QQPK Bans 724 Bot Accounts After Integrity Report

Four days. That is how long it took QQPK to go from "we will forward the information" to 724 permanent bans. On October 13, 2026, the Asian-facing room, which also trades as QQPoker, announced on X that its Security and Risk Control teams had confirmed 724 accounts "engaged in unauthorized automated play (bots) and severe violations of our Terms of Service," and that every one of them was banned with its balance frozen for good.

The trigger was a report posted to Two Plus Two on October 9, in which anonymous investigators flagged 2,587 accounts they believed were bots and claimed those accounts beat their all-in equity by 138,705 big blinds across 193,757 tournament all-ins. The bans confirm the room had a bot problem. They do not answer most of what the report raised, and a new, more technical allegation landed on the same forum on the same day QQPK made its announcement. We have kept our QQPK review live while this plays out. Here is where things stand.

What QQPK has said and done

QQPK's first replies were holding emails. Its customer service told the investigators on October 4 that "QQPK takes game fairness and related security matters very seriously," and on October 6 that the material had been "formally submitted to the QQPK Security Team." Things moved quickly once the report went public.

Forum accounts are easy to create, so on its own the October 10 post proves little about who wrote it. What gives it weight is that QQPK's verified X account followed through three days later. The post was specific: "All accounts in the report are under review for automated play, and their withdrawals are on hold pending review." It said QQPK holds complete records of every hand from July 2 to October 2, promised "We will publish our findings by October 17, including methodology," and added: "Confirmed bot accounts will be closed, and we will publish how affected players are handled."

The October 13 announcement delivered the closures. It did not deliver the second half. It says the 724 accounts' balances are "permanently restricted," but nothing about what happens to that money or to the players who lost it to them. That is the gap to watch on October 17. When GGPoker banned accounts for serious violations in 2025, it refunded harmed players from the confiscated funds. That is the standard QQPK will be measured against.

724 is not 2,587

QQPK banned 28% of the number the investigators flagged (724 / 2,587 = 27.99%), leaving 1,863 accounts unaccounted for. The announcement does not say whether the 724 came from the investigators' list at all. There are three ways to read the gap, and the October 17 report should tell us which is right.

One: QQPK's bar for a permanent ban is higher than the investigators' bar for "suspected," and more bans may follow. Two: QQPK found bots the investigators missed and cleared many they flagged. Three: the investigators' list swept up a lot of humans who simply play a tight, mechanical-looking style. The third would not surprise us. The report flagged accounts on playing patterns such as fixed round-number 3-bet sizes, used a "stat checker" whose rules were never published, and included a manual review by people who could see the results. That is exactly the kind of process that can lean, without anyone intending it, toward accounts that happened to run well.

The investigators raised a timing question of their own. QQPK's forum post went up at 05:30 UTC on October 10, saying every account in the report was under review, but the authors say they only sent QQPK the full account list at 12:37 UTC that day. "So which accounts had they actually placed under review and restricted withdrawals for when they published that response?" they asked. QQPK may have rebuilt the list from the published material. We do not know, and neither, publicly, does anyone else.

The bans don't answer the number that started this

Banning bots is one thing. Explaining how a group of accounts beat their all-in equity by a reported 14.6 standard deviations is another, and QQPK's announcement does not mention all-in results at all. That number needs care. It was calculated across all 2,587 flagged accounts, and it is very sensitive to how they were chosen. By our calculation, a list that leaned only slightly toward accounts that had run well, by an average of 0.29 SD per account, would produce the entire 14.6 with a perfectly fair deck (the group total grows with the square root of 2,587, about 50.9, and 14.6 / 50.9 = 0.29). With most of the list not banned, at least so far, that worry gets bigger, not smaller.

The fix is in QQPK's hands. It says it is re-running the all-in analysis on its own server records. The single most useful number it could publish on October 17 is the all-in result for the 724 confirmed bots alone. If those accounts still beat their equity by a mile, something beyond automated play was going on. If they ran close to expectation, the 14.6 was mostly a sampling artefact and the story is a bot ring that got caught.

If the edge does turn out to be real, the likeliest explanation is not a rigged deck but accounts seeing cards they should not. A tracker's all-in equity assumes every unseen card is still live. A bot that knows what its teammates folded knows better.

That one hand is worth 2.7 percentage points; the report's average edge when the flagged accounts were behind was 1.9. The report's own split fits the idea, too: the edge showed up at tables with three or more players seated and vanished at tables with only two, where there are no folded hands to know about. QQPK's statement mentions collusion only as something it has "zero-tolerance" for, not as a finding. Whether the banned accounts were sharing cards is another question for October 17.

A new claim: early deck keys and a dealing menu

On October 13, the same day as the bans, a poster called yegor, who carries a Verified Coach badge on Two Plus Two and has been a member since 2008, went further. He argued that "the main question here is not bot ring exposed but a question of games are fair," and described how QQPK's hand-verification system works: one backend endpoint hands out the encrypted, shuffled deck, another hands out the keys to decrypt it. His claim is that "in current implementation privileged users may receive secret keys before the hand ends," which would let them see the runout before it is dealt.

He also described a hidden menu in the client labelled with the Chinese term for a card dealer, which he translates as "manual deal," with one mode that sets the board cards and another that sets each player's hand. He says a live endpoint linked to it answered "unauthorized" when he tested it, and argued that if such a tool were ever used on real tables, "technically that does void the RNG certificate." QQPK's random number generator is certified by Gaming Laboratories International, whose certificate page confirms a pass on July 31, 2025. A certificate like that tests the shuffle. It does not rule out other ways of putting cards on a table.

Treat this with care. We have not tested any of it, and nobody else has published a replication. An "unauthorized" reply proves an endpoint exists and is locked, not who holds the key or whether it has ever been used. Game software often includes test tools for setting up specific deals during development; the real question is whether one is reachable in production, and by whom. There is also a cross-check against the original data: early deck keys would help just as much at a two-handed table, and the report found no edge there. As far as we can find, QQPK has not addressed the claim publicly. It should, on October 17.

Who checks QQPK's homework

Mostly QQPK. The room operates under a licence from Anjouan, an island in the Comoros off East Africa, and an ABC News investigation in December 2025 reported that the Comoros central bank has called several Anjouan licensing bodies "fictitious structures." Whatever you make of that, there is no strong regulator here to take a complaint to. In practice, the October 17 report is the accountability, which is why what goes into it matters.

Where we stand

Credit where it is due. QQPK put the flagged accounts' withdrawals on hold within a day of the report going public, banned 724 accounts within four, and put a date on a full explanation. That is a fast, public response. The original report remains unverified, and its headline number may not survive a closer look at how the accounts were chosen. On that basis our QQPK review stays live.

That is a position for now, not a clean bill of health. If the October 17 report slips, skips the refund question, or ignores the dealing-tool claim, we will say so here and revisit the review. If it delivers, we will say that too.

Sources: QQPK statements on X (@QQPKOFFICIAL, October 13, 2026); Two Plus Two investigation thread (October 9 to 13, 2026, including the QQPK Official, investigator and yegor posts); QQPK Game Integrity Investigation report; GLI certificate for the QQPK RNG; ABC News on Anjouan licences; PokerListings on GGPoker refunds. Ratios, the example hand and the selection-bias figure are our own calculations from the report's published numbers.