What it is
ADP (average draft position) is the average slot at which a player came off the board across a large pile of real or mock drafts. It is a record of behaviour, not an opinion — nobody decided it; it is the summed result of many people each deciding something.
That makes it a price. It answers "what does this player cost right now," and it is the only source in fantasy football that answers it. It does not answer "is this player good" or "how many points will he score" — those belong to a ranking and a projection, different objects produced by different processes. See ADP, rankings, and projections are three different things.
The distinction matters because a price can be right about what a player costs while being wrong about what he is worth, and both facts are useful to you at once, for different purposes.
What it is good for: timing
ADP's real job is scheduling, not evaluation.
You hold picks at known positions. At each one you face a question that has nothing to do with who is best: can I wait? If a player you want is priced later than your next pick, take someone else now and still get him. If he is priced before it, waiting means losing him. That is the whole calculation, and ADP is its only input.
This converts a draft from a sequence of isolated best-player choices into a plan. Two picks close together let you take the scarcer thing first and the more available thing second, and ADP tells you which is which. It also warns you when a whole position is about to go at once — several managers reaching the same conclusion inside three picks — turning a gentle price into a cliff for everyone downstream. Use it as a clock. It is a very good clock.
What it is bad for: being right
ADP is a consensus, and a consensus is wrong in exactly one way: everyone at once, in the same direction.
An average cannot be contrarian. If the drafting population is overrating a type of player — a class of role, a category of hype, a name everyone remembers from last season — the average records that overrating faithfully and presents it as information. The error does not wash out, because it is not random noise across drafters; it is a shared belief, and shared beliefs are exactly what an average preserves.
So ADP tells you what a player costs and nothing about whether that cost is sane. Treating it as a ranking imports every collective mistake in the hobby into your board. The corrective is to build your own valuation from something structural — projected points measured against what is free at that position, which is VOR — and keep the two numbers in separate columns.
It is platform-specific, and this is not a technicality
An ADP is produced by a particular population drafting under particular settings. Change either and the number changes.
Sites differ in default scoring, roster shape, user base, and — the one that matters most — the pre-loaded ranking sitting in front of every drafter as they pick. That suggested order is visible in the room, most people follow it loosely, and it feeds back into the ADP that platform generates. A site's ADP is partly a measurement of its own advice.
So an ADP from one site does not describe a draft happening on another. It is a real price in a different market. The broad ordering survives, but any claim of the form "he goes in the fifth round" is about the wrong room, and the disagreements are largest exactly where you care most: at positions whose value depends on scoring settings. When a board uses one site's ADP as a stand-in for another's, that substitution has to be labelled every time it is cited, and read as an ordering rather than a schedule.
The signal is the gap
Neither number is interesting alone. Your valuation tells you who you like with no sense of what he costs; the price tells you what things cost with no sense of what they are worth. Put them side by side and the useful quantity appears: the difference. A player your method rates well above his price is a place where you and the market disagree, and disagreement is the only thing that can produce an edge — if you are right. Everywhere the two numbers roughly agree, no decision is available: you and the room want the same thing at the same cost.
This reframes two words usually used as praise and insult.
A reach is paying above market. That is all it is. Not automatically bad — it is what you must do when the price is wrong and you know why. But it is justified only by private information, something you have computed that the market has not priced. Reaching without that is paying a premium for no reason.
A value is a player priced below your number. Also not automatically good. It can mean the market knows something you do not — an injury, a role change, a depth-chart move that has not reached your source. A large gap is a question, not an answer.
The honest caveat for a first real draft
The market is made by people who watch the sport constantly. They see role changes as they happen and have context you do not. Your gaps against that market will therefore mostly be your errors — a mispricing you think you have found is more likely a fact you are missing.
So a beginner's edge is not in finding mispricings. It is in not reaching: taking the structurally correct position at the structurally correct time, letting the price be the price, and never paying a premium for a conviction you cannot source. Discipline beats insight here, and discipline is available immediately. The insight is not.
One real exception. Gaps that come from your league's settings rather than from your opinion of a player are legitimate and findable, because a market priced under different rules is genuinely mispriced for your room. That is the one place to trust your own number over the market's.
Open questions
- How far two platforms' ADP actually diverge is unmeasured here — known to be nonzero and largest at scoring-sensitive positions, but the size is not established.
- Consensus bias is asserted directionally. Which categories of player the market systematically overrates changes year to year and this note cannot pin it down.
- That a beginner's gaps are mostly their own errors is a prior, not a measurement. It would take a season of logged disagreements to check.