The research desk
How our methodology actually works — without the marketing gloss.
Model architecture
The Aurevia Tradex platform runs three concurrent reinforcement learning models, each handling a specific stage of the trading lifecycle: signal generation (identifying opportunities), weight attribution (sizing positions given the current risk profile), and outcome feedback (continuously recalibrating against realised results).
Inputs
The platform ingests more than 14,000 data points per second covering consolidated order books from 35 regulated exchanges, interbank FX flows, scheduled and unscheduled macroeconomic announcements, and structured analysis of real-time corporate earnings transcripts. We do not use social-media sentiment, undocumented leak feeds, or any source whose provenance cannot be audited.
Limitations
No quantitative system anticipates every market regime. Low-correlation periods, sudden geopolitical disruptions, or unusually thin liquidity reduce the predictive value of historical patterns. Subscribers should expect periods of underperformance and structure their risk caps accordingly. Past performance is not indicative of future results. That is not a legal disclaimer — it is a statistical fact.
Editorial independence
The Research Desk does not coordinate with the Coverage team on what to publish or when. The principle is simple: if our research says the platform's current direction is wrong, we want our subscribers to learn it before we do. The editor retains the right to publish exactly what the Research Desk produces, without modification.
← Back to home