# 14. How do we check whether a new player-data API will integrate with our database?


**Test identity resolution, data meaning and change handling—not merely whether an API request returns successfully.** Wyscout’s documentation exposes provider-specific player identifiers and profile fields; its API reference also documents resources such as players, matches and events. Treat those as inputs to a club-controlled identity system, not an assumed universal identifier. ([support.wyscout.com](https://support.wyscout.com/players-wyid?utm_source=openai))

Use this **Player Data Joinability Check**:

| Test | Evidence to retain |
|---|---|
| Identity | Accepted matches, rejected matches, ambiguous cases and reviewer decisions |
| History | Player-team relationships valid at the match date, not only the current team |
| Competition mapping | Gender, season, phase and competition identifiers |
| Semantics | Definitions, units, denominators and missing-value meanings |
| Geometry and time | Pitch coordinates, attacking direction, periods and timestamp alignment |
| Updates | Correction, deletion, backfill and schema-change behaviour |
| Operations | Authentication, rate handling, monitoring and support escalation |
| Exit | Usable authorised exports and retention/deletion instructions |

Maintain a crosswalk with the club ID, provider ID, evidence used to match, review status and mapping version. Avoid automatic name-only acceptance; route uncertain cases to review.

Report matching measures separately:

- **Automatic-match coverage:** automatically accepted records divided by eligible records.
- **Reviewed precision:** correct accepted matches divided by reviewed accepted matches.
- **Exception workload:** unresolved cases and time required to resolve them.

Do not describe a high match rate as high accuracy without reviewing false matches.

Version tests matter. Hudl’s 1 September 2026 Statsbomb release notes document fixes involving competition/season export selection and missing-data displays. Use that as a prompt to test exports against the screen and preserve versions—not as evidence that one provider is uniquely unreliable. ([hudl.com](https://www.hudl.com/releases/statsbomb))