How the Draft Room works

There is no panel of experts behind this app. Every rating, grade and suggestion is worked out from public data by rules that fit on this page. This is all of them.

01What this is

A mock draft simulator. You take one seat, seven to thirteen computer opponents take the rest, and everyone drafts until the rounds run out. Drafting on your own, nothing leaves your browser and no server is involved. You can also send a link and draft with your league instead, which does use one — section 09 is how that works.

It was built for one ten-team league and still opens on those settings, but the league shape is now yours to set: 4 to 24 teams, 8 to 20 rounds, any starting lineup, and standard, half or full PPR scoring.

02Where the numbers come from

Two free public sources, no accounts and nothing to sign up for:

• Sleeper — the player list, injury designations, depth charts, season statistics back to 2018, week-by-week game logs and next-season projections.
• Fantasy Football Calculator — average draft position, taken from real drafts, as one set per scoring format.

Fantasy points are recomputed, not copied

Sleeper publishes its own points total, and it is thrown away. It assumes four-point passing touchdowns and flat kicker scoring, and bakes in choices you may not share. Every point you see is recalculated from raw components — attempts, yards, touchdowns, field goals by distance — under the rules below, which are the defaults and can all be changed:

Scoring format changes the board

Full PPR moves receivers up and running backs down, so each format has its own ADP list. When a draft starts, the board is built from the list matching your scoring, and every positional rank, tier and value figure is worked out from that list rather than from a fixed order.

Scoring happens as you watch

There is no points total stored anywhere. What is kept is the raw counting stats — yards, touchdowns, receptions, field goals by distance, points allowed — and every scoring rule is applied on the spot, as the page runs. All thirty-eight of them can be changed on the setup screen.

Change one and everything rescores at once: projections, every historical season, the week-by-week logs, replacement level, the Juke score and the draft grade. Nothing is rebuilt and nothing is refetched, because the numbers were never baked in to begin with.

This is why the daily rebuild records what happened rather than opinions about what it was worth. A stat that was never stored could never be rescored, so everything worth points to somebody is kept, even when the default rules score it at zero.

03The league, and everything that follows from it

Your league is described in one place, and every number on this page is worked out from it. These are the settings the app opens on, and the setup screen changes all of them:

That matters more than it looks. Nothing on this page is written down separately from those settings, so switching to twelve teams moves replacement level, the roster limits and the pick at which a kicker starts being worth taking along with it. Nothing is left behind describing your old league.

The setup screen refuses impossible leagues

Two combinations are rejected before a draft can start:

• The roster does not fill the rounds. Starters plus FLEX plus bench has to equal the number of rounds, or teams would finish the draft with seats unfilled or players with nowhere to sit.
• The draft is bigger than the room can use. Teams times rounds cannot exceed what the league is actually able to absorb — which is not the same as how many players are on the board. A ten-team league may hold ten quarterbacks, one each, and the board carries far more than ten; the spares are on it and draftable by nobody, so counting them as picks the room has space for would let a draft finish having spent its closing rounds on players who can never be started. The check counts, per position, the smaller of what the board has and what the room may legally hold. In practice it only has anything to say at the very deep end — the default ten- and twelve-team leagues run to twenty-two rounds before it objects, against the fourteen they actually use.

Replacement level

This is the most important number the app works out for you. It is the last player at a position who would realistically start somewhere in the league, and it feeds the draft grade, the Juke score and every value-over-replacement figure.

It is the number of starting slots the league creates at that position, plus that position's share of the FLEX slots, plus one:

The FLEX shares are how often each position actually wins that slot, which is why running backs and receivers run so much deeper than quarterbacks.

Roster limits

How many of a position a computer team will ever hold: its starters, its share of the FLEX, and a fixed depth allowance of 3 for quarterbacks, 5 for backs and receivers, and 2 for tight ends, kickers and defenses. At the default settings that works out to 4 quarterbacks, 8 backs, 8 receivers, 3 tight ends, 3 kickers and 3 defenses.

04How the CPU teams draft

Each computer team scores every available player and takes the lowest. The score starts at average draft position and is multiplied by how badly that team needs the position:

A lower score wins, so anything below 1 pulls a player up the list and anything above 1 pushes him down:

A team refuses a player outright when:

• it has hit the roster limit for that position;
• the player is ruled out — out, on injured reserve, suspended, non-football injury or did not report. These are never drafted by anyone and never suggested to you;
• it already has as many quarterbacks, kickers or defenses as the league starts. A superflex league that starts two quarterbacks gets two.

Kickers and defenses are priced, not scheduled

There used to be a rule here: no kicker before the last two rounds, no defense before the last three. It was tidy and it was wrong, and the board's own data said so — Seattle's defense carries an average draft position of about 81, which is round nine of a ten-team draft, while the room refused to look at a defense until round twelve.

Worse, a calendar has no variety in it. Measured across 60 simulated drafts at the default settings: the first defense went between picks 111 and 112 in every one, and the first kicker between 121 and 123, with defenses landing in only two or three different rounds. Real rooms do not look remotely like that — a one-pick spread is not a draft, it is a schedule.

So the two positions are priced the way every other position already is — what they cost against what else is on the board — plus one thing a shared rule cannot capture: ten managers do not all decide they need a defense on the same pick. Each computer team draws its own appetite for each of the two, fixed for the draft by the same seed that fixes the wobble, so one seat reaches, most take one about when the market does, and one waits it out.

The one thing the old rule did buy is kept. A team that has run out of picks to spare takes what it still has to start, so nobody finishes a draft with an empty kicker or defense slot.

The wobble

A draft where every computer team took the best available player would play out identically every time. So each player is nudged off his average draft position, by his own published standard deviation across the real drafts that average came from — which is why the very top of the board barely moves and the deep bench moves a lot. Decided once when the draft starts. It is settled rather than rolled fresh each pick: the same nudge holds all draft, so undoing a pick does not reshuffle how the room thinks, and a resumed draft behaves exactly like the one you left.

Picks are made one at a time on a three-quarter-second timer so you can watch the board fill in. Skip jumps straight to your turn.

Your suggestions answer to your scoring table, when it stops being a standard one

They are scored the same way as the computer teams', on your roster's needs rather than somebody else's. On a standard, half or full PPR table that is the whole of it, and deliberately so: average draft position is published one set per format, so the ADP behind those six names was drawn from drafts scored the way your league is scored. It has already priced your rules in. Discounting a player a second time for a projection built from those same rules is counting one fact twice, and measured over ten drafts it cost about five points of starting-lineup strength every time.

Edit the table, though, and ADP is describing a different game. Set receptions to five points and every number on a player's card moves, because the Juke score is worked out under your rules — while average draft position, a record of other people's drafts under theirs, does not budge. So once your scoring stops matching any of the three published formats, a player the projection rates highly gets a discount on his draft position: up to a quarter off, in proportion to his Juke score against the best player still on the board. In a five-point-a-reception league that is worth about four points of starting-lineup strength over advice that ignores the change.

05Tiers

A tier is a run of players at one position with no meaningful gap between them. The point is the sentence it lets the app say: not "take the best player available" but "take him now or lose the whole tier".

A new tier starts when the gap to the previous player is large relative to where you are on the board:

Proportional, because two picks apart means something at pick 5 and nothing at pick 120. A tier also breaks after six players regardless: gaps stop growing as fast as ADP does, so deep in the board nothing ever clears the threshold and a single tier would otherwise swallow half the position.

06The draft grade

Four parts. Each is worked out the same way for every team in the room, then scaled from 0 to 100 against the other teams, then weighted. Your grade is a ranking against the people you actually drafted with, not against the league at large — which is why somebody always gets the top grade and somebody always gets the bottom one.

The weighted total is ranked, and the ranks are handed letters from a fourteen-step scale running A+ down to F, long enough to give every seat in a fourteen-team room its own letter.

Why starter strength is measured against your draft slot

An early seat in a snake draft is worth a great deal on its own. Run a room where every team drafts by the identical rule, so nobody out-drafts anybody, and the first chair still fields a starting lineup worth about 190 projected points more than the fifth chair. That difference is real — but it is a fact about where you sat, not about how you drafted, and a grade that counted it was handing out most of a letter for the luck of the draw.

So your lineup is scored against par: the app simulates a room in which every seat simply takes the best player the consensus board offers, and reads off what your chair would have been worth in it. Your score is what you beat that by. Same idea as par on a golf hole — the number that matters is how far under it you went, not how long the hole was.

This is also why the figure under the bar reads two ways: the points your starters are worth over replacement, and that same figure against par. The first is your roster. The second is your drafting.

Why bye weeks are squared, and why every one counts

This used to read your worst bye week and stop. So three starters out in week 6 and three more out in week 8 scored exactly the same as a single bad week — the second one was invisible. Every week that leaves more than two starters out is charged now.

Squared, because the weeks are not interchangeable. Four starters out at once is a week you probably lose outright; three out twice is two weeks you can patch from the bench. So the fourth man out costs four times what the third does rather than twice, and one week with four off outweighs two weeks with three.

What "cover" means in roster construction

If one of your starters goes down, how far from startable is the next man up? That is the best player on your bench at that position, measured in places past replacement level. Nothing is docked if he could start today; the full 12 if he is fifteen or more places past it, or if there is nobody there at all.

Why kickers and defenses sit out of draft value

Not because the room picks the moment — it used to, and does not any more. Because of what their average draft position is measuring.

It comes from drafts that run more rounds than most leagues set up here, so a kicker's place on the board is often past the last pick that exists in your draft — which means taking one at all reads as taking him early. Measured across a full ten-team, fourteen round draft: all ten kickers scored as reaches, by an average of 35 picks, and not one came out neutral. "Biggest reach" had become a lottery among kickers.

So both positions are left out of draft value and out of the two callouts. It is not a thumb on the scale — every team drafts the same forced pair, and recomputing a full room without them moved no team more than two places. It takes out a constant that was drowning the thing the number is supposed to measure.

07The three signals on a player

Open any player and there are three meters. They are the app's own reading, not a consensus of analysts, and they disagree with the market on purpose — that disagreement is most of what they measure.

The Juke score

Projected points above a replacement starter at that position, expressed against the best such figure on the board. A player projected for exactly replacement production scores zero regardless of name.

It is a ranking against the rest of the pool, not a rating of the player, and two things follow from that which surprise people. Somebody always scores 100. And most of the board scores 0, because the board is about 220 deep and a ten-team league only ever starts ninety players at once — so a receiver one point below replacement prints what a receiver sixty points below prints. The player sheet says how far below, because those are not the same player.

The sheet prints last season's Juke score beside this season's for the same reason. Where the two disagree is where the projection is actually taking a position — and it is worth knowing how big those moves get. Across real consecutive seasons the score moves by 12 to 20 points on average, and of the players who scored above zero one year, between a quarter and two fifths were back at zero the next. Treat it as a strong opinion about ranking, not a measurement.

Upside

Starts at 20 and rises with: projecting at least four positional ranks better than his draft position (up to +35), three years or less in the league (+18), first on his depth chart (+14), age 24 or under (+10), and points per game up more than 20% across his last two full seasons (+15).

Bust risk

Starts at 12 and rises with: being drafted at least four ranks ahead of where he projects (up to +35), being ruled out (+40) or carrying a lesser injury designation (+22 or +12), age (+20 for a back at 28 or older, +15 for a receiver or tight end at 31, +12 for a quarterback at 36), sitting second or lower on the depth chart (+15), playing 12 games or fewer in his last active season (+14), and having missed real time in at least half of his seasons (+12).

08Saving and resuming

A draft on your own is stored in your own browser: your seat, the clock length, the wobble, the league settings and an ordered list of player names. Refreshing the page no longer destroys a draft.

Signed in, that same save is also copied to your account, and so is the locker of drafts you have finished — which is the whole of what an account is for here. The browser's copy is still written first and is still the one the page reads; the copy on the server exists so the next browser you sign in on can read it too. Signed out, there is no second copy and nowhere for one to go.

A draft with your league works differently, because everyone in it has to agree on what has happened. That one is held on a server while the room is open — the seats, the picks, the names people typed and the chat. A room needs no account and is never told your real name or email address, only the display name you type into it. Section 09 is all of it.

It refuses to restore rather than restore something broken, in two cases:

• The settings no longer match. Resuming a twelve-team draft into a ten-team league would corrupt the board — the snake turns in different places and the rounds run out at a different pick. The saved draft is kept and the app tells you which settings it needs, because setting the dropdowns back is all it takes.
• A player no longer exists. The player list is rebuilt every morning. If a saved pick cannot be resolved the board would be corrupt, so the draft is discarded rather than half-restored.

09Drafting with your league

Everything above describes a draft against the computer. You can also create a room, send the link, and draft with real people instead. Whoever opens it takes a seat, and any chair still empty when you start is drafted by the computer — so six of you can still run a ten-team draft.

The link

A room code is eight characters, and they are picked carefully: no vowels, so a code can never accidentally spell a word at somebody, and none of the characters that argue with each other on a screen — no O against 0, no I against l against 1. Short enough to read down the phone, long enough that guessing one is nobody's idea of an afternoon.

Seats

Seats are taken in the lobby, first come, and you can move chairs right up until the draft starts. After that the room is set.

If you drop out mid-draft — a tunnel, a locked phone, a laptop shut too fast — your seat is held and picked for until you are back, never handed to somebody else. Losing your roster because your train went underground would be far worse than having a pick made for you.

The clock belongs to the room

The room does not count down. It records when the current pick started and works out how much is left whenever anyone asks. That is what lets someone who just joined, or whose phone was asleep in a pocket, see the same number as everybody else instead of counting from whenever they woke up.

When it runs out, the pick is made for that seat the same way it is for an empty chair.

One board, agreed by everyone

Drafting alone, your own browser decides what is allowed, and there is nobody to disagree with. With ten people the room decides instead, so two managers hitting Draft on the same player half a second apart cannot both come away with him. Your browser sends what you want, and the board moves when the room says it moved.

The player list is rebuilt every morning, so a room also pins the version it opened on. Two people on different mornings' lists would watch the computer teams pick differently and drift apart inside a round.

Chat

Every room has a chat, with reactions and GIFs, and the picks appear in it as they happen. Three things about it are worth saying plainly, because chat is the only place in the app where the words come from another person rather than from the data:

• Anything typed is treated as words and never as instructions to the page, so nobody can slip something live into everybody else's draft.
• GIFs are only ever fetched from GIPHY, checked properly rather than by looking for "giphy" somewhere in the address — those are not the same test, and the difference is somebody else's website.
• Names are tidied up and capped at twenty characters, and the "typing…" line is never stored anywhere. It is true for about two seconds and a lie the moment a connection drops.

There is a ceiling of forty actions every ten seconds per person, which is far above anything a real draft reaches and far below what a script would want. Going over it costs you the message, never the draft.

10What it deliberately does not do

• No expert rankings, news or analyst commentary. That content belongs to the sites that produce it. Everything here is worked out from public statistics.
• No trade or keeper handling. This simulates one snake draft.
• Nothing to install, and no account required. Opening the page is the whole of it, and a draft on your own, signed out, asks nothing of anybody's server. An account is offered and optional: it exists so your drafts follow you between devices, and changes nothing about how a draft works.

ADP from Fantasy Football Calculator. Player, injury, statistical, projection and depth chart data from Sleeper. Fantasy points recomputed under the scoring rules in section 02. The data files are regenerated daily; the app shows you when, under the Start button.