The three objects
A projection is a forecast of points. Somebody predicts a stat line — so many carries, so many catches, so many yards, so many touchdowns — and those stats are then scored under a set of rules to produce a number. It is a quantity with a unit. Two projections can be subtracted, and the difference means something: 40 points over a season is about 2.4 points a week. Projections are the only one of the three you can do arithmetic with, which is why every value method, including Value over replacement in fantasy football, is built on them and not on the other two.
A ranking is one person's ordering. It says this player before that one, and nothing else. Crucially, a ranking is not simply a projection sorted — analysts routinely rank a player above someone they project fewer points for, because the ranking silently absorbs things the projection does not carry: injury risk, the chance the role changes, how wide the range of outcomes is. That is a feature, not sloppiness. But it means a rank has no distance in it. The gap between the 4th and 5th player might be enormous or nil, and the ranking cannot tell you which. Subtracting two ranks produces a number with no meaning.
An ADP is a market price. Average draft position is where players actually went, averaged across many completed drafts. Nobody's opinion — a record of behaviour. It is not a forecast and not a recommendation; it is the answer to "what will this cost me." Treated properly it is the most reliable of the three, because it describes something that already happened rather than something that might. See ADP as a market price.
The false friend from basketball
This is the distinction most likely to trip up someone arriving from fantasy basketball, because in basketball the three collapse into nearly one object. Basketball rankings are usually computed — projected per-game category lines run through a standardisation step to produce a value score, and the ranking is that score sorted. Ranking and projection are the same artifact viewed two ways, and disagreement between them is a bug.
In football they are separate artifacts made by different processes, and a site publishing both will happily contradict itself. That is not an error to resolve. The disagreement is the signal. When a site projects a player for a lot of points but ranks him below someone it projects fewer points for, it is telling you that the projection is fragile — usually because the player's role or health is uncertain. Read the gap, don't average it away.
Why ADP is platform-specific
An ADP describes a population of drafters operating under a specific rule set. Change either and the number changes:
- Scoring changes it. In a league that pays a full point per reception, a receiving-heavy running back is worth far more than in one that pays nothing for catches. Drafters in the two formats will not draft him at the same slot.
- Roster shape changes it. How many starters at each position, whether there is a flex, how deep the bench runs — all of it moves when positions get taken.
- League size changes it. A 10-team ADP and a 14-team ADP are different draws from different pools.
- The drafters themselves change it. Free public mock drafts are populated by people practising, often abandoning halfway, sometimes autodrafting. A real ESPN home league is populated by people who care, plus a few who did not open the app.
So an ADP taken from one site does not describe a draft on another. It is a related market, not the same market — closer to a comparable-sales figure from the next town over than to a quoted price.
The proxy in this build, stated plainly
The released ADP source is Fantasy Football Calculator’s 10-team PPR feed: 4,121 mocks dated 2026-07-26 through 2026-08-02. It is a labelled proxy for ESPN’s ADP, not ESPN’s own. The 2026-08-02 canonical freeze promotes 237 player rows; populated ADP values are sourced from ffcalc-ppr-10tm-2026-08-02, while intentionally unmatched source keys remain blank rather than inheriting July values.
Why that is defensible: it matches the confirmed 10-team, full-PPR market. It is the live frozen board input; the old July 12-team CSV is historical control material, not a relabelled current board.
Why it still is not the real thing: the drafter population is different, and free mocks skew toward people who follow fantasy football closely, which tends to push rookies and hyped breakout candidates earlier than a home league would. Every row of Data/players.csv carries an adp_source field recording exactly which pull it came from, so a later correction has a traceable blast radius. The honest fix is mock drafts on ESPN itself, which is one more reason those are on the critical path.
Anywhere ADP is cited in these notes, it should be read as approximately where the market has him, never as where he will go in this draft.
How to use each one
Projections build the value. Take projected stat lines, score them under your own league's rules, subtract the replacement baseline, and you have a board ordered by how much each player is actually worth to you. The released freeze uses FFToday’s complete 419-row 2026-07-30 native-stat snapshot. This is the only step that produces a defensible ordering, and it must use stat lines rather than an imported points total — otherwise you have imported someone else's scoring rules along with their numbers. The D/ST exception is deliberate: public FFToday D/ST values use incomplete brackets, so the validated modeled D/ST values remain canonical and the 20 raw source rows are audit-only.
Rankings sanity-check the board. Once the board exists, compare it to a consensus ranking. Ignore small disagreements. Chase the large ones: if the board has someone thirty slots away from consensus, there is a reason, and it is either something the board knows that the crowd does not, or a defect in the projection. Both are worth finding before the draft, not during it.
ADP times the pick. It tells you when, not who. In the requested 10-team snake, 1.01 is followed by 20/21; that 18-pick blind window makes current 10-team ADP essential. With two consecutive picks, take the player the market prices earlier first and the one it prices later second.
The compressed rule: build with projections, audit with rankings, spend with ADP.
Open questions
- How far does free-mock ADP drift from a real ESPN home league? Still unmeasured; ESPN mocks remain the direct test.
- The released Aug. 2 FFC snapshot must be rerun after the Aug. 30 cutdown and at draft −48h.
- Whether the board's disagreements with consensus are insight or defect has not been checked player by player. That audit is unrun.