Command center de fraud intelligence — Rakero Anomalies
Rakero · Anomalies

Fraud caught before the payout.

The fraud-intelligence layer that operates between the operator's transaction and the analyst's decision — in seconds, with a playbook, with an audit trail, fluent in Brazilian regulation.

Detection
Real time
Decision
Auditable
Compliance
COAF · LGPD · SPA/MF
Real time
1.247 tx/s
Anomaly
Withdrawal blocked
The real cost

Fraud in Brazilian iGaming is no
longer noise. It is a recurring P&L hit.

AML · Lavagem
R$ 4,4M
at risk · 1,929 players · 35% detection with current methods
Account Takeover
R$ 4,0M
1,184 compromised accounts · 30-day window
Payment Fraud
R$ 782k
9,084 alerts · chargeback + double loss
Match Fixing
Fed. crime
Law 12,299/10 · severe reputational risk
LIVE · figures observed at a single mid-size operator · 24h window
The gap

Everything that exists today
was built for another market.

Horizontal anti-fraud

Generic

Good at e-commerce. Has no idea what rollover, RTP, live-betting burst, sharp money are. Rewriting the rules takes months — and the fraudster has already moved on.

Enterprise AML

Expensive and batch

Runs at night, alerts in the morning, decides the day after. Doesn't know COAF, SISCOAF, Law 14,790/23. Hostile latency for iGaming.

In-house

Operator building it

Two analysts + SQL + chat. Works until the operation doubles. No audit trail, no network, no playbook. Whoever knows the rule leaves, and the rule goes with them.

There is no native Brazilian iGaming fraud intelligence. Until now.
Manifesto

Detecting is not enough.
Deciding in real time
is essential.

Anomalies decides between the operator's transaction event and the analyst's action, in seconds — with a playbook per rule, a signed audit trail and Brazilian regulatory grounding in every alert.

Engine
6 specialized agents
Decision
under 2 seconds
Workflow
Signal → Case → Decision
The engine

6 specialized agents.
One auditable ensemble.

Each agent carries its own rules, weights and temporal decay. The score is aggregated by weighted ensemble. No black box.

FINANCEIROActive

Laundering & smurfing

Deposit/withdrawal ratios, velocity, round amounts, proportional wagering, cross-country PIX.

F01 Saque despr.F03 VelocityF04 Round-trip+6
COMPORTAMENTALActive

Baseline deviation

Z-score vs historical profile, marathon session, post-dormancy burst, anomalous night pattern.

B04 MaratonaB06 BurstB10 Noturno+2
CASINOPhase 2

RTP exploit & bonus

Win ratio vs theoretical RTP, post-bonus game hopping, provider patterns, abnormal conversion.

C01 RTPC02 HoppingC06 Provider
SPORTSPhase 2

Match fixing & insider

Sharp money in obscure markets, late insider, live-betting burst, cross-book arbitrage.

S01 SharpS04 Live burstS06 Arbitragem
COMPLIANCEPhase 2

KYC, PEP & COAF

Incomplete KYC, PEP match, COAF threshold, self-exclusion and responsible gaming.

K01 KYCK02 PEPK04 COAF
NETWORKActive

Network & cluster

Clustering by IP, device fingerprint, email pattern and payment BIN. Farms and syndicates.

N01 IPN02 DeviceN03 BIN
Catalog

30 catalogued rules.
Each with a playbook.

In production
30
On roadmap
48
Avg FP rate
13.4%
F01
Saque despr.
30%
F02
Smurfing
25%
F03
Velocity
22%
F04
Round-trip
28%
F05
Estruturação
15%
F06
Valores red.
20%
F07
Z-Score
18%
F08
Spike depós.
8%
F09
Multi-Pay
15%
F10
Chargeback
22%
B01
Hor. anômalo
12%
B02
Pós-dormência
14%
B03
Maratona
15%
B04
Device switch
12%
B05
Geo jump
12%
B06
Bet pattern
16%
B07
Login pattern
8%
F11
PIX cross
17%
B08
Escalada
11%
B10
Noturno
13%
deterministic · weights per rule · temporal decay · ensemble aggregated per agent · scores 0.00–1.00 · configurable cutoffs
Patterns covered

Not a generic score. Six
risk patterns with grounding.

AML

Severe

Laundering via mismatched deposit/withdrawal, smurfing below the COAF threshold, cross-country PIX.

Lei 9.613/98 · COAF · SISCOAF

Account Takeover

Severe

Compromised credentials, post-dormancy burst, device-fingerprint switch, geolocation shift.

Attack window · < 1h

Payment Fraud

Critical

Stolen cards, fraudulent chargeback, multi-payment in 24h, blacklisted BIN.

Double loss avoided

Bonus Abuse

High

Multi-account farm, conversion spike, withdrawal right after wagering, post-bonus game hopping.

CAC + bonus margin

Match Fixing

Severe

Sharp money + late insider in obscure markets, correlated cross-house arbitrage.

Lei 12.299/10 · federal crime

RTP Exploit

High

Win ratio consistently above theoretical RTP, timing pattern (precise ms), statistical anomaly.

Provider · licensing
EVERY PATTERN →SuspendVerifyAuditReportDocument
Pipeline

Native streaming.
Decision in under 2 seconds.

01 · Ingestion

Native streaming

We consume events straight from your operation over an encrypted connection. No intermediary, no caching of sensitive data.

≈ 980 msg/s
02 · Routing

Agents

Each event goes to the right agent. Per-event, per-window and batch evaluation.

≈ 4.47 ev/s
03 · Scoring

Auditable ensemble

Signals per rule → agent score (weighted average × decay) → final score vs cutoffs.

lag p50: 11s
04 · Action

Alert + webhook

Persisted alert and signed webhook fired to your channels in real time.

≈ 183ms p50
Events / 24h
385.845
Signals generated
135k
Players analyzed
61.775
Critical open
19.270
Risk mitigated
R$ 2,58M
Análise de rede de fraude — Rakero Anomalies
Network analysis

Fraud is rarely solo.
80% of cases have a network.

Automatic clustering by shared attributes — IP, device fingerprint, email pattern, payment BIN. When you block one player, Anomalies already shows you the other 8 in the farm.

Clusters detected
10
5 IP · 5 device · 7d
Players in network
36
distinct · largest cluster: 9
LGPD Art. 6º VII · IDs masked by hash. Real metrics, identity revealed only with recorded authorization.
Brazilian compliance

Built for Brazil.
Not "adapted to it".

Every rule, every playbook and every audit log was written assuming COAF, LGPD and Brazilian iGaming regulation as the default context — not as an add-on.

§COAF · Lei 9.613/98

Smurfing detection calibrated to the COAF R$ 10k threshold. The AML playbook respects article 11: the report cannot tip off the suspect.

SISCOAF · 24hArt. 11 · tipping-offRetenção 5 anos

§LGPD · Lei 13.709/18

Synthetic identities (player hash) by default. Revealing identity requires recorded authorization. Metrics and network views are already masked in production.

Art. 6º VIIHash player_idAudit reveal

§SPA/MF · Lei 14.790/23

Network detection references SPA/MF Ordinances 1,143/24 (Art. 25) and 1,231/24 (Art. 33) directly. Regulatory grounding in every cluster.

Portaria 1.143/24Portaria 1.231/24Lei 14.790 Art. 36

§Match Fixing · Lei 12.299/10

Federal crime. The playbook includes a sports-integrity channel + criminal documentation protocol. Paying a confirmed bet = co-liability.

Reporte obrigatórioCanal de integridadeVoid T&C
Outcome

What Anomalies returns per month.

Risk mitigated / month
R$ 2,58M
withdrawals intercepted on severe/critical players · 24h window in production · ↗ 77% MoM · 61,775 players analyzed
SLA met
85%
last-hour alerts · dedup active
Avg resolution time
2.1h
vs. days in manual workflow
True positive rate
91.4%
analyst team average
Cases resolved / 7d
192
team of 5 analysts
ROIAt a mid-size operator, Anomalies pays for itself in under 30 days from blocking payment fraud + account takeover alone. AML and match fixing are regulatory protection — not quantifiable until it becomes a fine.
Frequently asked

Frequently asked questions.

How does detection work?
Six AI agents analyze every bet, transaction and session in real time and assign an auditable risk score per rule — you see exactly why each alert fired.
Does the alert arrive in time to act?+
Yes — analysis runs in real time over your event streaming, so fraud is flagged before the payout, not after the loss.
What about false positives?+
The score combines multiple signals (ensemble) with thresholds calibrated per operation, cutting noise and keeping the review queue lean.
Is it AML- and compliance-ready?+
Yes — a full audit trail, network analysis and reports ready for AML/CFT, COAF and SPA/MF.
The next decision

Every hour without detection is a
risk decision.

Anomalies is in production and the numbers on this page are real. The next step is to show yours — in shadow mode, over your real streaming, no noise.

Let's detect
Contact

Let's build the right game.

Fill out the form and our team gets back to you within 24h. Or talk now on WhatsApp.

WhatsApp
✓ Message sent! We will be in touch within 24h.