Transfer risk screening
Anti-Failure Model
Flagging the signings that won’t work out - before the money is spent.
A screening model that identifies the 20–25% of transfers that historically fail to establish, at 80% precision on the highest-risk tier.
What it does
The model evaluates a prospective move against the conditions that historically predict a player failing to establish themselves at a new club - maturity signals, availability history, positional adaptation risk, and the stability of the destination. It returns a calibrated probability of failure, defined against a concrete, measurable outcome.
Used in: Player deep-dive risk assessment.
Why it’s defensible
What makes this proprietary.
A measured outcome, not an opinion
Failure is defined precisely - a player falling below a minutes threshold over an 18-month window, adjusted for injury - and the model is trained against verified outcomes for that exact definition. That rigor is what makes the 80% precision figure meaningful rather than rhetorical.
Trained on transitions, not snapshots
The signal comes from studying completed transfers end-to-end - how the move actually resolved - across hundreds of verified transitions. Assembling a clean, labelled transition dataset of that quality is itself a significant proprietary asset.
Risk that compounds with age
The model surfaces structural patterns the market systematically underprices - for example, that the youngest signings fail to establish at roughly twice the rate of players signed at 24 or older. Encoding those interactions reliably is where the value sits.
How we validate it
Flagged players are checked against observable post-transfer outcomes: minutes accumulated, league-tier movement, and whether the player established themselves at the destination club. When the model marks a transfer as high-risk, that cohort fails to establish at the predicted rate - verified on out-of-sample transitions the model was never trained on.
What we don’t publish
The risk factors, their interactions, and the labelled transition dataset are proprietary. We disclose the precision and the outcome definition so the screening can be trusted, without publishing the recipe that produces it.
The underlying data and models are proprietary. We show the validated results, not the inputs that produce them.
See it applied to your shortlist.
Every report shows the model’s output with the per-league accuracy behind it.
Order a Report