The Ruleset

The belt begins with college football's first game: Rutgers over Princeton, 6–4, on November 6, 1869, in New Brunswick, New Jersey. Rutgers is the belt's first holder.

Status: locked, pending one open check. The original collegefootballbelt.com site (dormant since 2018) may have stated its own origin point or house rules, and we'd like to confirm this site stays continuous with it via a Wayback Machine snapshot — but the Internet Archive was unreachable when we last checked (it was reporting a full outage, not a blocked request), so that comparison is still pending. Nothing here depends on the outcome; 1869 is the standard origin every existing lineal tracker uses.

The belt is won on the field. Whoever beats the current holder in a game takes it; every other result leaves it where it was. This one sentence is the entire engine — everything below is just how it resolves specific situations.

Before overtime rules existed, college football had plenty of tied games; Division I adopted overtime in 1996 and there has not been a tie since, so this rule only ever applies to the historical record. The rule: the holder retained the belt on a tie. A tie was not a loss, so it was not treated like one.

Why this over the alternative: this is the standard lineal-title convention (see boxing's lineal titles, or the "you have to beat the man" principle generally), and it's also the version that reproduces the reference trackers' published totals almost exactly — the other convention would move noticeably more distinct teams through the belt than any tracker reports holding it.

If the reigning holder has no game in a given week — a bye, an opted-out bowl, a canceled game — the belt simply carries over untouched. It only moves when the holder plays and loses. This already came up for real: Notre Dame won the belt on November 29, 2025, then missed the 2025 playoff and opted out of a bowl game, so the belt sat frozen through the entire offseason with no belt game until the 2026 opener.

That means a reign's length in days always includes the offseason: a team that holds the belt on New Year's Day gets 200-plus quiet days added to its reign before anyone can take it. That is the lineal convention and it stays, but the site also measures every reign in belt games — the game that won it plus every defense — and the Full History and conference pages let you switch between the two. A reader, Elliot, suggested the switch.

Seasons that were canceled or drastically shortened league-wide — 1918 (influenza pandemic), and the 1943–44 seasons for many programs during World War II — are treated the same way as any other stretch where the holder didn't play: the belt carries over. No special-case handling; it falls directly out of the rule above.

Some games are later stripped from the official record — a vacated win due to NCAA sanctions, for instance. This site follows the on-field result, not the retroactive record. The belt's entire premise is that it is won and lost on the field on a given day; a sanction imposed years later doesn't change what happened during the game, so it doesn't change who left with the belt.

If the holder loses to an FCS program, the belt goes with it, exactly as it would against any other opponent — the FBS/FCS line is a modern administrative label, not a rule of the game the belt is based on. The same went for the club, athletic-association and military service teams of the pre-modern era, and still does: before the 1978 split, every game on record counts.

From 1978 on, though, only a game between two Division I teams (FBS or FCS, as they were classified that season) can move the belt. This is a limit of the record, not a judgment about the football: the data source has no schedules below Division I before 2021, so a belt that went to a Division II, III or NAIA team could not be followed — it would simply vanish until that team next met a Division I opponent, years later. A Reddit reader found exactly that on the What if? page in September 2026 (a flipped Baylor–Wofford 2013 sent the belt to Division II UNC Pembroke, whose next game on record was in 2021). So a holder's loss to a lower-division team is not a belt game, and the holder keeps the belt. In the real lineage this has never come up — no holder has met a team below Division I since 1978 — so the rule changes nothing about the history; it governs the live seasons, the What if? page and the alternate universes.

The same logic runs in both directions. A club or service team that wins the belt holds it like anyone else: its later games are belt games, it can defend against other non-college teams, and it only loses the belt by losing on the field. This has happened: the Olympic Club of San Francisco took the belt from Saint Mary's in November 1931, and the San Diego Marines and West Coast Army both held it before Stanford won it on October 15, 1932.

Who suits up doesn't change this either. In November 1931 the Olympic Club split its squad, sending half to Hawaii while the other half played a scheduled game at Loyola (Los Angeles). That half lost, and the belt went to Loyola. An official game on a team's schedule counts no matter who played in it, the same way a college's loss with its starters injured still counts.

Football in the 1870s–1890s was often played by athletic-club and university-association teams that don't map cleanly onto today's program structure. This site includes any game that shows up in the historical record with a completed score, regardless of whether the participants would be recognized as "college teams" by today's boundaries — consistent with following on-field results rather than modern classification, per the rule above.

The full chain — every reign, every belt game, every score — is computed mechanically from the College Football Data API's game-by-game record, not researched or hand-entered. The only exception is the short, cited list of games that record is missing (next section), and those are walked through the same rules as everything else. Given the rules on this page and every final score in order, the result is fully determined; there is no editorial judgment applied game to game. Game dates reflect the local kickoff time at the game's venue, not a UTC timestamp, so a late West Coast night game is dated the evening it was actually played rather than the following morning UTC.

The College Football Data API's record is close to complete for college teams, but it has gaps in the pre-war era, mostly games involving club, service and small-college teams. When a gap changes who held the belt, the missing game is added by hand, and only under these conditions:

• The final score and date come from a contemporary source, normally a newspaper report from the day after the game, and that citation is shown on the game's page.
• The game is added exactly as played. Its result goes through the same rules as every other game, so every reign, defense and total after it is recomputed, not edited.
• Added games are kept in their own file, separate from the data source. If the data source later adds the same game, its copy is used instead.
• A game is added because it was played, never to make the lineage come out a particular way. Whatever follows from it follows.

The first correction came in September 2026 from Ray, who runs rutgersstartedthis.com, another lineal belt tracker. The data source was missing Loyola's 13–0 win over the Olympic Club on November 21, 1931, so this site had the Olympic Club holding the belt until September 1932. Adding that game, and the five other missing games the belt then passed through, added four reigns (Loyola, the San Diego Marines, Fresno State and West Coast Army) and shortened the Olympic Club's reign to two weeks. The lineage rejoins the previous chain when USC beat Stanford on October 22, 1932, so nothing after that date changed.

If you know of a missing game that affects the belt, email hello@collegefootballbelt.com with a source.

Alongside the real belt the site runs companion belts under the same rule, each restricted to a smaller world of games. They are derived the same way — mechanically, from the game record — with three additions the real belt never needs:

• Which games count is judged game by game, in the season the game was played. A conference belt counts a game only if both teams were members of that conference at the time. The FBS-only and FCS-only belts count a game only if both teams belonged to a conference of that subdivision that season, so a program that moves up or down takes its games with it. The subdivisions themselves date from 1978, when Division I split into I-A (now FBS) and I-AA (now FCS): the FBS belt is simply the real belt through the 1977 season and diverges only after that, and the FCS belt cannot start before 1978. In practice it starts in 2003, the first season the data source has any real coverage of I-AA games; it has almost none between 1978 and 2002.
• A holder that leaves vacates the belt; one that stays keeps it. Membership is read season by season from every game a team played, not just the ones that count, so an independent that goes years without meeting another independent is still an independent and simply keeps the belt until it meets one — the belt waits, as the real belt does. A holder that actually left — it changed conferences, moved between subdivisions, or stopped fielding a team while its old world went on — vacates the belt as of its last game as a member, and the belt reverts to the most recent earlier holder that is a member of the season it left for. The belt never rests with a team that is no longer part of its world.
• A belt whose world ends retires. When a conference dissolves there is no one left to inherit its belt, so the final reign closes on the date the league formally dissolved (or the day after its last game, when no formal date is known) and the belt is retired with that holder. A league that only changed its name keeps one continuous belt under its current name; one that stopped sponsoring football and later resumed starts a fresh lineage with its next game. A silence that crosses seasons the data source does not cover (I-AA play before 2003) is treated the same way — the belt retires on the last game we can see and a fresh lineage starts with the next one on record — rather than pretending nothing happened in between.

These rules replaced an earlier version in September 2026 after Elliot, a reader, pointed out that the FCS belt listed a run of one-day reigns, that North Dakota State had somehow never held it, and that defunct conferences showed their last holder as still reigning "to the present." He was right on every count; the last one was a genuine bug, and the others came from judging every game by each program's classification today rather than in the season it was played. Days later a Reddit reader caught the next flaw: the FBS Independents belt showed Notre Dame "leaving" twice and the belt retiring with UConn, when both are still independents — the site had been inferring "left" from "hasn't played another member lately." Membership is now read from the games themselves.

• Origin cross-check against the pre-2018 collegefootballbelt.com site, pending Internet Archive access.
• Two dates in the 1932 correction are uncertain by a day: the San Diego Marines' wins over the Santa Barbara Athletic Club (listed as September 12, possibly played the 11th) and West Coast Army (September 18, listed by one source as the 17th). Neither changes the lineage.
• Pre-1900 inclusion boundary: the current rule includes any game with a final score, which is the maximally inclusive reading. If the original site drew a narrower line, the cross-check above may prompt a revision.

The College Football Belt, the College Basketball Belt and the pro-league belts on Belt Holders run on the same idea and differ only where the sports do: ties, draws and shootouts, who counts as a league member, and what happens when a holder stops playing.