Almost everything written about “reading” roulette results is wrong. Past spins do not predict future ones, and a run of reds says nothing about the next spin.
There is one exception, and it is a real one: a wheel that is physically faulty. This is the only circumstance in which a history of results carries information — and the bar for proving it is much higher than most people expect.
What bias actually is
A roulette wheel is a precision object that lives a hard life. It gets spun thousands of times a week, knocked, cleaned and moved. Over time things go wrong: a fret between pockets works loose and deadens the ball, the base settles slightly out of level, a pocket wears deeper than its neighbours.
When that happens the wheel stops being fair. Certain pockets genuinely catch the ball more often than one in thirty-seven. That is not a pattern in the results — it is a defect in the object, and the results are merely the evidence.
The man who did it
Joseph Jagger was an English mechanic who understood machinery wears unevenly. In the 1870s he paid clerks to record every result on several wheels at Monte Carlo, then went through the sheets looking for numbers appearing more often than chance allowed.
One wheel showed it. He bet the biased numbers and won a very large sum before the casino worked out what was happening and began moving wheels between tables overnight — which broke his records, because his data belonged to one wheel and he could no longer tell which one he was at.
That detail matters more than the winnings. A bias belongs to one physical wheel, not to roulette, not to a casino, and not to a table position. Mix two wheels' results together and you have destroyed the only thing you were measuring.
How much data it honestly takes
This is where nearly every modern attempt falls apart.
On a single-zero wheel each number should land once in thirty-seven. After 370 spins you would expect about ten hits each — but ordinary randomness produces a range of roughly four to seventeen. Some numbers will look scorching. None of it means anything.
To separate a modest genuine bias from that noise you need something in the region of five to ten thousand spins on the same wheel, and a proper statistical test rather than a glance at a leaderboard. The standard tool is a chi-square test, which asks one question: how likely is a spread this uneven from a wheel that is actually fair?
At a live table producing a spin roughly every forty seconds, five thousand spins is somewhere near sixty hours of recording. That is the real price of the technique, and it is why almost nobody completes it.
Look for neighbours, not numbers
A useful sanity check: physical defects are physical. A loose fret or a dip in the track affects a run of pockets sitting next to each other on the wheel, not scattered numbers across it.
So if your hot numbers are neighbours on the rim, that is consistent with a real fault. If they are scattered at random around the wheel, the far likelier explanation is chance — or a mistake in your recording.
Why it is nearly dead
Casinos are not passive about this. Modern wheels are precision-made, levelled and serviced on a schedule, rotated between tables, and monitored with the casino's own outcome statistics. A wheel drifting from expectation gets pulled long before a player could gather enough data on it.
The window Jagger walked through has been deliberately closed.
And it cannot apply here
One thing worth stating plainly: none of this touches an online game. Our roulette has no wheel, no frets and no bearing to wear — results come from a random number generator, which has no physical state to go wrong. There is nothing to find, and a hot number here is only a number that happened recently.
