The open datasets
Every register on this site is published as a CSV, generated at build time from the same files the pages render. A download and the table above it are two renderings of one set of records rather than two transcriptions, so the file cannot disagree with the page. Free to reuse and cite with attribution.
The datasets
| Dataset | Rows | Columns | Updated | Register | Method | Download |
|---|---|---|---|---|---|---|
| Casino payout index | 24 | 21 | 2026-10-02 | The casino register | Casino method | CSV |
| Exchange register | 24 | 28 | 2026-10-09 | The exchange register | Exchange method | CSV |
| Wallet register | 30 | 24 | 2026-10-07 | The wallet register | Wallet method | CSV |
| Provably fair implementations | 14 | 16 | 2026-09-22 | The implementations register | Provably fair method | CSV |
| VPN and location clause sweep | 24 | 10 | 2026-09-22 | The VPN sweep | Clause watch method | CSV |
| On-chain deposit volume, measured by Tanzanite | 24 | 17 | 2026-09-22 | Deposit volume | Why we cite it | CSV |
Casino payout index
One row per operator on the casino register. Every operator on the casino register, with the fields the payout score is computed from. The score itself is not a column: it is derived from these fields by the published formula, so a reader who disagrees with the formula can recompute the whole table from the inputs.
What it does not cover
- measured_median_minutes is empty on all 24 rows. No withdrawal test has been run, and the column is empty rather than filled with the operator's own stated time.
- hot_wallet_identified reads no on 24 of 24 rows. On-chain health can only be measured where a payout wallet has been identified.
- licence_verified_in_register false is not evidence that a licence is invalid. Several issuing authorities publish no queryable register at all.
Columns
Exchange register
One row per exchange. Which legal entity a user contracts with, which registrations were found in an issuing authority's own register, what a proof-of-reserves exercise is signed by, and which of six clause topics the binding document contains. There is no score column, because this register publishes no score.
What it does not cover
- confirmed_authorisations counts only registrations found in the issuing register. It is 0 on 10 of 24 rows, which records a failed confirmation and not a finding that no authorisation exists.
- binding_document_unread is true on 1 of 24 rows. Where it is true, every clause column on that row describes the European document only and must not be read across to a user outside the European Economic Area.
- An empty adverse_listings cell means no listing was found by the searches recorded in the file. It is not a search of every register in the world.
Columns
Wallet register
One row per wallet. What the marketing says about custody, what the binding document says about it, how many sub-service clauses route the user to a counterparty they never contract with by name, and whether anybody in the stack is disclosed as able to block an address. No score column.
What it does not cover
- sub_service_clauses totals 95 across 30 wallets and counts recorded clauses, not distinct companies. Two clauses can name the same counterparty.
- marketed_as_non_custodial is computed from the words in the recorded marketing quotes. A wallet that makes no custody claim in its marketing reads false, which is a finding about the marketing and not about the product.
- An empty liability_cap cell means no capping clause was collected, not that liability is uncapped.
Columns
Provably fair implementations
One row per operator on the casino register. Every operator on the casino register, whether it publishes a fairness scheme, whether that scheme is specified closely enough for a third party to recompute a result from it, and which games the scheme covers and leaves out.
What it does not cover
- Every boolean column is tri-state. An empty cell means the documentation does not say or could not be rendered, and 12 of 14 rows have an empty verifier_runs_client_side for that reason.
- 1 of 14 rows could not be retrieved at all. Those rows are excluded from every denominator the register publishes, and nothing is recorded about them.
- games_covered on a row where publishes reads false records the scope of the operator's own assertion, not of any documentation.
Columns
VPN and location clause sweep
One row per operator on the casino register. One clause read across every operator on the casino register: whether the binding terms prohibit a VPN, proxy or similar tool, what consequence is stated, whether the consequence is discretionary, and whether the drafting separates privacy from evasion.
What it does not cover
- document_retrieved reads false on 4 of 24 rows. Every other column on those rows is empty, no retrieval date is claimed, and the rows are counted in no published figure.
- The honest denominator for every ratio drawn from this file is 20 documents read, of 24 operators on the register.
- 1 of the 3 sweeps on the queue has been read. This file covers that one clause and says nothing about dormancy or unilateral amendment.
Columns
On-chain deposit volume, measured by Tanzanite
One row per operator on the casino register. Crypto deposits arriving at operator deposit addresses over one month, with the count of deposits, the change against the previous month and the chains they arrived on. The average deposit column is computed from the other two.
What it does not cover
- This is the one dataset here whose figures were measured by somebody else. Every row carries source, source_url and retrieved for that reason, and nothing in this file feeds the payout score.
- tracked_by_source reads false on 15 of 24 rows. Every figure on those rows is empty and not_tracked_reason says why, because an empty cell on its own reads as a zero.
- It counts money arriving and not money wagered, and it is not a measure of what an operator holds or pays out. No payout wallet has been identified for any operator on the register, by us or by the source.
- The window is 10 buckets of 3 days from 2026-08-25, and the final bucket was still open when the figures were read.
Columns
The casino index in full
The oldest of the five, printed here because it is the one most often cited and because the score in it is derived rather than stored. It is not a column in the CSV: it is computed from the columns by the published formula, so a reader who disagrees with the weighting can recompute the whole table from the inputs.
| Operator | Status | Score | Confidence | Licence in register | Register status | Measured median (min) | Wallet 30d | High conflicts | Casino Guru | Reviewed |
|---|---|---|---|---|---|---|---|---|---|---|
| BC.GAME | watchlist | 54 | low | yes | Valid | not tested | no wallet | 1 | 7.8 | 2026-09-02 |
| BetFury | watchlist | 44 | low | yes | Assessment in progress | not tested | no wallet | 5 | 9.1 | 2026-09-02 |
| Bets.io | watchlist | 58 | low | yes | Valid | not tested | no wallet | 0 | not recorded | 2026-10-02 |
| Bitcasino | watchlist | 73 | low | yes | Indefinite | not tested | no wallet | 0 | not recorded | 2026-10-02 |
| Cloudbet | watchlist | 58 | low | yes | Assessment in progress | not tested | no wallet | 1 | 9.2 | 2026-09-02 |
| Coins.game | watchlist | 47 | low | yes | Valid | not tested | no wallet | 0 | not recorded | 2026-10-02 |
| Duelbits | watchlist | 50 | low | yes | Assessment in progress | not tested | no wallet | 1 | not recorded | 2026-10-02 |
| Flush | watchlist | 32 | low | no | not recorded | not tested | no wallet | 1 | 8.3 | 2026-09-02 |
| FortuneJack | watchlist | 49 | low | yes | Valid | not tested | no wallet | 1 | not recorded | 2026-10-02 |
| Gamdom | watchlist | 26 | low | no | not recorded | not tested | no wallet | 3 | 9.8 | 2026-09-02 |
| Jackbit | watchlist | 44 | low | yes | Assessment in progress | not tested | no wallet | 4 | 8.3 | 2026-09-02 |
| Lucky Block | watchlist | 18 | low | no | not recorded | not tested | no wallet | 1 | not recorded | 2026-10-02 |
| mBit | watchlist | 76 | low | yes | Indefinite | not tested | no wallet | 0 | not recorded | 2026-10-02 |
| Metaspins | watchlist | 53 | low | yes | Valid | not tested | no wallet | 1 | not recorded | 2026-10-02 |
| Rakebit | watchlist | 50 | low | yes | Valid | not tested | no wallet | 3 | 8.1 | 2026-09-02 |
| Rollbit | watchlist | 39 | low | yes | Expired | not tested | no wallet | 1 | not recorded | 2026-10-02 |
| Roobet | watchlist | 69 | low | yes | Indefinite | not tested | no wallet | 1 | 7.9 | 2026-09-02 |
| Shuffle | watchlist | 57 | low | yes | Indefinite | not tested | no wallet | 3 | 4.8 | 2026-09-02 |
| Sportsbet.io | watchlist | 64 | low | yes | Active, with no expiry recorded | not tested | no wallet | 1 | 5.9 | 2026-09-02 |
| Stake | watchlist | 49 | low | yes | Assessment in progress | not tested | no wallet | 2 | 7.9 | 2026-09-02 |
| Thunderpick | watchlist | 47 | low | yes | Assessment in progress | not tested | no wallet | 1 | 2.9 | 2026-09-02 |
| Vave | watchlist | 9 | low | no | not recorded | not tested | no wallet | 3 | not recorded | 2026-10-02 |
| Wild.io | watchlist | 34 | low | no | not recorded | not tested | no wallet | 0 | 8.3 | 2026-09-02 |
| Winz.io | watchlist | 19 | low | no | not recorded | not tested | no wallet | 4 | 8.2 | 2026-09-02 |
What the casino columns mean
Licence in register is the single most useful column. It records whether we found the operator’s licence in the issuing authority’s own register, not whether the operator claims one. Most do claim one. Fewer can be confirmed.
Measured median is our own withdrawal test, in minutes, not the operator’s stated processing time. Where it reads not tested we have not deposited yet and we say so instead of reprinting the operator’s figure as though we had checked it.
Wallet 30d is the balance trend of the operator’s public hot wallet over 30 days. A wallet draining fast is a leading indicator of payment trouble; complaint volume is a lagging one. Where no public wallet has been identified, the column says so.
High conflicts counts documented contradictions between an operator’s marketing and its own binding terms or its licence register. Each one is cited on both sides on the operator’s page.
How to read an empty cell
Four tokens govern absence across every page on this site, and a CSV cannot carry them, so an empty cell in a file is ambiguous in a way the page is not. Each dataset above names which kind of absence its empty cells are, and every file that can distinguish them carries a retrieval column beside the field. A record whose document could not be retrieved asserts nothing at all: every derived column on it is empty and no retrieval date is claimed. The vocabulary itself is at the method hub.
Citing these datasets
Cite as: CryptoGuardz, dataset name, retrieved 2026-10-09, https://cryptoguardz.com/payout-index/. Licensed CC BY 4.0. If you are writing about a specific company, the underlying sources are listed with their retrieval dates on that company’s own entry, and the entry is the better citation because it carries the quotes these files only summarise.
Method and limits
One method page per register, at the method hub, each stating what that register measures, what it never measures and why four of the five publish no score. The casino scoring formula is at how we rate operators and is generated from the function that computes it.
Two limits worth stating plainly on every file here. Absence of confirmation is not evidence of falsity: several issuing authorities publish no queryable register at all, and a page that would not load is recorded as a page that would not load. And every one of these datasets is documentary. They record what companies published and what registers held on the dates printed beside each value. None of them records what any company did.
Errors in a file are corrections like any other. Send the primary source to the contact page and it is published with its date at corrections, with the change logged at the change log.