How Fanademy’s rankings are built

Why the board is ordered by value over replacement

Value over replacement is a player's projected points above the last startable player at his own position. Fanademy ranks by it rather than by raw projected points, because points alone are not comparable across positions: sorting by points puts almost every quarterback above almost every running back, which is arithmetically true and useless when you are setting a lineup or making a trade.

A running back 40 points above replacement helps your roster more than a quarterback 20 points above replacement, even though the quarterback scores far more in total. That is the whole reason the ordering looks the way it does.

Replacement level is not a fixed number. It moves with the league: team count, the legal lineup, whether quarterbacks start in one slot or two. This is why the paid tools recompute it for your league instead of applying one board to every format.

How the published accuracy is measured

Each published ranking is compared with what players actually scored, using the Spearman mean absolute error between each pre-game projection and the realised fantasy score.

The comparison uses pre-game projections only, so nothing is credited with information that was not available when it was published. Results cover a walk-forward backtest across the 2020 to 2025 seasons, and every subsequent change is versioned rather than replacing the record.

The full record, position by position, is on the accuracy page. It is published whether a given season reads well or badly.

What the published values mean

Projected fantasy points are the median projection for a player in a given week or over the rest of the season.

Dynasty values are a multi-year value-over-replacement measure, adjusted for age and shown on a 0-to-100 scale.

In the draft tools, the replacement line is recalculated as players leave the board and roster slots fill, so the board reflects the options that remain for that draft.

When the numbers change, and when they do not

Once a week. Rankings refresh on Tuesdays at about 09:00 UTC and nothing changes between refreshes — there is no intra-week update, no live update, and no in-game adjustment.

Each refresh republishes weeks 1 through 18 and the rest-of-season board together, in PPR, half-PPR and standard scoring. The published date and version are shown alongside the rankings, so you always know how old what you are reading is.

How players are identified

Every player is keyed on his canonical nflverse gsis id. Names are never used as identity keys, because two players share a name more often than people expect and a name-based join silently attributes one player's season to another.

Drafted rookies are ranked on explicitly labelled cold-start priors until they have enough NFL history for the weekly projection record.

What the rankings do not claim

A projection is an expectation, not a forecast of what will happen. A ranking that is right on average is still wrong about individual players every week, and no amount of testing changes that.

The published accuracy describes the seasons it was measured on. It is evidence about the past, offered so you can judge the product, and it is not a promise about the season you are playing.