Why an Average Can Mislead
Suppose two casinos both average an 8 out of 10 for payout satisfaction. On the surface they look identical, but the average hides how spread out the answers were. If one casino's players almost all said 8, and the other's answers ranged from delighted 10s to frustrated 4s, those are very different experiences wearing the same number. That spread is called variance, and ignoring it is one of the most common ways rating sites end up publishing figures that feel wrong to real players.
We ran the California survey data through statistical software, using variance analysis to understand not just where each casino landed on average but how consistent that experience was. A high average with low variance is a genuinely dependable site; a high average with wide variance is a coin flip that happens to have paid off for some players.
T-Tests: Is the Difference Real?
When one casino scores higher than another, the obvious question is whether the gap is real or just an artifact of who happened to answer. A t-test answers exactly that. It compares the two groups of responses and estimates the probability that a difference of the size we observed could have appeared by chance alone. When that probability, the p-value, falls below the conventional 0.05 threshold, we treat the difference as statistically significant and let it influence the score. When it does not, we hold back rather than pretend a coin-flip gap is meaningful.
Confidence Intervals: Showing Our Uncertainty
Every survey figure is an estimate, so we also calculated confidence intervals: the range within which the true value most plausibly sits. A narrow interval means we can be fairly precise about a casino's rating; a wide one is a signal to be cautious, usually because fewer players in the sample had first-hand experience with that site. This is why a strong score backed by many consistent responses carries more weight in our rankings than an equally high score built on a thin, scattered sample.
How to Read the Statistics
Two cautions are worth keeping in mind. First, statistical significance is not the same as practical importance: a difference can be real yet too small to change your choice of casino. Second, these tests describe the players we surveyed and generalize from them, they do not predict any individual session. With those limits stated plainly, the tests do their job, which is to make sure every gap in the score reflects evidence. The final chapter explains how the tested pillars combine into one number, and the methodology page summarizes the whole approach.
How We Sampled California: Random and Snowball Sampling Across the State
A score is only as honest as the people behind it. We stratified our sample to mirror California itself, then used random and snowball sampling to reach players other surveys miss.
By Chloe MarshMay 21, 2026
From Survey to Score: Weighting the Eight CAC Pillars Out of 100
Survey data on one side, hands-on testing on the other. Here is exactly how the two become eight weighted pillars and one number in a green poker chip.
By Ryan CarriganJune 4, 2026
The CAC team weighs in
The analysts behind the study talk through what this chapter means for a California player.
The crypto-versus-other payout gap is real, not noise: t = 3.31, p = .006, with a large effect. Banking model genuinely separates the field.
It's the single sharpest discriminator in the whole dataset. If you only looked at one pillar, payout would tell you the most.
Region, by contrast, barely moved the needle — eta-squared of .0035. That's what justifies one statewide score instead of six regional ones.
Which is reassuring for a reader: a player in Sacramento and one in San Diego can trust the same number for the same casino.
And we publish confidence intervals, so you see the precision behind each figure rather than a number floating on its own.
Keep reading
Our full 67-page methodology and dataset: a stratified survey of 4,217 verified California players aged 21+, the eight-pillar weighting model, complete data tables and statistical analysis behind every score on this site.
Download the full study (PDF, 67pp) →