How the Draft Kit ranks for your league
A season-long NFL ranking is only useful if you know what it was built from. Here is every assumption behind the Draft Kit, including the ones it does not make.
Most rankings are a list of the best players. A draft board is a different question: who is worth the pick you are actually holding, in the league you are actually in. The TalliedPicks Draft Kit answers the second question, and this page explains how, so you can decide whether you agree with it before you trust it in a draft room.
Three scoring formats, ranked three times
PPR, Half-PPR and Standard are not one ranking with a filter over it. Each format is projected and ordered separately, because a reception is worth a full point, half a point or nothing at all depending on your league, and that single difference reorders receivers against running backs from the very first round. Every ranking is labelled with the projection it came from, so you always know which set of rules produced the number you are reading.
What goes into a projection
The projection is deliberately modest, and every assumption in it is a named constant in the code rather than a judgement call. Per-game fantasy points are weighted across the last three seasons, most-recent first, at 0.6 / 0.3 / 0.1. Fantasy points already contain usage multiplied by efficiency, which makes them an honest backbone for a first version.
A thin sample is then regressed toward the positional average with an eight-game pseudo-count, so three hot games in a small role do not arrive at the top of a board. Projected games played are recency-weighted and clamped between 10 and 17 - a transparent availability adjustment, not a durability model, and it is not pretending to know who gets hurt. Finally a team-context term nudges the number by at most six percent from the NFL model's preseason team ratings, clamped so that offensive environment can never overwhelm a player's own usage history.
There is no machine learning in this version. That is a choice, not a gap: an interpretable projection you can argue with beats an opaque one you cannot, and the model gets to earn a more complex second version by how its calls read against the season it described.
Value over replacement, not a flat list
The overall order is by value over replacement, not by raw projected points. A player's value is his projection minus the projection of the player ranked at replacement level for his position - the 14th quarterback, the 30th running back, the 36th receiver, the 14th tight end, in a 12-team starter-plus-flex league.
This is the whole reason a quarterback who scores more total points than a running back can still sit below him on the board. If the difference between the best quarterback and the fourteenth is small, the top quarterback is not buying you much over what you could have had later. Positional scarcity gets priced in by arithmetic rather than by an arbitrary cut line somebody drew between positions.
Tiers cut where the projections actually gap
Tier boundaries are not round numbers and they are not groups of five. Within a position, the projection curve is walked looking for gaps, and a gap wider than 1.5 times the median gap starts a new tier. The result is that a tier break means something specific: the next player down is measurably further away than the usual step between neighbours.
That matters most on the clock. If four players are in one tier, you can trade down or fill another position and still get one of them. If you are looking at the last name in a tier, the drop is real and the board is telling you so.
What it does not do
Rookies and players with no NFL usage history are not projected and do not appear in the ranking. The projection is built from realized usage; there is nothing to regress for a player who has not played, and inventing a number for him would make every ranking around him quietly wrong. A later version can add a draft-capital prior for rookies. This one says the omission out loud instead.
Kickers and defenses are projected from opportunity and accuracy priors rather than from the same usage backbone, and they carry more variance than any other position. Average draft position comes from Fantasy Football Calculator and player data from nflverse, both credited on the page itself.
Why a draft tool sits next to a graded betting record
The same discipline runs through both. Every betting pick this site publishes is settled in public against the closing line and the official box score - win, lose or push, with nothing deleted or quietly re-graded. A draft ranking cannot be graded the same week, but it can be built the same way: from named constants, with its limitations stated, by a model that does not get to revise its answer after the fact.
Draft rankings are publisher content, not advice. They are a statement of what the data shows, and they can be wrong. So can the projections behind them - which is exactly why the numbers that produced them are written down.
Open the Draft Kit · How every pick is graded · The public record