Super RTA Explained: Data Pipeline, Detection, and Enforcement

Super RTA—third-generation Real-Time Assistance—runs on a fixed architecture: capture reads the table, relay moves data off the play machine, processing returns a solver recommendation, and a human clicks the action. The poker client looks clean. The decision was made upstream, seconds ago, on hardware the room cannot see.

PokerNews, Pokercode, and GGPoker's Security Ecology Policy (SEP) draw the same line: any external aid that influences in-game decisions in real time is cheating.

This article maps the pipeline, dissects one famous exposure, summarizes how platforms detect RTA in 2026, and explains why outsourcing decisions—even when you keep the winnings—leaves you unable to think when the feed goes offline.

Three RTA Tiers (Not All RTA Is Super RTA)

Classify by how data enters the system, not marketing labels.

Tier	Input method	Processing	Player workload	Platform stance
Tier 1 — Static	Printed or on-screen charts (preflop opens, push/fold grids)	Human lookup	High—player maps spot to row	Banned during play on GGPoker (any chart); limited carve-out on PokerStars for basic unopened-pot starting-hand lists only
Tier 2 — Manual	Player types stack, pot, and actions into a solver	Local or cloud solve on demand	High—5–30 seconds per spot	Strictly prohibited while client is open
Tier 3 — Super RTA	OCR, screen capture, HDMI split, or memory scrape automatically	Cloud solver API or pre-computed tree lookup	Low—glance second screen or phone, click	Industry red line; permanent ban + confiscation on confirmed use

Tier 3 is the arms-race category. The human no longer interprets a chart—they execute a pre-selected action from an automated pipeline.

Public RTA services (documented on sites like rtapoker.com and rta.poker) advertise this stack: OCR reads multiple tables, a Telegram bot or overlay shows "Hero's next move," backends call CheckMath-style cloud trees or LLM approximations. Sellers claim "you still click" as if that matters. Platforms classify the information channel, not the finger.

Case Study: Kruse Dual-Machine Setup (2020)

The clearest public autopsy is the Fedor Kruse operation exposed on TwoPlusTwo in September 2020.

Setup (per roommate whistleblowers, reported by PokerNews):

• Machine A: Standard poker client—no prohibited processes, clean for local scans.
• Machine B: Pre-solved database ("Dream Machine")—PioSolver-derived trees browsable by board texture and action line.
• Protocol: Hard spot on Machine A → glance Machine B → replicate sizing and action on Machine A.

The exploit used physical separation: integrity software on the play machine saw a normal session while the solve lived on separate hardware.

Fallout (GGPoker public statement, September 2020):

• 40 accounts banned for RTA; 40 additional accounts warned.
• $1,175,305.43 confiscated from 13 of those accounts.
• 4,329 players reimbursed from the confiscated pool (Pokerfuse reporting on GGNetwork communications).

GGPoker never named Kruse in the official release, but timing aligned with the forum exposé. Kruse later claimed ~$250,000 confiscated in media interviews—a figure in the same order of magnitude as per-account averages from that wave, but not independently verified as his specific line item.

The case proved: you do not need a bot clicking for you. A human relay with solver-perfect upstream orders prints money—and erodes trust in every non-intuitive sizing that might otherwise be skill.

2025–2026: Hardware Capture and Counter-Detection

The pipeline did not retire after 2020. It hardened on both sides.

Cheat-side evolution (documented in industry reporting and vendor marketing):

• HDMI capture cards feeding a dedicated processing box.
• Phone cameras pointed at the monitor, OCR on a separate device.
• Cloud APIs returning moves within a few hundred milliseconds—vendors publicly claim sub-200ms reads for multi-table setups.

Platform counter-detection:

Layer	What it catches	Publicly documented by
Process / environment scan	Solvers, VMs, remote-desktop tools alongside client	GGPoker SEP §1.10; PokerStars prohibited-tools policy
Solver timestamp correlation	Training-site solve logs matched to live hand timing	GGPoker–GTO Wizard partnership (March 2025); 31 bans announced at launch
Behavioral signature review	Sizing mixes, timing uniformity, deviation from population norms	PokerStars cites >95% proactive RTA detection before player reports; human review follows flags
Live coaching / ghosting	Real-time advice over Discord, screen-share	Oct 2025 GGMillion$ case—RealOA permanently banned, Ren Lin suspended; $346,903 redistributed per GG statement

What is not on this list: arbitrary "98% GTO match" thresholds. Operators hunt patterns that track solver outputs across large samples, combined with environmental evidence. One solver-perfect river is variance. Ten thousand mixed frequencies with machine-like timing is a flag.

GeoComply and similar vendors market runtime integrity—flagging injected libraries, disguised processes, and overlay HUDs at the system level. Detection moved from "is Pio open?" to "does this session behave like a human decision-maker?"

Skill Erosion: Why Real-Time RTA Weakens You

Fair-play ethics aside, Super RTA creates a skill vacuum:

1. No range construction practice. You never defend why 67% pot is right—only that the screen said so.
2. No pressure rehearsal. Solvers do not feel tilt, clock anxiety, or the urge to hero-call.
3. No transfer. Kill the second monitor and the same player in a live tournament, a new pool, or a security interview exposes baseline ability immediately.

Pokercode and PokerCoaching draw the same line: study after the session, decide during it. Post-session review—hand histories, spot quizzes, deliberate re-solves with the client closed—is how knowledge migrates into memory. Real-time consultation skips that step entirely.

Compliant Alternative: Study With the Client Closed

The consensus across PokerStars, GGPoker, WPT Global, and major tours:

• Client open → zero external decision aids (charts beyond PS's narrow exception, solvers, coaching feeds, Telegram bots).
• Client closed → solvers, trainers, and review tools are permitted.

Structured post-session work—hand import, spot replay, drilling—is the compliant path. One session review beats a second screen every time, because the lesson sticks in your head instead of a relay buffer.

Summary

Super RTA is a four-stage pipeline: capture, relay, processing, human click. Platforms in 2026 invest in cutting it at every stage: environment scans, solve timestamps, behavioral review, and human case files.

The winning move for an honest player is slower and simpler: review after the session ends. Deconstruct hands, drill spots, build intuition. At the table, the only input that should reach your fingers is the one you generated yourself.