# 6. What would convince you that a role-fit or league-translation model works for our market?


**Interview objective:** Separate descriptive similarity from a validated prediction of performance after a move.

**Put to:** Sarah Rudd, Arielle Dror and the supplier’s modelling lead.

**Why it matters to buyers:** A cross-league modelling paper published on 26 August 2026 reports held-out transfer evaluation but explicitly limits its framework to Premier League destinations. That stated boundary is important: evidence for one destination does not itself establish applicability to MLS or NWSL. ([journals.sagepub.com](https://journals.sagepub.com/doi/10.1177/22150218261481583))

**A strong answer should cover:**

- The actual prediction target: style similarity, future minutes, specified actions, role suitability or another outcome.
- Training and evaluation populations, including leagues, seasons, roles and the treatment of women’s competitions.
- Testing on later, genuinely unseen cases rather than information available only after a transfer.
- Comparisons with a simpler baseline, uncertainty ranges and results for relevant subgroups.
- Low-minute players, unsuccessful moves, players who never move and other selection limitations.
- Conditions under which the model declines to rank or requests more evidence.

**Sharp follow-up:** Give us a documented failure and explain whether the limitation was data coverage, the predicted target, context change or the model itself.

**Request this artefact:** A model card and validation table matched to the club’s recruitment market. Where evidence is insufficient, label the output exploratory and use it to prioritise further assessment—not to imply a calibrated probability of success.

Proposed expert interview; not yet conducted. No expert quotation is attributed.