Can a roulette wheel be biased? Wear, levelling and what testing checks

Yes, a physical wheel can in principle be biased: some pockets then land more often than 1/37 (single zero) or 1/38 (double zero). Bias is a measurable departure from equal chances, and a short run of spins cannot show it. Kenyan rules require certified games and displayed odds, but they do not set roulette-wheel procedures.

By Nairobi Ultimate Editorial (prepared with AI tools under our editorial policy) · Last updated 2026-10-30

What "biased" means

A fair single-zero wheel has 37 pockets, and each should receive the ball with probability 1/37, about 2.70%. A fair double-zero wheel has 38 pockets, each with probability 1/38, about 2.63%. A wheel is biased when one or more pockets have a lasting probability different from that, so the chances are no longer equal.

This is different from a streak. A fair wheel still produces runs, repeats and long gaps, as the guide to ten reds in a row shows. Bias is about the long-run frequency of a pocket, not about what the last few spins looked like.

How a physical wheel could become uneven

The following causes are described in general terms only. No figure is given for how much any of them changes the odds, because none is stated in the Kenyan rules read for this guide.

• Wear. Pocket dividers (often called frets) and the pocket floors meet the ball thousands of times. If some wear faster than others, the way the ball bounces and settles could shift slightly.
• Levelling. If the wheel head is not level, the ball can drop from the track more often on one side of the bowl. The wheel then no longer treats all pockets alike.
• Ball track and rotor. A dent, scratch or uneven track, or a rotor that does not spin smoothly, could affect where the ball leaves the track.
• Damage or dirt. A chipped divider or debris in a pocket could change how the ball behaves there.

These are reasons why a physical wheel needs inspection and maintenance. Whether any given wheel is affected is a question of measurement, not of appearance.

What a house edge assumes

The edge printed in the rules is derived from equal pockets and a fixed payout. A straight-up bet pays 35 to 1, meaning a winning stake of 1 returns 35 plus the stake. The derivation is set out in full in the guide to the roulette house edge; the key lines are repeated here.

Both lines use the probability 1/37 or 1/38 for the chosen pocket. If a real pocket had a different probability, the line would use that value instead, and the average result would move away from the printed figure. The wheel types themselves are compared in the guide to European and American roulette.

Why bias needs many spins to measure

Each spin lands in one pocket, so the number of times one chosen pocket appears in N spins has an expected value of N x 1/37. Its spread is measured by the standard deviation, the square root of N x p x (1 - p), explained in the guide to standard deviation in casino games. For a single-zero wheel p = 1/37 and 1 - p = 36/37.

The numbers come out exactly when N is 1,369 (which is 37 squared) times a perfect square, because then N x 36/1,369 is itself a perfect square.

The same working for larger runs gives the table below. The last column is the standard deviation as a share of the expected count; it shrinks slowly as the number of spins grows.

Even after 49,284 spins an honest wheel's pocket can easily finish 36 above or below 1,332. A tester therefore needs a deviation well beyond this range before suspecting a fault. Because 37 pockets are being watched at once, the single most extreme pocket will often look unusual by chance alone, so a test has to allow for that.

How small a bias can be and still matter

The straight-up result per unit staked on a pocket with probability p is 35p - (1 - p) = 36p - 1. At p = 1/37 this is -1/37. It reaches zero at p = 1/36.

This is why regulators and operators rely on repeated, large-sample checks and physical inspection, not on watching results. It is also why a printed edge cannot be audited from a screen of recent numbers. An unusual-looking run is not evidence, in line with the points in the guide to the gambler's fallacy and session limits.

What testing under the Kenyan rules covers

The Gambling Control Act, 2025 defines "gambling equipment" as including a gambling table, board and any instrument used in connection with gambling, and defines a "testing agent" as a person licensed to test and calibrate gambling machines, equipment or devices. Form 11 of the Licensing Regulations, 2026 asks a testing laboratory to state its scope, with listed options such as RNG evaluation and certification, game logic and maths verification, hardware integrity and security testing, and payout percentage verification.

The Conduct of Gambling Operations Regulations, 2026 add several general duties. Regulation 20 says casino games must be authorised by the Authority and that a licensee shall display clearly the rules, odds, house edge and average return to the player. Regulation 58(1) requires a licensee to operate gaming equipment meeting the technical standards prescribed by the Authority. Regulation 59(3) and (4) say that imported machines are inspected and tested by the Authority, and that, where the Authority advises, they undergo standardisation testing by the Kenya Bureau of Standards or another approved entity.

None of these provisions mentions roulette, wheel levelling, or a statistical test for pocket frequencies. They are general rules for equipment, so this guide does not claim that they prescribe any particular wheel check.

Live studio wheel versus virtual roulette

In a live game a real wheel is spun in a studio and the result is read from the wheel itself, as described in the guide to how live dealer games work. Such a wheel is subject to the physical points above. A virtual roulette game has no wheel: a random number generator selects a pocket, and the question becomes whether the generator and game logic match the declared rules.

For the virtual case the Regulations are more specific. Regulation 45(2) requires the RNG used for games of chance to be certified by the Authority or its approved agent to statistically test the randomness and the theoretical return to player, and Regulation 45(3) requires game rules to be evaluated or certified annually. Regulation 47(2) provides that the Authority shall annually conduct and certify, or procure an independent entity to carry out, a test and audit of the approved online gambling system or platform. See how RNG testing works in Kenya for the wider process.

What this means for a player

The house edge, 1/37 or 2/38, is the average cost of play when the wheel is fair, and it applies to every standard bet whatever the previous spins were. Check the displayed rules, odds and edge before staking, and decide your limits in advance. Gambling is addictive: play responsibly, and only if you are 18 or over. If it stops being a limited pastime, the responsible gambling page lists the GRA toll-free line and self-exclusion.