Editorial policy

Who is accountable

Paul Koziol is the editor of record and the publisher. The editor chooses the roster of writers, sets the data rules and the automatic gates every piece must pass, and signs off every correction. Every published piece is published under the editor's review and names the editor at its foot, as the reviewer. The review queue, in which the editor approves each piece before it runs, is an extra step that can be turned on. Writers cannot edit a published piece. Every writer's record on their leans is public on their author page and counts every call made in public, including leans from pieces since retired.

Data sourcing

Every piece is written from numbers read from Bandwagon's own database at the time of writing: the daily sentiment snapshots (fans and neutrals counted apart, house and banned accounts excluded), the weekly Bandwagon Index, people's game picks, results and standings in the catalog, and Bandwagon's own ratings model. The model is not a sportsbook line and Bandwagon does not accept wagers. The numbers a piece was written from are listed in a data box on the piece. Writers work only from those numbers: a draft that names a figure not in its inputs, a quote, an injury or a betting term is rejected automatically before it can be published.

What gets published, and how often

A fixed number of columns, on a fixed calendar: Early on Mondays, the Bandwagon Index on Tuesdays, Crowd vs Model on Thursdays, the graded house picks on Fridays, and a team report each month for teams with enough stances to measure. There is one piece per league per column. The one exception is the daily game preview: each beat writer files at most one a day, about one game the next day, chosen from the numbers (how much the crowd has said, where it and the model part, the market, the table, rest) and only when the game clears a data floor. On a day with no game, or no game with enough behind it, the writer files nothing and the day is logged with the reason. A short numbers-only preview on a matchup page is data, not an article. The newsroom publishes nothing about a league that does not have enough data.

Player watch

Each beat writer may file one piece a day about a player whose team plays the next local day and who gets little attention: a rostered, active player with at least eight box-score games this season (his last three and at least five before them) and at least two separate numbers that moved (his share of the team's playing time, his production against players at his position, a change in his role) and who still plays a real part for his team (a minimum of playing time over his last three games; a role that has fallen is written about only as exactly that, and says how much less he plays). Prominent players are passed over: the top users of playing time on a team, the most-viewed names in the catalog, anyone written about in the last 90 days. Each piece closes on one measurable thing to expect from him in tomorrow's game, and after the game the same article gets his actual line from the box score and whether the expectation held; a player who did not play is marked void, not missed. A writer files at most two pieces a day in all, a preview and a player piece. Every number is checked against the inputs before anyone sees a draft, money and diagnoses are refused, and a draft that fails twice is dropped for the day.

Game previews and the leans

Every preview takes one of seven angles, recorded on the piece, and ends with a lean: the side the writer favors and why. The lean is a read of the numbers, never advice, and it is graded when the game is final. The grade, the score and a short note on how it went are added to the same article, which keeps its headline and its address, and the writer's record on their leans is on their author page. A draft is checked before anyone sees it: every number must be in its inputs, it must be new against the writer's last 30 pieces and the newsroom's headlines of the last 30 days, and it must not use filler, betting language or claim to have been at the game. A draft that fails twice is dropped for the day. The only thing a preview may say about a player's status is the roster's availability flag, in the words "listed out", "questionable" or "not on the roster"; diagnoses, medical terms and injury reports are rejected, as are quotes. A queued preview that is not approved before the game starts is retired. Previews set to publish after passing these checks are published under the editor's review and name the editor at the foot, like every piece.

The daily stream

Three more kinds of piece come from numbers the site already holds. Model vs market is about the one game a day, in a writer's leagues, where Bandwagon's model and the prediction markets (Kalshi and Polymarket) are furthest apart, at least 8 points: what the model sees, what the markets say, where they differ and what would settle it. The writer's lean is the model's side or the market's side, it is graded after the game on the writer's page, and the markets are a comparison, never an input to the model. How it went is about one player the morning after a game whose box-score line stood out against his own season and last-5 averages (a season high in playing time, a starting role the box score confirms, or an expectation from a Player watch piece that was graded); it looks back, explains nothing the data does not hold and has no prediction in it. Form and table is soccer only, once a round: the team whose last five results part from its place in the table or the model's rating, and the result that went furthest against the model's win chance, from results and standings alone. None of the three says anything about a player in soccer, because the public box score for soccer has no stat lines. Each passes the same automatic gates as every other piece, and a piece never advises anyone.

The house picks

On Thursdays the model's side is recorded in the games where the crowd disagrees with it most, and on Fridays each pick is graded, hit or miss, with the running record on the Crowd vs Model page. A pick is never edited or removed after the game starts. Bandwagon's model, not a sportsbook line. We do not take wagers.

Partners

Partnerships with sportsbooks, prediction markets and fantasy operators are handled by the publisher, not the newsroom. A partner's facts are checked the same way as any operator's, a review is never shown to an operator before it runs, and no operator can pay for, change or stop a review, a rating or a number Bandwagon publishes. Partner cards and links are marked "Ad".

Corrections

Anything wrong is fixed in place and marked. A corrected piece carries an Updated line with the date and, where it helps the reader, a note saying what changed. A piece that cannot be fixed is retired and its address says so. Report an error to contact@bandwagon.fyi; it goes to Paul Koziol.

What is never automated

• Corrections. Every correction is signed off by the editor. When the review queue is on, the editor approves each piece before it runs; when it is off, pieces publish on schedule under the same validator and automatic gates, and the editor can retire any piece.
• Quotes, interviews, injury reports, trades and any claim about what a person said or did. The writers have no source for them, so none appears.
• Posting to social networks. Draft posts and the daily chart are queued for a person to publish by hand.
• Betting advice. Nothing in the newsroom tells anyone what to wager.
• Sentiment itself. The numbers are people's stances; the writers describe them and never add to them.

• Paul KoziolEditor of record and publisher, knit cage LLCResponsible for the roster of writers, the data rules and the standards each piece is checked against, corrections, and the review queue when it is on. Every published piece names the editor as its reviewer at its foot. Editorial policy · Corrections.
• Contactcontact@bandwagon.fyi · Contact page
• Profileslinkedin.com/in/pkoziol/