Data and methodology
How the play data on this site is collected, what it measures, and the specific ways it is wrong. For how the compliance score is built, see how we score.
Where the data comes from
Every casino below publishes a live bet feed on its own site, readable without an account. We read those feeds directly and count the results ourselves. We do not buy this data, take it from another analytics site, or accept figures supplied by an operator.
| Casino | Read via | What each bet carries |
|---|---|---|
| Stake.com | GraphQL | game, studio, stake, multiplier, currency, timestamp, and the operator's own declared house edge |
| Winna | REST | game, studio, stake, multiplier, currency, timestamp |
| Roobet | REST | game, studio, stake, multiplier, timestamp, amounts already in USD |
| Shuffle | GraphQL | game, stake, multiplier, currency. No studio and no timestamp |
Only Stake publishes a house edge, so it is the only operator where our observed figure can be set against the operator's own. For the others there is nothing to compare against and we say so on the row rather than leaving a blank.
How often we read
A feed is a rolling window, so the only thing that matters is reading it faster than it turns over. Window depth differs per operator, so the cadence does too.
| Casino | Read every | Window |
|---|---|---|
| Stake.com | 30s | 50 rows, ~64s |
| Winna | 15s | 2 windows x 50 rows, ~29s, rate limited |
| Roobet | 6s | 11 rows, fixed |
| Shuffle | 6s | 10 rows, fixed |
How observed return is calculated
Return is the mean payout multiplier. A multiplier is dimensionless, so bets placed in usdt, ltc, eth, sol and inr all average together without an exchange rate. A stake-weighted figure would need an FX rate for every currency at the moment of every bet and would import that error into every number we publish. This costs a little precision and buys correctness.
Every figure carries its sample and its interval. Slot returns are driven by rare large multipliers, so a mean over a few hundred rounds can sit far from the true rate by luck alone. We publish the 95% interval next to the figure. Below 1,000 observed bets a figure is shown only with its interval and never as a settled rate; below 100 we do not show it at all and the row says "collecting".
Feed curation is reported where it can be proven. No casino game returns more than it takes over the long run. So an observed return above 100% on a real sample is not a generous game, it is proof the feed is not publishing every round, and we label it "curated feed" rather than showing it as a finding. Below 100% we cannot establish it either way from the data, and we say the figure is uncorrected rather than implying it is clean.
What every number is tagged as
observed | Counted directly from the operator's public feed. |
derived | Computed from observed data through the method described above. |
reported | Published by the operator. We repeat it, we do not verify it. |
On-chain attribution
Money moving inside the games is public at every casino we read. Money moving out to players is private at all of them, so any payout figure has to come from the chain, which means knowing which addresses belong to whom. Every address we attribute is published with the method behind it and must trace back to an address the operator itself handed us. See the wallet attribution registry. No addresses are attributed yet, which is why no deposit or payout figures appear anywhere on this site.
What we do not claim
We do not claim this is every bet.
Each feed is a rolling window the operator publishes on its own site. We read the window; we do not see the stream behind it. Every figure here is a sample.
We do not claim the sample is unbiased.
A bet ticker is a marketing surface, and the selection leans toward winning rounds. Our observed return is uncorrected for that and is biased upward by an amount we cannot measure, which differs between operators and therefore does not cancel out when two are compared.
We do not claim a game's true RTP.
A game's certified return is set by its studio and measured over hundreds of millions of rounds. Nothing here replaces that. Where an operator publishes a house edge we show it beside our figure so the two can be compared, and the operator's figure is the one with the audit behind it.
We do not publish deposits, withdrawals, revenue or profit.
Those need on-chain wallet attribution, which we have not built. Other analytics sites publish them as modelled ranges over assumed turnover multiples and assumed game mixes. We would rather publish nothing than publish a model output that reads like a measurement.
We do not predict anything.
Past rounds do not indicate future rounds. A game that returned above its declared rate this week is not due to return below it next week, and nothing on this site should be read as a reason to play.
Independence
Niekie carries affiliate links to some operators. The collection pipeline reads every casino's feed identically and has no knowledge of which operators we have a commercial relationship with; the readers are selected by which casinos publish a readable feed and nothing else. No operator can pay to change a score, a return figure, or a position in any table on this site.
Corrections
If a figure here is wrong, tell us and we will fix it and say that we did. Contact us. Operators are welcome to dispute a number; a disputed figure is checked against the stored observations behind it rather than removed.
Coverage right now
535,321 bets across 7,630 games from 4 operators in the last 7 days. Coverage is a floor: we see what these operators publish and no more.