Paragonix Primex applies backtested predictive models to structured and unstructured market data, producing recommendations that are logged publicly and open to independent review — not just presented in a private dashboard.
The distance between a terabyte of market data and a decision you can act on is where most tools fail. Our pipeline is built to close that gap in four auditable stages.
Structured feeds (pricing, volume, macro indicators) and unstructured sources (filings, sentiment data) are normalized into a common schema before any modeling begins.
Model parameters are tuned against historical regimes using heuristic search rather than static rules, so the system adapts when market conditions shift.
Every candidate model is run against multi-year historical windows it has not seen during training, isolating overfitting before it reaches production.
Approved signals are timestamped and published to the live performance log at the same moment they become available internally — no retroactive edits.
Every signal we publish is recorded with a timestamp and outcome, visible to anyone before and after the fact. This section reflects the same data available in the full log.
Paragonix Primex is designed to sit alongside a portfolio you already manage, not replace your judgment. The following are the three most common ways professionals integrate it.
Rather than concentrating exposure in familiar sectors, the module surfaces uncorrelated asset clusters identified through statistical distance measures across historical price movement.
Volatility clustering and drawdown patterns from comparable historical periods are used to flag positions where downside risk has increased faster than expected returns.
When a monitored data source shifts beyond a defined threshold — earnings surprises, volume spikes, sentiment reversals — a timestamped alert is generated and logged alongside the underlying trigger.
Public logs are only credible if someone other than the platform operator can confirm them. That is the role of our validator network.
Registered users with verified track records who independently review published signals and flag discrepancies before entries are marked confirmed.
Share of log entries where validator review matched the platform's original recorded outcome, based on the last four full quarters.
Read our full transparency policy for details on validator qualification, review cycles, and dispute handling.
The terminology on this page is precise on purpose. Here is what it means in practice, and where the boundaries of our claims sit.
A backtested model has been run against historical data it did not train on, showing how it would have performed. It does not guarantee identical future performance — markets change, and past accuracy is one input among several we monitor, not a promise.
No. Signals are timestamped at the moment of publication and cannot be altered retroactively. If an error is identified, it is corrected with a visible audit note rather than removed, which is why the log occasionally shows flagged corrections.
It is a method of tuning a model's internal settings by testing many variations against historical outcomes and keeping the ones that performed best, rather than relying on a single fixed formula. It allows the model to adapt as conditions change, within defined limits.
No. The platform generates signals, alerts, and risk flags for your review. Execution decisions remain with you or your existing broker relationship; we do not hold custody of funds or place orders on your behalf.
Validators are registered users who meet a track-record threshold and opt in to review published signals. Applications are reviewed against activity and accuracy criteria described in the transparency policy linked below.
Every claim on this page is drawn from the same log available to registered users. Creating an account gives you full-resolution access to historical signals, validator activity, and per-asset risk flags — no commitment required to browse the logs first.
Initialize AnalysisStandard access includes full performance log visibility and alert configuration. Extended data exports and API access are available on request; details are provided during setup.