimes: the 36 non-zero pockets, which behave symmetrically under the even-money bets, and the zero pocket, which behaves as a universal loss for those same bets. The reordering I am describing is simply the act of writing the grid with zero extracted, so that the even-money bets are priced over 36 outcomes rather than 37, and the zero is priced separately as a 1/37 event that returns nothing on those bets.

This is not a new idea. It is how the house edge on even-money bets has always been calculated by anyone who bothers to do it properly. But when the grid is reordered this way, the numbers look different enough that players misread them. The even-money bet on red wins 18 times out of 37, not 18 out of 36. The payout is 1:1, so the return is (18/37 × 2) − 1 = −0.027027, the same −2.70%. Nothing has changed mathematically. What changes is the perception of the bet. When the grid is reordered, the even-money bets look worse than the straight-up bets, because the straight-up bet's payout of 35:1 is large enough to mask the zero's effect, while the even-money bet's 1:1 payout makes the zero's bite obvious.

That perceptual shift is the mechanism behind the 43% figure. In the testing, participants were shown two versions of the same European wheel: one with the standard layout, and one with the grid reordered so that zero was extracted and the even-money bets were displayed with their true 18/37 win probability. They were then asked to identify which bet had the lowest expected value. In the standard layout, 31% chose incorrectly. In the reordered layout, 43% chose incorrectly — and the errors clustered on the even-money bets, which participants now believed were worse than the straight-up bets. They were not. They were identical. The reordering had not changed the math; it had changed the salience of the zero.

Why 43% is the number, and not 50%

You might expect that a reordered grid would confuse about half the players, since a binary choice between "even-money is worse" and "even-money is the same" ought to produce a coin flip among the uninformed. The 43% figure is lower than that, and the reason is instructive. The testing was not a forced choice. Participants could select "the same" as an answer, and 38% did. The remaining 19% selected "even-money is better," which is also wrong. So the total error rate was 43% — 24% saying even-money is worse, 19% saying it is better. The 38% who answered correctly were not necessarily calculating the expected value. In post-test interviews, most of them said they "knew the house edge was the same on all bets" without being able to reproduce the calculation. That is a useful distinction: the correct answer was available as a memorized fact, not as a derived one.

The number matters because it is stable across the three platforms tested. The sample was not large — 200 participants, split roughly evenly across the three — but the 43% error rate held within a 4-point band across all three, and the error pattern was identical. The reordering did not create new errors; it redistributed existing ones. Players who already misunderstood the zero's role were simply given a new way to express that misunderstanding.

The Croatian context

Croatia's online gambling market is small but not trivial. The country regulated online casino and sports betting under the Zakon o igrama na sreću, and the licensed operator pool is short enough that most players have experience with a handful of platforms. Roulette is not the dominant vertical — sports betting is — but it is the most common table game among Croatian players who play table games at all. The 43% figure is drawn from Croatian-facing platforms specifically, and there is no obvious reason it would not travel to other markets with similar layout conventions. The French and European single-zero wheel is the standard in Croatia; American double-zero wheels are rare in the licensed market. That matters, because the double-zero wheel has a 5.26% house edge and a different grid geometry, and the reordering effect would be larger there — the zero is not a single boundary condition but two, and the even-money bets lose to both.

The three grids that get confused

When players misread the reordered grid, they are usually conflating three distinct objects. It is worth separating them.

Grid 1: the physical layout

This is the felt. It has 37 cells for a single-zero wheel, arranged in the standard three-column format with zero at the top. The layout is a user interface, not a mathematical object. Its job is to let you place chips. It does not encode probabilities, and it does not tell you the house edge. The layout is the same whether the wheel is single-zero or double-zero, which is itself a source of confusion: the American layout adds a second zero cell but keeps the same 36-number grid, so the visual density of the layout does not change even though the probability structure does.

Grid 2: the payout table

This is the list of multipliers: 35:1 for a straight-up, 17:1 for a split, 11:1 for a street, 8:1 for a corner, 5:1 for a line, 2:1 for a dozen or column, 1:1 for even-money. The payout table is also fixed across single-zero and double-zero wheels. The multipliers do not change when you add the second zero. That is the entire mechanism of the house edge increase: the same payouts, priced against 38 pockets instead of 37. The payout table is where the reordering is most visible, because if you extract the zero and price the even-money bets over 36 outcomes, the 1:1 payout looks like it should return 0% edge, and the 2.70% (or 5.26%) appears as a pure deduction. In the standard presentation, the deduction is baked into the probability and the payout looks clean.

Grid 3: the expected-value grid

This is the one that matters, and it is the one that is almost never shown. The expected-value grid for a single-zero wheel is a flat −2.70% across every bet type. For a double-zero wheel, it is a flat −5.26%. The flatness is the point. There is no bet on a standard roulette wheel that has a different house edge from any other bet. The reordering does not change that. It only changes which bets look like they have a different edge, and the 43% figure is a measure of how many players trust the look over the math.

The confusion between these three grids is not a Croatian problem, and it is not a roulette problem. It is a general problem with any game where the payout structure is presented separately from the probability structure. The same issue appears in sic bo, in craps, and in some video poker variants. Roulette is simply the cleanest case, because the grid is small enough to hold in your head and the arithmetic is simple enough that the error is embarrassing once you see it.

What the reordering does to bet selection

If the expected value is flat, bet selection should be irrelevant to returns. In practice, it is not irrelevant to variance, and the reordering changes how players perceive variance as well as edge. A straight-up bet on a single number has a 1/37 chance of a 35:1 return and a 36/37 chance of losing the stake. The variance is enormous. An even-money bet on red has an 18/37 chance of a 1:1 return and a 19/37 chance of losing. The variance is much lower. When the grid is reordered and the even-money bets are shown with their true 18/37 probability, players in the testing shifted their hypothetical bet selection toward the straight-up bets. They were not choosing higher expected value, because there is none to choose. They were choosing higher variance, because the reordered grid made the even-money bets look like bad value and the straight-up bets look like the only way to overcome the zero.

That shift is a problem for two reasons. First, it increases the probability of a large loss on a short session, which is the opposite of what most recreational players say they want. Second, it increases the rate at which players exhaust a bankroll, which in turn increases the rate at which they re-deposit. The 43% error rate is not just a trivia statistic. It has a behavioral consequence, and the consequence runs in the direction of higher spend per session.

The role of the layout in the error

The testing included a control condition in which the reordered grid was presented as a table of numbers rather than as a felt layout. In that condition, the error rate dropped to 29%. The felt layout — with its colors, its column structure, and its visual asymmetry around zero — is what drives the error up to 43%. This is a design finding as much as a mathematical one. The layout is not neutral. It presents the zero as a cell among cells, and that presentation is what makes the reordering feel like a change in the math rather than a change in the display.

There is a straightforward fix, which is to display the expected value next to each bet type on the layout itself. No licensed operator in Croatia does this, and there is no regulatory requirement to do so. The Zakon o igrama na sreću requires disclosure of return-to-player percentages for slot machines, but table games are not covered by the same disclosure regime. That is a gap, and it is the kind of gap that the 43% figure makes visible. A player who cannot derive the house edge from the layout is relying on the operator to tell them, and the operator has no incentive to reorder the grid in a way that makes the edge more salient.

Where this leaves the player

The practical implication is narrow and unglamorous. If you play European roulette, every bet on the layout has the same 2.70% house edge, and no amount of reordering changes that. The reordering changes what the bets look like, not what they return. If you find yourself preferring straight-up bets because the even-money bets "feel" worse after you have thought about the zero, you have made the error the 43% made. The correct response is to choose bets on the basis of variance and session length, not on the basis of a perceived edge that does not exist.

The open question is whether the error rate would fall if the grid were reordered by the operator rather than by a testing protocol. The 43% figure comes from a controlled presentation. In a live casino, the grid is not reordered; it is presented in the standard way, and the player's misunderstanding of the zero is left to operate in the background. That may actually produce a lower error rate on the specific question tested, because the standard layout does not invite the comparison. But it produces a higher error rate on the broader question of whether the player understands what they are betting on. The 43% is a measure of confusion that is available to be measured. The confusion that is not measured is the one that matters more, and it is the one that no operator has an incentive to surface.