Original research
Crypto Casino Withdrawal Study 2026: 21.3% of 1,958 Mediated Complaints Ended With the Player Paid
Across 1,958 closed Casino.Guru complaints at 17 crypto casinos we have reviewed, 417 ended with the player getting paid, 1,520 were rejected, and 21 ended unresolved. That is a player-win rate of 21.3%, or roughly one in five. This is our own computation over our own review corpus, not a figure lifted from anyone else's report.
21.3%
of closed complaints ended with the player paid
417 of 1,958, Casino.Guru
1.1%
ended with the casino defying the mediator
21 of 1,958, Casino.Guru unresolved
24 of 28
casinos have actually used a confiscation lever
86% of the reviewed corpus
24 of 28
defer identity checks to the moment of cashout
the most universal mechanism we found
1.1 months
median time to extract a USD 50,000 win
at published monthly ceilings
r = 0.51
correlation, Casino.Guru Safety Index vs our score
across all 28 reviewed casinos
The finding in one paragraph
Across 1,958 closed Casino.Guru complaints at 17 crypto casinos we have reviewed, 21.3% ended with the player paid and only 1.1% ended with the casino defying the mediator. Large crypto casinos rarely ignore a ruling. They win on the merits, on player silence, or on terms - and 24 of the 28 casinos in this corpus have a documented case of actually keeping a player's funds under one of those terms.
What this number replaces
On 12 July 2026 we published a figure of 20.6% across 1,322 closed complaints at six casinos in our guide to getting your money out of a crypto casino. This study is its successor. The six-casino subset is embedded unchanged in the larger pool and reproduces exactly: 272 of 1,322 is still 20.6%. Extending the corpus from six casinos to seventeen moves the rate to 21.3%. The "roughly one in five" headline survived a near-tripling of the sample, which is the main reason we are willing to stand behind it.
Per-casino, the median player-win rate is 22.7%. The range runs from 0.0% at Cryptorino, where the docket records no mediation win for any player, to 51.9% at Weiss. Rainbet is second-worst at 4.2%. Among the largest brands, mBit is highest at 38.3% and Stake lowest at 15.4%.
Method
The corpus is 28 casinos we have reviewed - 27 published reviews plus one held as a draft because every figure in it is secondary. Seventeen of those 28 have a full, readable Casino.Guru complaint split (resolved / rejected / unresolved); those seventeen are the only ones inside the 1,958-complaint pool. The other eleven are in the corpus for the licence, cap and lever analysis but not the win-rate arithmetic, either because the split was not retrievable on this pass or because the docket is effectively empty.
How cases were counted. A case counts as closed when Casino.Guru has marked it resolved, rejected or unresolved. Open cases are excluded. The player-win rate is resolved divided by closed. We summed per-casino splits rather than reading a platform-level aggregate, so every casino's contribution is auditable against its own review page.
Two platforms, two vocabularies, never blended. Casino.Guru marks a complaint unresolved specifically when its mediator sided with the player and the casino did not comply, so "zero unresolved" is a real and meaningful claim. AskGamblers' Resolved label is a different thing entirely: it routinely means the case was closed in the operator's favour. We proved that on two dockets during vetting, an 18,000 dollar case and a 140,000 dollar case, both labelled resolved, both decided against the player. Every player-win figure on this page is Casino.Guru's metric only. Where we cite an AskGamblers number, it is named as such and never used as evidence a casino treats players well.
Selection bias, stated up front. We choose which casinos to review, and we choose newsworthy and partner-relevant brands. Casinos with visible payout problems are over-selected on purpose, because that is what our readers ask about. This is not a random sample of the online casino market and the 21.3% should not be read as an industry-wide rate. It is the rate inside a deliberately adversarial corpus.
No money was staked. Nothing here rests on a deposit or a withdrawal we made ourselves. This is document and docket research: casino terms and conditions read for section numbers, regulator registers and seal validators checked at source, and complaint dockets read case by case. Our review methodology explains the scoring the corpus is built on.
Finding 1: casinos engage the mediator, then win
The counter-headline to the low player-win rate is that outright defiance is rare. Only 1.1% of closed cases, 21 out of 1,958, ended with the casino ignoring a ruling that went the player's way. The platform-wide unresolved rate recorded in the July dataset that seeded this study is 13.4%. Large crypto casinos are an order of magnitude below it.
That is not the reassurance it first looks like. It means the money is not usually taken by refusing to answer the mediator. It is kept by winning the argument: on the merits, on player abandonment mid-complaint, or on a term the player did not know applied. Casino.Guru's "rejected" bucket mixes all three, which is why every player-win rate here is a floor, not a ceiling.
Finding 2: 86% have actually used a confiscation lever
24 of the 28 reviewed casinos (86%) have at least one documented case where the operator actually kept player funds under a discretionary or terms-based lever - not merely a clause sitting on paper. Two casinos have no documented confiscation of any kind: Jackbit, where the widely repeated "3.3 million dollar confiscation" turned out to be a player's typo for 3,311 dollars that was released within two weeks, and Tsars, which has no complaints filed directly against it at all. Two more, Fairspin and Weiss, show only isolated or defensible use.
The nuance the corpus keeps proving: many of those 24 uses are defensible. Multi-account rings the casino evidenced to the mediator and genuine max-bet breaches are in there. The pattern worth avoiding is not enforcement itself, it is enforcement without disclosed evidence, arriving only at the withdrawal of a win.
Finding 3: the levers, ranked by how many casinos use them
| Lever | Casinos with a documented instance |
|---|---|
| KYC deferred to cashout (the win-then-wall) | 24 of 28 |
| Linked or multi-account accusation at withdrawal | 16 of 28 |
| Discretionary advantage-play or bare terms-breach | 12 of 28 |
| Bonus fine print (retroactive max-bet, wagering on no-wager offers) | 12 of 28 |
| Geo or VPN clause, including retroactive geo change | 10 of 28 |
Deferred KYC is the near-universal mechanism. Verification is pushed to the moment money tries to leave at 24 of 28 casinos. Only Weiss, Fairspin and King Billy run an upfront design, and King Billy gets there by blocking Bitcoin deposits until an ID is submitted. Our KYC policy database holds the clause-level detail: the trigger, the documents, and the deadline clauses that turn a routine check into a locked balance.
Below that, the ranking is: linked or multi-account accusation at withdrawal (16 casinos), then discretionary advantage-play or a bare terms-breach claim (12) tied with bonus fine print such as a retroactive max-bet rule (12), then the geo or VPN clause (10). Two retroactive geo-change confiscations are documented: Winz, where the casino admitted it on the mediator's record for Polish and German balances, and King Billy, in Gibraltar in 2018 and across Russia and Belarus in 2022 to 2023. The largest single geo enforcement on record in the corpus is Cloudbet's 15,000 dollar seizure under its United States restriction clause.
Finding 4: the drip, and what a 50,000 dollar win costs in time
Fourteen reviewed casinos publish a weekly withdrawal cap. The median is EUR/USD 15,000 per week. Converted to published monthly ceilings, the median time to extract a 50,000 dollar win is 1.1 months - but the spread is what matters: from under two days at TrustDice, whose clean crypto accounts face roughly a 1,000,000 dollar monthly ceiling, to 5 months at Vavada's base status tier, where the ceiling is 10,000 dollars a month with an installment clause on top. BC.Game is worse on paper still: its secondary big-winner clause reads as 5 to 10 months, because the monthly allowance halves once a balance reaches ten times deposits.
The tier ordering inverts at the top. The highest published ceilings sit on the weakest paper: Winz at 100,000 a week on an Anjouan and Tobique pair, TrustDice at roughly a million a month on Curacao. Both are untested at that size - the largest public Winz dispute we found is about 11,000 Canadian dollars. The tightest caps sit on verified Curacao licences: Weiss at EUR 4,000 a week, Vavada at 5,000 dollars a week on the base tier. If you want to see what a cap does to a specific win, our cashout time calculator runs the same arithmetic.
Finding 5: no published cap usually means an unpublished throttle
Six reviewed casinos publish no numeric withdrawal cap at all: BitStarz, Cloudbet, Stake, Duelbits, Gamdom and Rollbit. In every one of them, the big-win drag has simply moved somewhere less visible. Cloudbet operates an undisclosed "betting integrity review" hold of up to six months that its own support confirmed is not in the terms. Gamdom states a manual review of up to 90 days on large withdrawals. Stake publishes no processing timeframe anywhere and reserves, in section 8.2, the right not to explain the nature of an investigation at all. Rollbit and Duelbits both run open-ended manual reviews. BitStarz is the mildest case: its FAQ says no limits are enforced, while term 8.7.3 reserves the right to pay substantial wins in monthly installments. An absent cap is not the absence of a throttle.
Finding 6: licence tier predicts recourse, not conduct
Mean trust score by licence tier across the 28 casinos: EU-regulated 3.5, Curacao verified 3.45, Anjouan 1.9, off-licence 1.75. The ordering holds, but within-tier variance swamps it. Anjouan spans Tsars at 3.5 down to BC.Game and FortuneJack at 1.0. Curacao spans BitStarz at 4.5 down to Cloudbet at 3.0.
The sharper signal is licence state, not licence jurisdiction. Both casinos whose Curacao terms had lapsed into "Assessment in progress" when we checked - Duelbits and Rollbit - carry a trust score of 1.0, the lowest in the corpus, and both pair that lapsed paper with unreversed five and six-figure confiscations. A jurisdiction tells you where to complain. Whether the paper is current tells you whether anyone is listening.
The datapoint that keeps this honest runs the other way. The only tier-one regulator finding in the corpus is against an EU-tier casino: the UK Gambling Commission fined TonyBet GBP 442,750 in January 2023 for exactly the practice this study measures, demanding identity documents at withdrawal and confiscating on failure. Tier buys recourse. It does not buy conduct.
| Licence tier | n | Mean trust score | Mean overall score |
|---|---|---|---|
| EU-regulated (Estonia, domain-verified) | 2 | 3.50 | 3.25 |
| Curacao GCB verified (active or indefinite) | 10 | 3.45 | 3.49 |
| Curacao GCB listed, not re-confirmed | 3 | 3.50 | 3.50 |
| Anjouan + Tobique dual (domain-verified) | 2 | 3.00 | 3.05 |
| Anjouan (domain-verified) | 5 | 1.90 | 2.66 |
| Off-licence (NOT FOUND standard met) | 2 | 1.75 | 2.55 |
| Curacao GCB lapsed ("Assessment in progress") | 2 | 1.00 | 2.15 |
| Tobique entity-only | 1 | 1.50 | 2.40 |
Finding 7: safety indexes fail high, not low
Fourteen of the reviewed casinos carry a Casino.Guru Safety Index of 8.5 or above, which the platform labels High or Very High. Four of those fourteen (29%) landed at 3.0 or below on our own five-point score. Across all 28 reviewed casinos the Pearson correlation between the Safety Index and our overall is r = 0.51, so the index explains about a quarter of the variance in what our docket-level reads found.
The two clearest inversions: Gamdom holds a 9.8 Very High index alongside a 950,000 dollar rejected and unpaid case and a stated 90-day review ceiling, and scores 2.8 with us. Winna holds roughly 9.3 while its terms cannot be read before you log in, and scores 2.4.
The failure is asymmetric. Where the index is low, it agrees with us - Rollbit at 4.1 against our 2.2, Cryptorino at 4.9 against our 2.5, Betpanda at 5.4 against our 2.6 - and no reviewed casino scoring below 7 on the index came out above 3.0 with us. The index measures published terms and operator size. It does not measure behaviour at a withdrawal.
| Casino | Casino.Guru Safety Index | Our overall | Gap (index halved to 5-point, minus ours) |
|---|---|---|---|
| FortuneJack | 8.7 High | 1.9 | +2.45 |
| Winna | ~9.3 Very High | 2.4 | +2.25 |
| Gamdom | 9.8 Very High | 2.8 | +2.10 |
| CosmoBet | 8.3 High | 2.6 | +1.55 |
| BC.Game | 6.9 Above average | 1.9 | +1.55 |
| Duelbits | 7.2 Above average | 2.1 | +1.50 |
Finding 8: the extremes, and the publicity effect
The largest documented paid win in the corpus is 1,350,000 dollars at Gamdom - a blackjack balance, paid in full, but only after the player was banned mid-play, the balance sat frozen for about three and a half months, and Casino.Guru mediated. Against that sits a single court-recorded claim of 1,482,162 dollars inside BC.Game's roughly 2.5 million dollar bankruptcy pool from the Curacao court's November 2024 finding against its prior operating entities. The largest paid win and the largest unpaid claim in this corpus are within ten percent of each other in size.
Three operators independently show a pay-only-after-public-escalation pattern. FortuneJack released 73 SOL the same day the player went public. Rollbit unlocked a 20,000 dollar balance within hours of a public thread. BC.Game restored funds, in its own words, following positive feedback from the Pakistani community. Publicity is not a rung on any casino's official escalation ladder, but in these three dockets it is the rung that moved the money.
The full corpus
Every reviewed casino, its Corner Score, licence position, published cap, drip time and the levers documented in its record. Casino.Guru splits are shown only where the full split was readable; a blank is a gap in our retrieval, not a clean sheet. Scroll the table sideways on a narrow screen.
| Casino | Overall | Trust | Licence | Verification state | Published weekly cap | Months to USD 50k | Casino.Guru player-win | Levers documented |
|---|---|---|---|---|---|---|---|---|
| BitStarz | 4.0 | 4.5 | Curacao GCB | Domain-verified | None published | No published cap | 22.7% 42 paid / 142 rejected / 1 unresolved | Bonus max-bet (retroactive), KYC and fake-ID, AML |
| Jackbit | 3.8 | 3.5 | Curacao CGA | Primary-verified | USD 50,000 | 0.5 | Full split not in corpus; zero black points | None documented as used |
| Tsars | 3.8 | 3.5 | Anjouan | Domain-verified | EUR 15,000 | 1.7 | No complaints filed directly against it | None documented as used |
| Cloudbet | 3.7 | 3.0 | Curacao GCB | Domain-verified | None published | No published cap | 27.0% 10 paid / 27 rejected / 0 unresolved | Restricted-jurisdiction geo, undisclosed integrity hold, multi-account |
| 7Bit | 3.6 | 3.5 | Curacao GCB | Domain-verified | 0.16 BTC (about EUR 10,000) | 1.25 | 21.2% 21 paid / 77 rejected / 1 unresolved | Bonus max-bet, chargeback allegation, geo, KYC-process failure |
| TrustDice | 3.6 | 3.0 | Curacao CGA | Listed, could not re-confirm at primary | None published (about USD 1,000,000/month) | 0.05 | Full split not in corpus; 2 small unresolved | Advantage-play flag, multi-account, unenforced geo-block trap |
| Stake | 3.5 | 3.0 | Curacao GCB | Domain-verified | None published | No published cap | 15.4% 58 paid / 316 rejected / 2 unresolved | Undisclosed investigation, source-of-funds wall, geo and VPN |
| Roobet | 3.5 | 3.5 | Curacao CGA | Listed, could not re-confirm at primary | Not published (about USD 10,000/day reported) | 1.0 | Full split not in corpus | Group or syndicate betting clause, thin-evidence enforcement |
| Fairspin | 3.4 | 4.0 | Curacao CGA | Listed, not re-confirmed at primary | EUR 5,000/day above EUR 10,000 | 0.33 | Full split not in corpus; 1 relevant unresolved | Isolated only: alleged false-ID bans, one unresolved case |
| KatsuBet | 3.3 | 3.5 | Curacao GCB | Domain-verified | 0.14 BTC (about EUR 10,000) | 1.25 | 27.3% 6 paid / 16 rejected / 0 unresolved | Bonus max-bet, third-party payment, account block |
| Mirax | 3.3 | 3.5 | Curacao GCB | Domain-verified | EUR 10,000 | 1.25 | 23.8% 5 paid / 16 rejected / 0 unresolved | Discretionary low-risk-play clause, retroactive bonus-term swap, wrong-chain refusal, multi-account |
| TonyBet | 3.3 | 3.5 | Estonia EMTA (EU) | Domain-verified | EUR/USD 15,000 | 1.0 | Only 2 cases, 1 of them unresolved; too few to rate | KYC-at-withdrawal wall (UKGC-fined 2023), malfunction voiding, IP association |
| Weiss | 3.3 | 3.5 | Curacao CGA | Verified in register | EUR 4,000 | 2.9 | 51.9% 14 paid / 13 rejected / 0 unresolved | No upheld pattern; rejections trace to clear breaches |
| Golden Star | 3.2 | 3.5 | Estonia EMTA (EU) + Anjouan + Tobique | Domain-verified on three registers | EUR 20,000 (deposit-tiered) | 0.8 | Split not readable this pass | Interrogation-style KYC, max-bet, an unlocatable rule (reversed), bonus terms |
| mBit | 3.2 | 3.5 | Curacao GCB | Entity and number verified, indefinite term | 3 BTC | Excluded (BTC-denominated) | 38.3% 18 paid / 29 rejected / 0 unresolved | Multi-account, double-spending, third-party payment, manufactured bonus lock, max-bet |
| Vavada | 3.2 | 3.0 | Curacao GCB | Verified active | USD 5,000 (base tier) | 5.0 | 22.6% 19 paid / 65 rejected / 0 unresolved | Duplicate-account, bonus void on any withdrawal request, installment drip |
| Winz.io | 3.1 | 3.0 | Anjouan + Tobique | Both domain-verified | 100,000 USDT/USD | 0.12 | 26.6% 21 paid / 58 rejected / 0 unresolved | Retroactive geo change (casino-admitted), anti-fraud policy, bonus fine print, KYC unwind |
| King Billy | 3.0 | 3.0 | Anjouan + Tobique | Domain-verified | EUR 15,000 | 1.7 | Split not readable this pass | Multi-account, joint-account rule, undisclosed bonus expiry, game-bug void, geo change, stall plus 25% cancel bonus |
| Rainbet | 2.9 | 2.0 | Anjouan | Domain-verified | None published | No published cap | 4.2% 1 paid / 20 rejected / 3 unresolved | 72-hour KYC window, multi-account, VPN and geo accusations |
| Gamdom | 2.8 | 2.0 | Anjouan | Domain-verified; register expiry had passed while status still read valid | None published (up to 90-day manual review) | No published cap | 21.4% 18 paid / 65 rejected / 1 unresolved | Account-under-review ban mid-play, bare terms breach, VPN detection, multi-account |
| Betpanda | 2.6 | 2.0 | Off-licence | NOT FOUND standard met | EUR 15,000 (secondary) | 1.1 | 10.0% 6 paid / 51 rejected / 3 unresolved | Sportsbook advantage-play accusations with no evidence disclosed, discretionary KYC |
| CosmoBet | 2.6 | 2.0 | Curacao GCB (reported) | Could not check; review held as draft | EUR 7,500 (secondary) | 3.3 | Split not in corpus | Win-triggered internal review, terms-breach allegation after KYC passed |
| Cryptorino | 2.5 | 1.5 | Off-licence | NOT FOUND standard met | EUR 15,000 (secondary) | 1.1 | 0.0% 0 paid / 12 rejected / 1 unresolved | Risk-team discretion, multi-account accusation, KYC after a win |
| Winna | 2.4 | 1.5 | Tobique | Entity-only, nothing ties the licence to the domain | USD 10,000 per 24 hours (secondary) | 0.17 | Split not retrievable | Stall until the balance is re-gambled, retroactive term change, card-counting flag |
| Rollbit | 2.2 | 1.0 | Curacao GCB | Lapsed, "Assessment in progress"; entity-only | None published | No published cap | Excluded: the mediator records that the casino does not cooperate on any complaint | Linked-account, AML and provider flags with evidence never disclosed, geo and VPN |
| Duelbits | 2.1 | 1.0 | Curacao GCB | Lapsed, "Assessment in progress" | None published (open-ended manual review) | No published cap | 28.6% 20 paid / 47 rejected / 3 unresolved | KYC document rejection, bare terms-violation, self-affiliation, open-ended manual review |
| BC.Game | 1.9 | 1.0 | Anjouan | Domain-verified for bc.game only; mirrors not on the licensed list | None in primary (EUR 10,000/month secondary) | 5.0 to 10.0 | 21.5% 146 paid / 527 rejected / 6 unresolved | Advantage-play clause added after the ban, KYC at cashout, insolvency |
| FortuneJack | 1.9 | 1.0 | Anjouan | Domain-verified | None published (discretionary installments) | No published cap | 23.5% 12 paid / 39 rejected / 0 unresolved | Multi-account accusation at withdrawal, post-settlement void, discretionary KYC-confiscation clauses |
Individual reviews for every linked casino are in the casino review index. CosmoBet is not linked because its review is held as a draft: every figure in it is secondary and we have not validated the licence at the regulator.
Limitations
- Corpus size and selection bias. 28 casinos, 17 with full complaint splits. We pick which casinos to review and we pick newsworthy ones, so warning-worthy brands are over-represented by design. Read 21.3% as the rate inside this corpus, not as an industry rate.
- The dockets are platform-sourced and the platforms are commercial. Casino.Guru, AskGamblers, Trustpilot and LCB all monetise casino relationships, and Casino.Guru rules by its own internal codex rather than any law. Its "rejected" bucket mixes cases where the casino was right with cases the player abandoned, so player-win rates are floors.
- Counts drift and were captured on different days. Splits were read between 8 and 27 July 2026, through routes that occasionally mangled amounts. Where a snapshot conflicted with another we flagged the variance in the underlying research file rather than picking the flattering number.
- Some caps are secondary-sourced. Caps for Betpanda, Cryptorino, CosmoBet, Roobet, BC.Game and parts of Winna and Rainbet come from review-site reproductions of terms we could not reach directly, and are labelled secondary in the table. Primary-verified caps carry section numbers in the source research.
- Currency is approximated. Euro and dollar figures are treated at rough parity for the drip arithmetic. Bitcoin-denominated caps, at mBit in particular, are excluded from dollar medians because the corpus holds no conversion for them.
- Date range. The underlying cases span roughly 2016 to July 2026. Era effects are real: Golden Star's confiscation cases cluster in 2016 to 2018 while its 2021 to 2026 record is delay-then-pay. The aggregate figures are not weighted by date.
- Dockets measure disputes, not typical experience. Players who get paid rarely file anything. Everything on this page describes what happens when a withdrawal goes wrong, which is a different question from how often it goes wrong.
- The measurement surface is itself a target. Winna's record includes a Bitcointalk moderator admitting he was paid to delete scam-accusation threads and a withdrawn Trustpilot rating. Mirax and KatsuBet have documented private blog networks manufacturing instant-payout copy. Cloudbet ran a paid review campaign. Where reputation is bought, our inputs inherit the distortion even after adversarial reading.
- Nothing here was tested with money. We made no deposits and no withdrawals. Player-reported figures are labelled as reports, not as facts we verified.
Cite this study
Journalists, researchers and other affiliates are welcome to cite these figures with attribution to Rate Corner and a link to this page, under CC BY 4.0. The per-casino dataset is available as JSON. Suggested citation:
Rate Corner, "Crypto Casino Withdrawal Study 2026", 2026-07-30. 21.3% player-win rate across 1,958 closed Casino.Guru complaints at 17 reviewed crypto casinos. https://ratecorner.net/research/crypto-casino-withdrawal-study-2026/
If a casino's terms, caps or licence state have changed since publication, or you think a figure is wrong, tell us and we will check it and log the correction below.
Change log
- 2026-07-30 - Initial publication. 28 reviewed casinos, 17 with full Casino.Guru resolved/rejected/unresolved splits, 1,958 closed complaints. Supersedes the six-casino 20.6% figure published on 2026-07-12 in the withdrawal guide; that subset is embedded unchanged in this pool.