Access Probex

QUBO
Probex

Real-time probabilistic consensus for predictive markets

Institutional-grade infrastructure built for compliant predictive markets — cryptographic finality in under 90ms, embedded across 47 regulatory frameworks.

99.42%

Consensus

< 90ms

Latency

840+

Markets

Access Probex

[ DATA PIPELINE ]

From Raw Signal
to Consensus Truth

Probex's four-stage processing architecture transforms noisy, multi-origin predictive feeds into cryptographically finalised consensus states in under 90 milliseconds end-to-end.

01INGESTION
02NORMALISATION
03CONSENSUS ENGINE
04COMPLIANCE GATE
01INGESTION

Raw Predictive Feed

Multi-source event streams from prediction markets, liquidity pools, and sentiment oracles are ingested in parallel. Each packet is timestamped at receipt, validated against schema checksums, and queued for normalisation — no data touches consensus without a clean signature.

Ingestion Latency

< 14ms

Sources Active

840+

Packet Integrity

99.97%

02NORMALISATION

Schema & Weight Calibration

Ingested streams are resolved against Probex's unified schema layer. Conflicting signals are probability-weighted using a rolling Bayesian calibration model. Stale events are expired on a per-source TTL; recency decay functions ensure the consensus state always reflects live market reality.

Normalisation

< 8ms

Calibration Cycle

50ms

Decay Precision

±0.03%

03CONSENSUS ENGINE

Probex Consensus Layer

Normalised signals enter the core Probex consensus engine — a distributed probabilistic voter that resolves multi-party event states with cryptographic finality. Quorum thresholds, deviation bands, and outlier rejection rules are enforced deterministically before any state update is committed.

Consensus Precision

99.42%

Quorum Latency

< 22ms

Finality

< 40ms

04COMPLIANCE GATE

Regulatory Output & Audit

Every committed state transition is routed through Probex's compliance guardrails before reaching end-consumer APIs. Jurisdiction rules, exposure caps, and mandatory reporting triggers are evaluated in-band. All outputs are written to an immutable audit log with full provenance tracing.

Gate Latency

< 6ms

Audit Coverage

100%

Jurisdictions

47

Live
End-to-end:< 90ms
Consensus P99:38ms
Events/sec:62,400
Uptime (30d):99.97%

[ PROBEX ARCHITECTURE ]

Three Operational
Layers. One Engine.

Probex separates concerns into three independently verifiable layers — consensus, compliance, and liquidity — each with its own failure domain, latency budget, and monitoring surface. They compose to produce a single, auditable predictive market infrastructure.

L1CONSENSUS VERIFICATION

Distributed Truth Resolution

Multi-node quorum validation with cryptographic finality. Probex resolves contested event states by aggregating validator votes against deviation bounds, rejecting outliers, and committing only when a configurable quorum threshold is met.

Quorum threshold

67%

Validator nodes

128

Finality

< 40ms

L2COMPLIANCE GUARDRAILS

In-Band Regulatory Layer

Jurisdiction rules, position caps, mandatory reporting, and KYC/AML hooks are evaluated inline — never as a post-process. Probex supports 47 regulatory frameworks with live rule updates and full immutable audit trail per state transition.

Frameworks

47

Gate latency

< 6ms

Audit coverage

100%

L3LIQUIDITY ENGINE
SPREAD

Predictive Market Depth

Real-time synthetic order book construction from aggregated position flows, automated market-maker liquidity, and institutional participation tiers. Spread optimisation and depth balancing run continuously to minimise slippage on large position changes.

Book depth

L5

AMM refresh

200ms

Slippage (avg)

0.04%

[ MATHEMATICAL FRAMEWORK ]

The Mathematics
Behind Probex

Probex's predictive accuracy rests on four interlocking mathematical frameworks — chosen for their formal guarantees, not their familiarity. Every layer can be independently audited against its specification.

A single deterministic core executes every framework below in lockstep — consensus, calibration, and attribution resolved on one clock, at institutional throughput.

01INFERENCE ENGINE

Hierarchical Bayesian Inference

Probex's core inference layer applies hierarchical Bayesian updating across all consensus inputs. Each new event updates a rolling prior distribution — signal confidence degrades gracefully under sparse data rather than collapsing to binary outcomes.

P(H|E) = P(E|H) · P(H) / P(E)

Posterior · Likelihood × Prior / Evidence

Specifications

Update frequency50ms rolling
Prior distributionBeta(α,β)
Convergence< 8 epochs

Consensus Precision

99.42%

Rolling 2,000 resolved events

[ PROBEX PLATFORM ]

Live Trading
Operations Console

The Probex operator console tracks BTC prediction-market positions, live edge and PnL, feed health, and adaptive risk state — the real trading surface, streamed in a single dense view.

probex // trading-opsPaper
FEED LIVE
00:00:00

BTC Price

$93,482

▲ live

Feed Latency

42ms

good

Active Positions

5

tracked

Edges Detected

1,284

rolling 24h

Orders Exec

318

avg 84ms

Total PnL

$338.00

unrealised

Open Positionsread-only
MarketDirEdgeProbabilityPnL
BTC above $95k at Fri closeYES6.2%
63%
$214.60
BTC below $88k this weekNO4.1%
74%
$78.20
24h realised vol above 4%YES5.4%
41%
-$63.40
Funding rate flips negativeYES3.3%
39%
$41.70
ETH/BTC below 0.052 at EODNO4.8%
69%
$66.90
Net unrealised$338.00
Survival StateHealthy

Capital preserved

96.8%

Kelly modifier

0.62×

Min edge

3.0%

Daily target

+1.4%

Runway

∞

Component Health
Price Feed
42ms
Main Loop
tick 500ms
API Access
reachable
Memory
61%
Warnings:0
Errors:0
Restarts:0
Uptime:18d 04h
PROBEX::OPS::v3.2.1

[ RESEARCH INTELLIGENCE ]

Research
Intelligence

Featured ResearchWhitepaper · 2025.11

Byzantine Fault Tolerance in Prediction Markets

Adapting BFT consensus to adversarial validators and conflicting event states at sub-40ms finality.

Read
Institutional InsightsAnalysis · 2025.09

In-Band Regulatory Logic Without Latency Penalty

Embedding jurisdiction rules at the execution layer with no measurable latency overhead.

Read
WhitepaperMethod · 2025.08

Hierarchical Priors for Multi-Source Signal Fusion

Constructing Bayesian priors that stay calibrated under sparse-data regimes.

Read
PublicationJournal · 2025.06

AMM Liquidity for Probabilistic Outcomes

Extending constant-function market makers to binary markets with asymmetric tail risk.

Read
Featured ResearchWhitepaper · 2025.05

Adaptive Kelly Sizing Under Capital Constraints

Survival-first position sizing that preserves capital through drawdown regimes.

Read
WhitepaperMethod · 2025.03

Brier-Calibrated Consensus for Event Resolution

Continuously scoring validator forecasts and down-weighting persistent miscalibration.

Read
Institutional InsightsNote · 2025.02

Edge Decay in High-Frequency Prediction Feeds

How informational edge erodes across the ingestion-to-execution path, and where to recover it.

Read
PublicationJournal · 2024.12

Survival-First Risk Models for Autonomous Trading

A formal treatment of capital-preservation states and recovery dynamics under loss.

Read
Carousel auto-scrolling

[ ACCESS THE PLATFORM ]

Ready to trade on
real-time consensus?

Probex is live. Institutional and professional counterparties can access the full platform today — live markets, real-time consensus feeds, and compliance tooling for 47 jurisdictions.

SOC 2 Type II

Certified

Consensus Precision

99.42%

End-to-End Latency

< 90ms

Access Probex Platform
probex-cli
CONNECTED
$probex auth --mode=institutional
✓ Authentication successful
$probex market list --active

PRX-001 Fed rate cut Q3 YES: 67.4%

PRX-002 BTC > $120k EOM YES: 44.1%

PRX-005 NVDA earnings beat YES: 72.9%

... 837 markets active

$probex consensus --stream PRX-001

Streaming consensus · latency 11ms · quorum 98/128

$
qubo-probex.duckdns.org/dashboard/OPEN