Methodology

Two separate measures

Local-developed share is the share of all roster spots filled by players who meet the selected local definition and were not previously rostered elsewhere in the data. The second measure is the share of roster spots filled by players who appeared on another college roster in an earlier season. The site keeps these measures separate because a local player can also have been rostered elsewhere.

The three lenses

"Local" needs a boundary, and there's no single correct one, so the site computes every composition metric under three definitions of local:

• In-state: the player's home state matches the program's state.
• 250 miles: the player's hometown is within 250 miles of campus.
• 100 miles: the player's hometown is within 100 miles of campus.

Plain in-state and distance shares use players whose hometown can be classified under that lens as the denominator. Local-developed share and the share of players previously rostered elsewhere use the full listed roster. Reader-facing labels name the denominator when missing hometown data can affect the result.

The front-page national trend uses a fourth, combined home-region definition: a player is local when his listed home state matches the school or his listed hometown is within 250 straight-line miles of campus. It is a union, so a player who meets both tests is counted once. This keeps a nearby player across a state line and a same-state player in a large state from being treated as outsiders.

Team pages and rankings default to the in-state lens because it has the most consistent historical coverage. State lines remain an imperfect definition of local. A player just across a border may live closer to campus than a player on the far side of the school’s state.

The long national trend

The front-page line is a purely geographic measure. Transfer history is not comparable across the older official archives, so it is not part of this chart. A player counts as local when his listed home state matches the school or his listed hometown is within 250 straight-line miles of campus. A player who meets both tests counts once.

First, the site calculates a local share for every qualifying team-season. Then it takes the median of those team shares for the season. It does not pool players nationally, so a school that lists a larger roster cannot outweigh another school. The annual cohort is allowed to change as coverage grows. The underlying series begins in 1979, the first season with at least 10 qualifying teams. The front page currently displays 1986 onward.

A team-season qualifies only when it has at least 70 roster rows, a known home state for at least 75% of the roster, and enough information to classify at least 75% of the roster under the combined state-or-250-mile definition. The team local share uses classified players as its denominator.

Sources are selected for a whole team-season. A reliable CFBD roster is used first. If CFBD is missing or fails a reliability gate, a reliable official athletic-site roster is used instead. Otherwise the team-season is omitted. CFBD and official rows are never blended within one team-season. This fallback is especially important from 2004 through 2008, when some CFBD roster listings are incomplete.

Homecoming Radius and average distance

The Homecoming Radius is the median distance between a team's located roster hometowns and its campus. Half of the located players grew up inside that distance and half outside it. Team histories and the campus map use this median because it is less distorted by a few very distant hometowns.

The Closest to Home and Farthest From Home rankings for the 2026 roster use the arithmetic average distance among located players and label it as an average. The Most Backyard Players ranking uses the share of located players whose hometowns are within 100 miles of campus. These ranking measures are separate from the median Homecoming Radius.

Previously rostered elsewhere is a lower bound

A player counts when the same athlete ID appears on another college roster in an earlier season. This is alower bound, not a census. It can miss prior stops at programs outside the source archive, missing earlier rosters, and duplicated or reissued athlete IDs. The measure is useful for comparing programs and tracking a team's trend, but the percentage for one season remains an estimate rather than a complete transfer ledger.

Deep history is geography first

For 134 programs, official athletic-site archives extend the hometown record before CollegeFootballData's 2004 floor. Coverage is uneven by design: LSU is reliable back to 1981, while several programs reach the 1960s or earlier and others do not clear the reliability bar until much later. Team pages show each program's actual span instead of pretending every school has the same history.

Those archive rows have no stable cross-school athlete ID, so they do not receive a detected-transfer value or a local-developed classification. When an official roster has a genuine, sparsely populated “previous school” field, the site labels its share as a previous-school proxy. It remains separate from the previously-rostered measure everywhere. In overlapping seasons, CollegeFootballData remains canonical. Archive rows fill earlier or otherwise missing team-seasons.

Which seasons count

Early-era roster archives are thinner and less complete than recent ones. A team-season is treated asreliable and used in headline stats and rankings only when it has at least 70 roster rows and a known home state for at least 75% of them. In the current build, 136 of 140 programs clear that bar by 2016; each team page notes its own first reliable season.

The front-page historical line adds the 75% combined-classification gate described above. That extra test is chart-specific because the line uses the state-or-radius union rather than any single team-page lens.

Rankings use the 2026 roster. Where a change is shown, the comparison is against the 2016–18 average. The team-page comparison is different: it compares the 2016–18 baseline with the completed 2021–25 portal-era window. Teams without enough reliable history remain in current-roster rankings but show no change.

Data sources

The 2004-forward roster and earlier-roster series is built from CollegeFootballData.com, an independent, actively maintained college football data project. Earlier hometown history comes from official school athletic-site roster archives for 134 programs, with exact source URLs retained per player-season and hometowns geocoded against the U.S. Census Bureau place gazetteer.

This site expands on the original analysis in Andrew Dunn's essay, "How local is your team, really?", which covered a 25-program cohort using the public SportsDataverse ESPN roster archive and 247Sports recruiting pages. Credit to both SportsDataverse and 247Sports for the data behind that original published research. This site's 140-program build uses CFBD as the canonical modern source and official archives to extend each team's history where the source passes the reliability rules above.

Walk-ons are included

Roster composition counts every listed player, including scholarship players and walk-ons. It does not weight players by playing time, starts, or recruiting rank. Walk-on rates vary by program and walk-ons are more local on average, so two similar scholarship rosters can produce different local shares when their walk-on programs differ in size. This measures who appears on the listed roster, not who plays.

Why the numbers here differ slightly from the essay

CollegeFootballData roster snapshots are about 9% larger on average than the SportsDataverse ESPN archive used by the original essay. This was measured across the 25-program cohort from 2016 through 2025. The extra rows are disproportionately walk-ons, who tend to be local. As a result, this site's in-state shares usually read one or two points higher for the same team and season, while the shares previously rostered elsewhere usually read one or two points lower.

Across 250 matched team-seasons, the average difference is about +0.3 percentage points for home-state share and −0.9 points for the share of players previously rostered elsewhere. The original essay's headline findings still reproduce from this pipeline, but individual values can differ slightly because the roster sources contain different rows.

Other limits

• "Local" is based on a player's listed hometown, not birthplace or where they actually grew up.
• State borders aren't equally meaningful. A small, densely-populated state and a large, sparsely-populated one don't offer the same pool of in-state talent.
• This measures roster composition, not competitive outcomes. A rising or falling local share says nothing on its own about whether a team is getting better or worse.