How Valuation Works

Value is a range, not a single number — a low, mid, and high estimate of what a player's performance and tools are worth against your program's own configured budget benchmarks. It's a planning tool, not a promise: it is not an NIL offer, not a guarantee any player will sign for that figure, and not an appraisal any recruit, family, or coach should treat as authoritative on its own.

The range itself is deliberate, not a hedge. A player with thin data (one season, a couple of measurable events) gets a wide range, reflecting real uncertainty. A player with a deep, corroborated track record gets a tighter one. A narrow range on a thin profile would be a false confidence, not a better estimate.

Six kinds of input feed a Value estimate:

A .350 average at a small HS program and a .350 average in a power-conference D1 lineup don't mean the same thing. Stats and tools are adjusted based on the level they were produced at before being compared to your program's level — a translation, not a raw pass-through.

This adjustment is honestly imprecise in places. Some level-to-level relationships (for example, how a HS hitter's numbers typically translate toward college competition) are grounded in published transfer-performance research. Others — most notably how a JUCO player's stats translate to a four-year program — have no public research to draw on at all; no study we could find measures that shift at a population level. Where that's true, the model applies a general, clearly-flagged estimate rather than pretending to a precision the data doesn't support, and that shows up as wider uncertainty in the resulting range.

Makeup, coachability, and character are not in this model at all — they matter, and Sightline doesn't attempt to quantify them. Injury history and risk are only as good as what's entered; the model doesn't independently assess medical risk. Small samples are handled by widening uncertainty, not by pretending a short run of good or bad games means more than it does.

Several inputs — including a handful of advanced batted-ball-quality metrics — are currently priced against seed estimates rather than measured population data, because no public research exists yet to calibrate them precisely. Where that's the case, the underlying constant is documented internally and gets replaced the moment real data is available, rather than being presented as more certain than it is.

Every input carries its own uncertainty, and that uncertainty is carried through to the final range rather than averaged away. A thin stat line, a single measurable reading, and a level-translation estimate with no real research behind it all widen the range they touch. The result: a wide range is the model being honest about what it doesn't know, not a hedge to protect against being wrong.