Specialist data
Turn permitted source material into records you can use and inspect. Dataset availability, coverage and licensing are discussed privately for each requirement.
Synthetic specimen · downloadable CSV
Inspect a price observation.
A field dictionary should show what a record means and what it cannot establish. A quoted price and size are not a confirmed fill.
| Field | Example | Meaning |
|---|---|---|
source_id | example-source | Provenance identifier; this source is synthetic. |
event_id | example-001 | Stable identifier for this event. |
market_key | two_outcome_match | Market definition, including the outcome set. |
selection | A | Outcome to which the quote applies. |
decimal_odds | 2.10 | Decimal odds at observation. |
available_stake_gbp | 100.00 | Quoted size; not a confirmed fill. |
source_time_utc | 2026-10-19T10:00:00Z | Timestamp supplied by the source; its event meaning must be defined. |
observed_at_utc | 2026-10-19T10:00:02Z | Time the collector observed the record. |
rule_version | example-v1 | Settlement-rule reference. |
quality_flag | valid_quote | Validation status under the stated checks. |
The CSV contains two synthetic rows for one event, two outcomes and one observation time. Dates are example values. It demonstrates a schema, not dataset availability, historical depth, a live feed or source access.
Download the synthetic CSVTwo kinds of data work
- Bespoke acquisition and processing
- Define permitted sources, reconcile records, document fields and produce the required schema with validation results.
- Dataset or licence enquiry
- Check whether the required records can be supplied, with the actual coverage, dates, rights and commercial basis stated in the proposal.
The public specimen is not a product catalogue. Game rules, poker hand histories, price observations and promotion terms have different provenance and validation requirements. Rights to one client’s source material do not transfer to another engagement.
Gaps, clocks and corrections
A missing row can mean no event occurred, a source was unavailable, collection failed or validation rejected it. Showing only accepted rows can make coverage look better than it is. A completeness percentage needs an expected observation count.
For this schema, checks would include unique observation keys across source, event, market, selection and observation time; decimal odds above 1.00; non-negative size; valid timestamps; known rules; and a record of rejected or delayed rows.
The example shows a two-second source-to-observer timestamp difference. Interpreting it as collection delay requires defined timestamp semantics and aligned clocks. The quoted price may have been older or already unavailable.
What a data specification covers
- Coverage and rights
- Date range, events, markets, provenance, permitted use, exclusions, known gaps and onward distribution.
- Delivery
- Field semantics, format, update cadence, delay definition, and snapshots versus changes.
- Quality and changes
- Expected and received observations, validation results, rejected rows, corrections, deletions and backfills.
- Commercial basis
- Processing project, dataset delivery or licence; scope, term and any exclusivity.