Version 1.0 / Published July 21, 2026 / Author: Expreci
How accurate is Expreci?
Expreci does not yet publish a quantified production accuracy, coverage, latency, or completeness result. Polymarket is the authoritative venue for its live trading state. Expreci is an independent analytics and historical reconstruction layer, and a reconstructed state is valid only when its starting snapshot, ordered inputs, gap state, and reconstruction engine version are known.
This means accuracy is not represented by a general claim. It must be measured over a named market set and time window using explicit sequence, reconciliation, and gap criteria. Expreci will not present intended architecture as measured production performance.
Authoritative source boundary
Expreci is independently developed and is not affiliated with, endorsed by, or operated by Polymarket. Polymarket remains the authoritative source for its markets, orders, trades, resolution rules, and live venue state. Expreci is designed to retain and transform venue-published observations for historical research and portfolio analysis.
If Expreci and the venue disagree about current state, users should rely on the venue. A historical Expreci result should identify its source records, capture times, gap status, and engine version so the disagreement can be investigated rather than concealed.
Data acquisition design
The intended acquisition design combines venue-published REST snapshots with WebSocket market updates where those interfaces provide the required evidence. Every retained record should keep the source payload or its content hash, source identifier, exchange timestamp when supplied, and local receive timestamp.
Source sequence is used only within the ordering scope guaranteed by the source. When no strict global sequence exists, Expreci must publish the applicable partition and deterministic tie breaker. Receive time can establish capture order, but it cannot create an exchange ordering guarantee that the source does not provide.
Production endpoint selection, collector regions, snapshot cadence, capture start dates, and supported market coverage have not yet been published. Those facts must accompany any future production coverage claim.
Reconstruction and reconciliation
- Begin from a validated snapshot with a known source scope.
- Normalize subsequent updates without discarding their source identity, timestamps, or raw evidence.
- Apply events in documented source sequence order and use named tie breakers only where required.
- Reject exact duplicates idempotently and retain conflicting duplicates for investigation.
- Stop declaring state valid when continuity or starting state cannot be established.
- Reconcile reconstructed state against later validated snapshots and record book state hashes and engine versions.
Identical inputs and the same engine version should produce the same result. Corrections create a new version and do not silently overwrite the interpretation used by an earlier result.
Gap policy
A gap includes a missing source sequence, a reconnect without established continuity, a missing valid starting snapshot, a conflicting duplicate, or another source integrity failure that prevents deterministic state transition. Affected intervals are flagged and are not filled forward as if the book remained valid.
Recovery can use retained source events or a later validated snapshot. The gap record remains part of history even after recovery and identifies the invalid interval and the evidence used to restore state. Results that depend on an unresolved interval must expose that limitation.
Current capability scope
| Surface | Status | Current boundary |
|---|---|---|
| Methodology, schemas, and interface design | Public now | Published for technical evaluation. Interfaces remain subject to revision. |
| Requirements intake | Public beta | Prospective users can submit specific data and analytics requirements. |
| Historical order book access | Planned direction | No production coverage dates or completeness levels are claimed. |
| Real time market data stream | Planned direction | No production latency, uptime, or service level is claimed. |
| Portfolio and scenario analytics | Beta design | Analytical scope is published. Production availability is not claimed. |
Latency, completeness, and coverage reporting
No aggregate or percentile production metrics are published yet. Before Expreci makes a measurable data quality claim, the report must define the population, time window, source interfaces, exclusions, clock method, engine version, and interpretation limit.
| Measure | Required publication detail |
|---|---|
| Coverage | Eligible markets and market hours, capture start date, exclusions, and reporting window. |
| Completeness | Observed sequence intervals, intervals without a detected gap, and unresolved intervals. |
| Latency | Receive latency distribution with median and tail percentiles, clock method, and sample scope. |
| Recovery | Gap count, affected duration, recovery source, and time to restored valid state. |
| Reconciliation | Snapshot comparisons, book state hash results, discrepancies, and correction versions. |
Known limits
- No production market coverage is claimed.
- No historical capture start date is claimed.
- No production latency, completeness, or uptime is claimed.
- An order book view does not establish executable value, fill priority, or future venue availability.
- Venue corrections, resolution changes, interface changes, and source outages can require a new data or reconstruction version.
Citation
Suggested citation: Expreci. “Polymarket Data Quality and Reconstruction.” Version 1.0, July 21, 2026. https://www.expreci.com/data-quality.
Continue with the reconstruction methodology, schema reference, or submit a data quality requirement.