RISK & RESEARCH
Auditable probabilities you can defend
Market-implied probabilities with provenance and methodology attached, for research that has to stand up to review.
Risk and research workflow
For risk and research teams: can a market-implied figure be traced to a suitable source, observation time and method before it enters a memo or report?
- Record the contract, source identity and source observation time where available.
- Check the quality state, relevant mapping and whether the current review supports the intended comparison.
- Attach the applicable method/version, assumptions and unresolved limitations to the research note.
Research output
A source-linked figure with its status and method context, plus an explicit record of assumptions that remain analyst-owned.
Limitations to carry forward
Cross-source consensus is conditional on two entitled sources, human-reviewed mapping and eligible observations. The separate portfolio-risk workspace is private beta; its analysis is not venue margin or a margin requirement.
Worked example
One figure as a memo would cite it, with everything that travels alongside.
Provenance
- Source
- POLYMARKET
- As of
- 2026-09-02 14:32 UTC
- Quality
- DELAYED
- Freshness policy
- STALE after 10 minutes
- Methodology version
- intel-methodology-r1@1+intel-consensus-v0.1.1
- Taxonomy version
- taxo_v1
When a computation changes it gets a new version identifier and a documented backfill. History is never silently rewritten.
What you get
A probability with no source, as-of time or stated method cannot be defended in a risk memo. PH Composite is implemented for qualifying reviewed inputs, while US and Kalshi production collection remains disabled.
| Capability | Status | Explore | Delivery |
|---|---|---|---|
| Source, as-of time and quality state on every figure | Available | Limited | Event workstation |
| Versioned methodology ids, with a documented backfill when one changes | Available | Limited | Event workstation |
| Rules-changed and confidence-change signals on your watchlist | Available | Not included | Alerts |
| The same records programmatically, for a model or a report | Available | Not included | Production API |
| Positions and collateral mapped into the canonical event graph | Private beta | Not included | Risk workspace |
| Terminal-state exposure, liquidity stress and PH Shadow Initial Margin (analytical, not venue margin) | Private beta | Not included | Risk workspace |
| PH Composite — implemented as gated, organization-scoped consensus | Implemented; eligibility conditional | — | Not delivered today |
Portfolio & risk
A second workspace on the same event graph: what you hold, what it is worth, and what happens to it under stress. Analytics only — Predict Here holds no collateral, sets no margin requirement and executes nothing. This workspace remains in the published private-beta profile. The risk model is version 0.1.0, and collateral cannot be observed automatically on the one collected venue — it has to be supplied. Reached through an Institutional Evaluation or Institutional plan; not included on Explore.
- Positions and collateral
- Position snapshots are mapped into the same canonical event graph the intelligence layer uses, and marked against the same published figures. Collateral is reported as observed: if any contributing account is unobserved, the total is left blank rather than partially summed.
- Event-state exposure
- Contractual loss across every terminal state of a canonical event, so deterministic offsets between related contracts net out instead of being counted twice.
- Stress and liquidity
- Probability shocks on a fixed grid, and a liquidity-stress figure walked from the same visible order-book depth the execution ladder uses. Where depth cannot support the position, the shortfall is stated rather than extrapolated.
- Shadow initial margin
- An indicative capital-at-risk estimate under the scenarios above, including concentration and near-resolution add-ons. It is Predict Here’s own model output — not a margin requirement, and not recognized by any venue, broker or clearing house. The near-resolution parameter is not yet calibrated, and every affected run is stamped with that warning.
Every run records the model version and a hash of the exact parameters behind it, so a figure can be reproduced under the rules that produced it. A run whose inputs are too stale to trust is recorded as such, with its capital figures left blank rather than estimated. This is analytics, not a clearing service: Predict Here sets no margin requirement, issues no margin call, holds no collateral and executes nothing.
Your evaluation
Three steps, and no commercial conversation until you have seen it against your own events.
Request access
Tell us your desk and your event set. We aim to reply within one business day.Scoping call
Thirty minutes on the events you actually cover and what a useful answer looks like.Institutional Evaluation
You evaluate Monitor against that event set before any commercial terms are discussed.
WORKING SAMPLE · SYNTHETIC DATA
Follow the research from observation to evidence.
Open Monitor, screen the universe, inspect contract terms, compare eligible outcomes and retain a research brief. Watchlists, saved views and adjustable workspaces keep the same event in context.
No account required for the sample. Its fictional records, local alerts and research stay separate from organization data. Portfolio scenarios are a bounded private beta; organization access still requires sign-in, plan and source permissions.
Put it in front of your team
Connect observations, terms, calendar context and a reproducible research note. Request an evaluation to review your organization’s source rights and workflow.
