American soccer acquires players through two entirely different mechanisms, and both end up in the same database. Merging them creates a record that looks unified and behaves inconsistently.
A draft pick is an allocation, not a purchase
Drafting assigns rights to a player without any negotiated price, because the order is set by standings rather than by willingness to pay. No market clearing happens.
The only price signal is the pick number, and that reflects a club's position in a queue as much as the player's estimated worth. Two different quantities share one field.
A model that treats pick order as a value proxy therefore learns something about league standings while believing it has learned about talent.
The two paths select different players
Drafted players are typically several years into college soccer, with a long record against a narrow standard of opposition. Transfer signings arrive from professional competition at varying ages.
Their prior playing histories are not comparable in workload, coaching or level, so a statistic carried into the database means different things depending on the path.
Pooling them and fitting one model produces coefficients that describe neither group, because the model is averaging across two populations with separate structures.
Age curves break in the middle
Valuation models lean heavily on age, since expected remaining career drives most of what a player is worth. That assumes a continuous development history.
College players enter professional soccer later than academy graduates in most of the world, having developed under a different calendar with a short competitive season.
Applying an age curve derived from academy pathways to a drafted player overstates how much development has already occurred, which is a systematic error rather than random noise.
Identity resolution is genuinely hard
The same person appears in college records, youth national team records and professional databases, often with different name spellings and no shared identifier.
Getting those records onto one player requires matching on birth date, position and club history, and every unresolved case silently splits one career into two partial ones.
Split records depress career totals and make players look less experienced than they are, which flows directly into any model reading those totals.
What separating them buys
Keeping the two sources apart until the point of comparison lets each carry its own assumptions, then converts them onto a common scale explicitly rather than implicitly.
The conversion is an estimate and should be labeled as one, but an acknowledged estimate is far better than an unnoticed assumption embedded in a merged table.
Analysts working in American soccer spend a surprising share of their time on this plumbing, and the ones who skip it produce confident numbers built on a contradiction.