Network surveillance across every online poker app · game-integrity.com

Teams
within networks.

We map colluding rings, bot farms, and chip-dumping clusters — and the overlaps between them. Ban-ready evidence packets, delivered the same day, backed by cross-corroborated network graphs.

Targets Rings · Farms · Dumps
Scope Every major poker app
Bar ≥90% · ban on delivery
01

Three surface patterns.
One underlying operator.

Modern cheating is never two accounts. It is rings sheltering bots, farms funding cash-outs, and clusters that trace back into rings. We work the pattern behind the pattern.

Threat 01

Colluding rings

Three, five, sometimes a dozen accounts playing as one. We fingerprint the ring as a whole, not its loudest pair, and deliver the full membership.

  • Ring membership via co-play graph community detection
  • Soft-play, whipsaw, and information-flow patterns
  • Cross-table recurrence across multiple stake pools
Threat 02

Bot farms

Coordinated bot networks share infrastructure, schedules, and stat fingerprints. We cluster them, then isolate the winning nodes that fund the whole farm.

  • Device, GPU, canvas, and connection linkage
  • Scheduled-play and profit-target session shape
  • Cross-account stat stability far above human variance
Threat 03

Chip-dump clusters

Multiple accounts funnelling chips into one cash-out — a laundering typology regulators recognise. We reveal the entire cluster: sources, sink, and payment linkage.

  • Directed pot-level chip-transfer graph, rake removed
  • Source-to-sink funnel and cluster centrality
  • Payment, KYC, and device correlation across nodes
02

The evidence bar for a ban

Anonymised operating metrics from live surveillance. Every finding a club or union receives clears the same evidence bar — decisive, corroborated, and ready to action without waiting on an app operator's confirmation.

≥90% confidence to escalate

Minimum composite confidence before a network case is escalated for immediate action. Below that bar, cases stay under observation — never delivered as a soft flag.

3+ accounts per network

Cases are graphed at the ring or cluster level. A single suspicious pair alone is not a case — it is a starting node for community detection across the pool.

3 signal families / case

Behavioural, financial, and technical evidence must all agree before an escalation leaves the review floor. Two-of-three is under-observation only.

≤24 hours to packet

From the analyst pressing Scan to a signed evidence packet in a club's hands — a typical high-confidence network case closes inside a day.

100% app-agnostic

Our methodology is platform-independent. If the platform produces hand histories, transaction logs, and connection metadata, we can map its networks.

0 unilateral disclosures

Evidence goes to the club or union that engaged us. Nothing is broadcast, and nothing is delivered to a poker app operator unless our client asks us to.

The evidence packet clubs can act on.

Every high-confidence case ships as a signed PDF: network diagram, member roster, confidence and evidence scores per family, corroborating chip-transfer graph, device and connection linkage, and the retained qualifying game IDs behind the finding. Clubs and unions have everything they need to remove the ring the same day — without needing an app operator to sign off on the decision.

03

Overlapping networks
from the file cabinet

Three anonymised cases from one review cycle. Separate findings on the surface — one operator, three masks, underneath.

Case 01 · Colluding ring CS-0142 · redacted

Seven-seat ring with two embedded bot seats

Accounts 7 Case 02: 2
Confidence 96%
Action Full-ring ban
Pattern
Seven accounts rotating across six-handed tables with fold-to-partner rates three-plus standard deviations above baseline. Two turned out to be automated seats embedded to guarantee chip mobility — a pattern only visible once the ring was mapped whole.
Cross-case linkage
Two ring members share device fingerprints with the Case 02 farm; one is a chip source in Case 03.
Outcome
Full ring banned same-day — bots included, no separate investigation required.
Case 02 · Bot farm CS-0187 · redacted

Twelve-account farm sheltering ring seats

Accounts 12 Case 01: 2 · Case 03: 3
Confidence 92%
Action Winners banned
Pattern
Twelve accounts sharing devices, connection ranges, and scheduled play windows. Five nodes ran the winning strategy; seven fed them chips through disciplined losses at adjacent tables.
Cross-case linkage
Two farm accounts sit inside the Case 01 ring; three appear as sources in Case 03 — same operator, three ways.
Outcome
Winning nodes banned immediately; losing siblings held to prevent identity refresh, then swept together on cross-case corroboration.
Case 03 · Chip-dump cluster CS-0206 · redacted

Nine-source funnel fed by the farm

Accounts 10 Case 02: 3 · Case 01: 1
Confidence 97%
Action Ban + funds hold
Pattern
Nine sources funnelling net-negative chip flow through under-defended pots into a single cash-out. Seat-level, the sources looked unrelated. They converged on the same payment rails.
Cross-case linkage
Three sources overlap with the Case 02 farm; one is a Case 01 ring seat. The sink was new — a laundering identity for the other two networks.
Outcome
Cluster banned, funds held. Cross-case corroboration also unlocked action on the Case 02 under-observation siblings.
03 · b

The pattern behind the pattern

Three independent cases. One operator underneath — running a ring, a farm, and a cash-out sink against the same pool. Under-observation cases turn ban-ready the moment a second case corroborates through a different signal family.

CASE 01 · RING CASE 02 · FARM CASE 03 · CHIP-DUMP RING ∩ FARM 2 accounts RING ∩ DUMP 1 account FARM ∩ DUMP 3 accounts ALL THREE 1 account
29
Unique accounts across the three cases, after de-duplication
6
Cross-case members appeared in two or more networks
1
Meta-operator controlled all three networks — one payment-rail, one identity

Overlap converts observation into action. A device-only flag in one case clears the confidence bar the moment behaviour or chip flow corroborates it in another.

04

How the operator came apart

A twenty-one-day trace of the review cycle that surfaced all three cases. Each stage names what a single, deliberate analyst action produced — and what it revealed about the account underneath.

  1. Behavioural scan

    Colluding ring surfaces on the co-play graph

    A routine analyst scan on a client union’s mid-stake pool flagged a seven-account subgraph with fold-to-partner rates three-plus standard deviations above the baseline. Confidence closed at 88% — above the review floor, below the 90-percent ban bar. Case 01 opened.

    State CASE 01 · UNDER OBSERVATION
  2. Technical scan

    Bot farm collapses onto shared infrastructure

    A parallel device-and-connection scan on the same union produced a twelve-account cluster on shared GPU fingerprints, connection ranges, and profit-target session shapes. Behavioural stability across the cluster exceeded every human comparable in the pool. Case 02 opened at 92% confidence.

    State CASE 02 · BAN-READY
  3. First overlap

    Two ring seats collapse into the farm

    A graph join of Case 01 members against Case 02 device fingerprints returned two exact matches. Two accounts inside the ring were bot-controlled seats embedded to guarantee chip mobility. Case 01 confidence rose from 88% to 96% on cross-family corroboration — no fresh scan required.

    State RING ∩ FARM · 2 shared
  4. Financial scan

    Chip-transfer graph reveals the cash-out funnel

    A directed pot-level chip-flow scan across the pool surfaced a nine-source funnel converging on a single previously-unflagged cash-out identity. Source accounts shared no device links at the seat level, but the sink pattern was canonical. Case 03 opened at 91% confidence.

    State CASE 03 · BAN-READY
  5. Second overlap

    The farm funds the cash-out cluster

    Joining Case 03’s source list against Case 02’s roster returned three exact account matches; a further join against Case 01 returned one. Case 03 confidence rose to 97%. The three cases were no longer three cases — they were one operator running three networks against the same pool.

    State FARM ∩ DUMP · 3 shared · RING ∩ DUMP · 1 shared
  6. Delivery

    One packet. Three networks. Same day.

    A single evidence packet — network diagram, roster of 29 unique accounts, per-case confidence, chip-transfer graph, device-and-payment linkage — landed with the client. Every account, including the under-observation Case 02 siblings, was banned that afternoon and the cash-out sink’s funds were held.

    State 29 accounts banned · 1 meta-operator dismantled
05

From scan to signed packet

Analyst-driven, network-scale. Only cases that clear the confidence bar leave the review floor.

  1. 01

    Capture

    Hand histories, chip flow, timings, device and connection signals — from any major poker app.

  2. 02

    Scan

    Analyst configures window, shared-game floor, and evidence score, then executes. Deliberate, never autonomous.

  3. 03

    Cluster

    Community detection surfaces rings, farms, and funnels. Pairs are starting nodes; cases are graphed at network scale.

  4. 04

    Corroborate

    Behavioural, financial, and technical families scored independently. All three must agree.

  5. 05

    Deliver

    Signed packet — network diagram, roster, per-family scores, game IDs. Club bans the same day.

05 · b

The confidence model

Three independent signal families feed one composite score. Cases only leave the review floor when all three agree and the composite clears 90%.

A

Behavioural

  • Ring membership via co-play graph
  • Fold-to-partner and soft-play patterns
  • Whipsaw and squeeze recurrence
  • Positional VPIP flattening across a cluster
  • Timing correlation across seats
B

Financial

  • Directed chip-transfer graph, rake removed
  • Source-to-sink funnel shape
  • Cluster centrality and one-way persistence
  • Variance-adjusted anomaly on edge weights
  • Cash-out sink identification
C

Technical & identity

  • Device, GPU, canvas fingerprint linkage
  • IP and subnet co-location
  • Payment and KYC linkage
  • Behavioural velocity and scheduling
  • Emulator and multi-VM artefacts

Two-of-three agreement holds a case under observation. All three over the bar signs the packet. Cross-case corroboration counts as an additional family — under-observation flips to ban-ready without a fresh scan.

06

How we operate

A decisive posture built on high-confidence evidence, applied consistently across every poker app.

High confidence, decisive action

Cases only leave the review floor above the 90-percent confidence bar. Below it, we observe. Above it, the evidence is decisive enough to ban without waiting on an app operator's confirmation.

Network-scale, never a lone pair

A single suspicious pair is a starting node, not a case. We graph the whole ring, farm, or cluster and deliver the full membership list — cutting the network's edge, not one of its seats.

App-agnostic methodology

We work with whichever platform the club or union operates on. Our detection is built on hand histories, chip flow, and connection metadata — not on any single vendor's cooperation.

Client-owned evidence

Evidence packets go to the club or union that engaged us, and to no one else. Nothing is disclosed to a platform operator unless our client instructs us to.

07

Speak to the surveillance team

For clubs and unions considering an engagement. Tell us who you are and the scale you cover — we’ll come back with the next step.

Sent to security@game-integrity.com. We reply from a real human, not an automated inbox.

game-integrity.com

Network-scale evidence.
Same-day decisions.

Game Integrity Surveillance builds ban-ready evidence packets clubs and unions can act on without waiting — across every poker app in the ecosystem.