College football betting research methodology

Where the facts come from

Two kinds of document are used, and they are not treated as equals.

Primary documents are produced by the body that has authority over the fact: a state statute for a legal question, a conference's own release for a conference result, a university athletics department's game record for a score. They are the default, and every legal finding and every score on this site rests on one.

Secondary documents are reports about a primary fact rather than the fact itself — a news agency's poll story, for instance. Secondary documents are used for one purpose only: the national polls, which have no primary document of their own because the poll is the publisher's own count of ballots. A secondary document is never used to establish a legal question or a score.

Every page that makes factual claims carries its own source register: the documents it rests on, numbered, each labeled primary or secondary, with the publisher, the document's own date where the publisher states one, the date it was read and, for legal findings, the part of the document relied on. Those page-level registers are the evidence; nothing on a page depends on a document the page does not show. Where a page says nothing about a question, that is usually because no document in reach answered it.

How a game is checked

A result is recorded from the competing universities' own published accounts, not from an aggregator. Where both universities publish a record of the same game, both are read, and the score, the venue, the date and the home and away designation must agree before the game is written up.

That last item causes more errors than any other. A neutral-site game, or a game reported by the visiting team's own site, is easy to record backwards; four such reversals were corrected during this season's early reporting by going back to the athletics departments' own pages.

Where an official account and a widely repeated report disagree, the official account wins and the disagreement is worth mentioning in the article. Where no official account can be read, the claim is left out — including claims as ordinary as a team's current win-loss record.

How rankings are handled

The rankings desk reproduces published polls. It does not produce a ranking of its own, and it does not average or reconcile the polls it carries.

Three different things get called "the rankings" in college football, and they answer different questions. The Associated Press poll is a vote of media members. The US LBM Coaches Poll is a vote of participating Football Bowl Subdivision head coaches. The College Football Playoff Selection Committee's ranking is the one that bears on the postseason field, it is produced by a different group under different instructions, and at the time of this edition its first release of the 2026 season was scheduled for November 3. The site carries no committee ranking.

A poll is an opinion taken on a date. This publication treats a movement in a poll as evidence about voters, not as evidence about a team.

How state betting rules are handled

The state records are the part of this site where an error would matter most, so they are built differently from the articles. Each record is a set of individual findings — the overall status of sports wagering, the regulator, the minimum age, whether college events may be bet on, whether a state's own college teams are excluded, whether player props on college athletes are allowed — and each finding carries its own statute or regulator document and its own date.

Coverage is deliberately incremental. 2 jurisdictions are published so far: Kansas and New Jersey. A state is added when its rules have been read in the state's own law or regulator material, not when a summary of them has been found. A state that is absent from this site is absent because it has not been done yet, and that says nothing about its law.

Each finding is labeled with one of four statuses, and they are never collapsed into a yes or a no:

Permitted

The cited document establishes that the activity is allowed.

Restricted

It is allowed, but a stated limit applies — a category of wager, a class of team or event, or a condition.

Prohibited

The cited document establishes that the activity is not allowed.

Unknown

No document in reach settles the question. This is used, and it is not a synonym for prohibited: the absence of an answer here means the research has not produced one, never that something is forbidden.

Where a state's rules changed, the record shows the change as a dated sequence rather than overwriting the old position, so a reader can see what applied when.

What the review labels mean

A record also carries a review label. It describes the state of the checking, not the state of the law, and it is set for the record as a whole, not line by line: one review covers the record and carries the date it was done, the desk that owns it and the date it falls due again. All three are printed on the page.

Verified

The record was reviewed against the documents it cites, on the date shown. Each finding names not just the document but the part of it relied on — a constitutional paragraph, a statute subsection, a page — so a reader can go to the same place and check. This is the only label currently in use on the site.

Pending review

A conclusion is not yet settled, or the cited documents disagree with one another.

Outdated

The record was reviewed once, but enough has changed, or enough time has passed, that it needs reviewing again before it can be relied on.

One thing the label deliberately does not report is whether a cited document can be opened on the web today. That is a question about a link, and a link that stops answering does not undo a reading that was done — nor does a link that answers prove one. The two are tracked apart: the citation always names the section it rests on, and when a cited document moves or disappears the record goes back through review.

Findings are also marked according to whether the underlying question is settled, contested between documents, or unresolved. Everything published so far is settled and supported by the document cited; a contested or unresolved finding would say so on the page, not disappear from it.

Dates: what each one means

Three dates can appear on a page and they are not interchangeable.

The research cutoff is the point after which no new information was taken into a piece. It matters most on anything written to describe a moment — a preseason preview keeps its preseason cutoff and is not quietly updated with what later happened, because a preview rewritten with hindsight stops being a preview.

The publication date is when the piece first appeared. The review date is when it was last read through and confirmed. The preview and recap addresses each hold one dated edition — currently the Week 10 preview and the Week 9 recap — and show the dates of that edition. A new edition replaces the previous one at the same address. Weekly editions are produced by an automated process from the NCAA's own records and published only when every automatic check passes; such pages carry a system validation date rather than a reviewer's name.

Corrections

When a published claim turns out to be wrong, the claim is corrected and the correction is visible. On a state record that means the dated sequence gains an entry; nothing is taken out of it. On an article it means the text changes and the review date moves.

When a cited document is itself replaced or corrected by its publisher, the record keeps the relationship, and does not swap one link for another, so the trail from the current claim back to what was originally read stays intact.

Betting analysis, and where it stops

This publication writes about the football questions that sit underneath a betting market: what a team's record actually demonstrates, which parts of it were tested, what the schedule asks next, where a result was decided. That is the analysis on offer.

The scope is deliberate and worth stating once. Everything published here is analysis of football; a price, a line or a selection would be a claim about a market, and a claim about a market needs that market observed and timestamped at the moment of writing. Where an article discusses a spread or a total, it therefore treats it as a concept — the questions a number would rest on — rather than quoting one.

Images

The photographs on this site are generated images in a photographic style. Every one of them is marked "AI" on the image itself, says so in its caption, and repeats it in the description read by screen readers.

None of them documents a real game, a real venue, a real team or a real person, and none is presented as a photograph of an event. They illustrate a piece of writing; the facts come from the documents cited on the page, never from a picture.