Pearson & the Monte Carlo Roulette Wheel
The 19th-century statistics project that helped invent the chi-squared test — using roulette data.
Long before casinos published RNG certifications, one of the founders of modern statistics tried to answer a simple question: does a real roulette wheel actually behave the way pure probability theory says it should?
Study overview
In the early 1890s, the Monte Carlo casino's outcomes were regularly published in a local newspaper, Le Monaco. Statistician Karl Pearson, looking for large, real-world data sets to test his emerging theories of probability, used this published data as a stand-in for genuinely random events — a kind of 19th-century "big data" source, long before computers could simulate randomness directly.
Methodology
Pearson compared the observed frequency of outcomes reported in the newspaper — colors, columns, and individual numbers across thousands of recorded spins — against the frequencies a perfectly fair, independent wheel should theoretically produce. Historical accounts of his notes describe him analyzing over 16,000 recorded throws collected over roughly four weeks of play, comparing the standard deviation of the actual results to the theoretical standard deviation predicted by chance alone.
Key findings
Pearson found that the newspaper-reported results deviated from what a truly fair, memoryless wheel would be expected to produce by more than chance alone could explain. At the time, he suggested this pointed toward genuine bias in the physical wheels. What the study is actually best known for today, though, is methodological: to make this kind of comparison rigorous, Pearson's work in this period helped lay the groundwork for the chi-squared goodness-of-fit test — now one of the most widely used tools in all of statistics, applied everywhere from medicine to quality control.
Limitations
This is where evidence and interpretation need to be kept separate. Later historians of statistics have pointed out a key weakness in Pearson's source data: the newspaper results were not a verified, complete record of every spin, and how they were compiled for publication is not fully documented. Most modern accounts of this episode conclude that the apparent "bias" Pearson detected more likely reflects problems with the newspaper's reporting process than genuine mechanical bias in the actual wheels. Pearson never independently tested a physical wheel himself under controlled conditions.
Practical meaning
For a modern player, this episode isn't evidence that roulette wheels are secretly biased — quite the opposite, it's a caution about how easily "the data looks off" conclusions can be produced from imperfect data, human recording, or too small a sample. Its real legacy is the statistical method it helped produce: the same chi-squared logic Pearson developed partly through this roulette data is exactly what regulators and labs use today to certify that a physical wheel or an RNG is behaving fairly. It's a good reminder that "the wheel looks non-random to me" is a hypothesis to test rigorously, not a conclusion to trust on its own.