At first glance, it seemed that QQPK’s users had gone crazy for all-in bets. After users analyzed hand history data, however, it turned out that thousands of QQPK users made the exact same decisions and, overall, the group performed significantly better than indicated by probability theory.
Following this discovery, an important integrity investigation into QQPK was initiated in October. QQPK claims to have identified a card-sharers’ syndicate, prohibited 893 user accounts permanently and created a $300,000 compensation fund for damages suffered. In addition to these actions, QQPK issued an official apology for not having detected this operation sooner.
Therefore, QQPK has acted appropriately, but we are left with many unanswered questions concerning the reasons behind the extraordinary all-in rates.
Who was responsible for discovering the QQPK cheating scandal?
The primary investigation into the QQPK bot / cheating scandal was conducted by an anonymous poster named “QQPK_enjoyer” on the poker forum Two Plus Two. He collaborated with several other researchers who analyzed tournament performance data collected from QQPK from July 2nd to October 2nd, 2026 – in total, approximately 4.18 million hand histories containing around 26 million player-hand records across 27,936 tournaments. By examining the similarities within betting statistics, identical numbers of third-round bets and nearly identical limping-third bet sequences for 2,587 potential QQPK bot accounts, they identified those accounts as suspect.
In order to determine if the suspected cheaters engaged in something more than just good decision-making based on additional information, the investigators ran a separate analysis on the outcomes of all-in bets involving players outside of the suspected group. In total, the suspected cheaters engaged in 193,757 qualifying all-in bets against non-suspect players and generated an approximate +138,705 BB net gain over their respective all-in equity values. The investigators modeled this outcome as a 14.6-standard-deviation result under their models. A separate check on the flops and turns of 33,958 situations showed an additional +19,006 BB gain.
By definition, running at greater-than-EV (expected-value) rates during all-in situations is normal. However, given the massive sample size of qualifying all-in situations that occurred here, it called for a deeper investigation.
To assess their conclusions, the researchers compared their findings to those produced by two different hand-history analyzing programs and two different EV-engines. While QQPK’s researchers claimed that they did not intend to demonstrate conclusively that QQPK either cheated or assisted others to cheat via its dealt cards, QQPK acknowledges that it was aware of the evidence prior to posting of the research by the authors.
Prior to publication of the authors’ study, players in a Two Plus Two thread launched by koerck88 on September 20 complained of unusual results in cash-games. These complaints will remain separate from the tournament study findings and may not indicate involvement in the same scam.
Timeline of Investigation: October 10–17
Here is what happened and when.
October 10th: Account restrictions begin
QQPK stated it was investigating possible collusive activity and suspended account withdrawals. A dispute arose regarding timing: the researchers provided their entire list of 2,587 accounts at 12:37 UTC; QQPK responded that it began freezing certain accounts at 05:30 UTC. Later QQPK explained that its initial freeze included some of the accounts identified by its monitoring program prior to receiving the researchers’ full list.
October 11th: Technical analysis begins
QQPK states that its technical staff reviewed both the authors’ calculations against the server records for QQPK and evaluated if missing or duplicate hand histories contributed to the abnormally high figures presented in the authors’ research.
October 13th: Initial bans
QQPK announced it was banning 724 permanent accounts due to unauthorized automatic play and significant infractions of rules.
October 15th: Collusion confirmed by QQPK
In a detailed report, QQPK increased total number of banned accounts to 893. Out of those accounts, 567 belonged to the researchers’ original list; QQPK claims that its subsequent investigation identified an additional 326 through checking devices/networks/playing patterns.
QQPK asserts that none of the remaining 2,020 accounts showed direct evidence of cheating as defined by QQPK. Therefore, these accounts were subject to restriction and further examination — they were neither accused nor convicted.
October 17th: Reimbursement process begins
Instructions for processing claims provide details on challenges related to calculating compensation and specify a date (November 30th) by which players must file disputes.
How did QQPK cheating occur?
QQPK claims members of a coordinated group were seated at the same table and exchanged their hole-cards in real-time.
Picture yourself playing Texas Holdem with knowledge of exactly what two of your fellow opponents hold. Instead of having only your own hand information available to you, you now have six pieces of information. Your opponent’s hidden cards are irrelevant because you have access to information that will help you decide when to commit chips and when to fold.
Collusion is cheating. With a randomly shuffled deck that was un-biased in terms of cards, collusion is not fair.
QQPK claims the group utilized this knowledge to avoid poor all-in calls and select advantageous all-in situations. Using login information/login devices/betting behaviors/network tracking data, QQPK’s security team traced the behavior of the group.
However, there is another question. Making better decisions on all-ins is not equivalent to winning more than expected value already earned from all-ins entered previously.
When EV does not include cards being secretly known to a group of cheaters, it distorts expectations. How much this distorts explanations for observed results is still unclear. On October 19th, yegor posted on TwoPlusTwo asking QQPK to present recalculation data showing the impact of card-sharing rather than assuming that card-sharing provides all explanations.
Were QQPK’s RNG manipulated?
There is no indication of RNG tampering by QQPK. QQPK states that it has audited internally and has found no evidence of altered shuffling (dealing), unauthorized access or secret ways to reveal concealed cards. QQPK references its Gaming Laboratories International (GLI) RNG certification as supporting evidence of RNG integrity.
Additionally, QQPK defends itself with respect to dataset issues. QQPK notes that 4,866,164 raw records submitted from outside sources resulted in 1,757,070 unique hands after eliminating duplicates/hand IDs. Its larger review contained an estimated 22.18 million hands from approximately 24,026 tournaments. QQPK states that incomplete sampling and/or duplication of records biased outwardly extreme results relative to what actually happened on its site.
Furthermore, QQPK noted that it constructed one outside sample calculation resulting in a +192,063-BB surplus – approximately 14.6 standard deviations — versus the researchers’ reported +138,705-BB net gain. Although different hand counts and/or tournament totals exist between reports and have not yet been reconciled in published format.
QQPK claims that anomalous results were largely limited to banned offender accounts and that approximately 1.22 million hands (a subset) were generally consistent with expected results for remainder 2,020 monitored accounts. As stated previously however, QQPK’s published report does not provide a clear identification of a z-score for full-dataset resolving researchers’ central statistical issue.
Although QQPK’s audit and GLI certification are germane – they do not constitute independent public verification of each assertion made about these particular hands.
What about manual-dealing tool?
In addition to discovering collusive play behavior, players found an apparent test interface for manually selecting cards from the client software. QQPK states that this is legacy code from an early version of insurance tests — it did not function as a means to select cards during actual games. Furthermore, QQPK indicates that server-based test functionality had already been eliminated before October; therefore, any remnants existed only until technical modifications allowed viewing of the old interface.
As part of its transparency initiative, QQPK also described a verification mechanism whereby players can encrypt/shuffle/deal a deck before a hand begins and then decrypt/verify keys released after settlement has concluded; thereby allowing players to verify if received cards match those encrypted/shuffled/dealt prior to game-play commencement — it does not verify that initial shuffles were truly randomized or that keys could not possibly leak prematurely.
Before October 25th, QQPK promises to delete unused code from client software.
October 17: Compensation Fund Details
The compensation fund consists of monies obtained from two separate sources:
| Source | Amount |
|---|---|
| Monies seized from prohibited accounts | $64,627 |
| Contribution from QQPK | $235,373 |
| Compensation Fund (guaranteed) | $300,000 |
Given this context and considering that almost three-fourths ($235k / $300k) of funding originated from QQPK’s own contributions rather than monies forfeited by prohibited accounts — QQPK’s response represents a significant effort towards resolution.
According to QQPK’s October 17 compensation notice explaining procedures for filing claims including contact information for customer service representatives; providing player ID; submitting claims prior to midnight EST November 30th — QQPK retains final authority for determining eligibility for payment and amounts paid.
While establishment of a $300K compensation fund does not mean $300K has already been distributed to players — the reimbursement notice does not state otherwise; furthermore, it establishes no minimum number of recipients nor a minimum number of completed payments as factual measures — both should be assumed as speculative without factually documented reference.
Finally, in response to allegations surrounding automated players (bots), QQPK offered an Integrity Bounty to encourage independent reviewers to identify detection methods and requested permission from independent reviewers to analyze anonymized all-in hand histories for additional scrutiny.
Is there more to QQPK bot problem?
October 15th saw poker analyst Vojtech Cervinka publish a review titled “11,469 Nicknames Likely Bots” (MTT Database Review). Using an entirely separate database review focusing upon approximately 24M player-hands from July 16th thru October 11th; Cervinka estimated 17.5M of those represented hands played automatically.
Please note these are Cervinka’s estimates — NOT 11,469 confirmed bans in QQPK. Nickname does not always represent individual account. Additionally his sample should NOT be viewed as representative of ALL activity on platform.
QQPK merits recognition for taking ban actions; issuing apologies; contributing financially toward reimbursement efforts; and making available historical data for scrutiny. To answer questions posed herein regarding all-in statistics; provide fair reviews for additional flagged accounts; and confirm that applicable reimbursement funds reach intended recipient accounts; players have a practical deadline of November 30th, 2026.
Our key takeaways from the story:
- No online poker site in today’s market can guarantee safe heaven with preventative measures. But the important factor here is the reaction – fast timing and big compensation nearly 4-times bigger than the confiscated funds from the guilty accounts.
- The overall collusion and bot problem was found on the MTT tables
- Manual dealing-tool problem – QQPK addressed this in a long and descriptive manner on 2+2 and on X.
- Investigations into popular online poker sites are a good thing, and players advocate for data mining to be allowed by platforms, as a measure against bots/collusion.
- QQPK remains one of the most liked and preferred poker platforms, especially in Asia. Bigger incidents have happened to bigger platforms, and they continue to exist and even grow. The response here by the QQPoker team has been satisfying so far, taking responsibility and reimbursing the players, instead of denying that problems exist ( contrary to many other “regular” poker sites)
