Monthly data release
The Niekie Review, July 2026
Measurements from 653,995 casino rounds broadcast across 6,875 games between 2026-07-01 and 2026-07-31, counted from operators' own public feeds. Generated, not written.
Feed completeness
How much of its own play each operator publishes. A real casino game loses on most rounds, so a feed reporting few losses at a high aggregate return is broadcasting a selection of wins rather than a record of play. Every figure below comes from the operator's own endpoint and can be recounted from it.
We earn a commission from every operator measured here: Roobet, Winna, Shuffle, Stake.com. That is disclosed because it cuts against us, not for us. The two feeds this table treats worst, Stake.com and Winna, are among them. Every figure is computed by the same query, from the operator's own endpoint, and the scoring code cannot read which of them pay. How we score.
| Feed | Games | Rounds broadcast | Rounds that lost | Aggregate return | Reading | Our relationship |
|---|---|---|---|---|---|---|
| Roobet | 3,613 | 336,303 | 86.2% | 0.999x | Consistent with a feed of ordinary play | Paid link |
| Winna | 1,029 | 177,953 | 32.7% | 2.652x | Wins heavily over-represented | Paid link |
| Shuffle | 1,924 | 108,275 | 66.8% | 0.960x | Consistent with a feed of ordinary play | Paid link |
| Stake.com | 1,903 | 31,464 | 56.7% | 9.621x | Wins heavily over-represented | Paid link |
What this does not measure
Three things other than concealment move these numbers, and none of them can be separated out from a public feed alone.
- Game mix. Crash and dice lose on a different share of rounds than slots, so an operator whose players favour one will differ from one whose players favour the other, with both publishing everything.
- Which endpoint exists. Some operators publish a feed of biggest wins by design. That is a product decision about a marketing widget, not evidence of hiding anything.
- Unequal windows. Feeds are still being added, so a feed collected for part of a month sits beside one collected for all of it.
This is a measurement of what an operator broadcasts. It is not a measure of payout speed, of fairness, or of any game's return to player, and it must not be read as one. For what is measured about payouts, see how we score.
The corpus behind this issue
Take the data
The figures on this page, as published, in both formats. Re-run them against the operators' feeds and you should get the same numbers.
SHA-256 of this issue's figures:2237d93cae2209dd0597050308be3cbfe3d617ba078e3ebd0a3eb90ace36f90a
The hash covers the measurements, not the page. A published issue never changes, so this value is
stable and a download can be checked against it.
Earlier issues
August 2026 June 2026 May 2026 April 2026 March 2026 February 2026 January 2026 December 2025 November 2025 October 2025 September 2025