How do NFL DraftKings projections work?

An opportunity projection does not start from a fantasy-point guess. It starts from how many plays a team is likely to run, assigns those plays with snap counts and market shares, and scores many versions of the game under DraftKings rules. What you get is a range: a down game, a middle, and a ceiling, with the rushes, targets, and yards written out beside the score.

That is the whole method. The rest of this note walks through each step in the order the model runs it, and names the columns that show the result on dfs-stack.ai.

The week starts as plays

Before any player gets a number, the game gets a size. Three team inputs do that work: implied points, pace, and rush rate.

Implied points are the Vegas team total. Home implied points are the game total plus the home spread, divided by two. Away implied points are the game total minus the home spread, divided by two. A positive spread means the home team is favored. A 24-point implied total is a bigger scoring environment than a 17-point implied total, and it is what sizes expected offensive touchdowns. It does not, by itself, say who scores them.

Pace is how many plays that team is expected to run. A play here is a targeted throw or a rush attempt, and quarterback scrambles count as rushes. Kneels, spikes, and two-point tries are left out. Two teams with the same implied total can still project very differently if one is lined up for 68 plays and the other for 58. More plays means more chances to hand out.

Rush rate is the share of that pace that is a rush. Targets are what remains: pace times one minus the rush rate. A fast, pass-heavy team creates a large pool of targets. A slow, run-heavy team creates carries instead, even when the game total looks friendly. If a receiver projection looks low in a high-total game, rush rate is one of the first places to look. The points may be real and still not be passing points.

These inputs are the environment. They do not name a player yet. They decide how much work exists.

Snaps decide who is on the field

Snap share is the average offensive snap rate over games before this week. The column is Snap%. Week 1 uses last season, because there is no current-season game to average yet. Later weeks blend last season with the games already played this season.

A snap rate of zero means the player was rostered while the team played and did not take an offensive snap. That is different from a player with no snap history listed. Defenses and kickers do not get a snap rate on this board.

The season rate is the role the projection trusts. The last game can disagree with it, and that disagreement is visible on the snaps board. A back at 60% for the season and 25% last week is a different situation from a back at 30% for the season and 80% last week. The projection follows the blended snap rate. The last game is there so you can see when the role may already have moved, before the average catches up.

Market share hands out the touches

Once a team has a count of rushes and targets, market share splits them among the players on the field.

• Pass share is the share of the team's targeted throws thrown by this player. The slated starting quarterback is given the full share. Backups are not assigned a slice of the passing game on top of the starter.
• Rush share is the share of the team's rush attempts, including quarterback scrambles. The skill players on a team cover the rushing game.
• Target share is the share of the team's targets. The skill players on a team cover the target pool.

A share is not a volume. Volume is the share times the team pool, and the team pool came from pace and rush rate. A receiver at a 25% target share on a 35-target team is a different projection from a receiver at a 25% target share on a 22-target team. Same role inside the offense, different number of chances.

Efficiency turns that volume into a stat line. Catch rate is receptions divided by targets, and it turns expected targets into expected catches. Yards per carry turns rushes into rushing yards. Yards per target turns targets into receiving yards, and it includes incompletions, so it is not yards per reception. Touchdowns follow each player's share of the team's rushing or receiving scores. Interceptions follow the quarterback's interception rate times pass attempts.

A strong per-touch profile on a small share of a small pie still projects quietly. A modest profile on a large share of a large pie can project ahead of it. That ordering is the point of an opportunity model. Role and team volume come first. The stat line sits on top. The fantasy score is last.

DraftKings scoring, many times

Each simulated game draws a pace and a rush rate around those team numbers, assigns the rushes, targets, and passes from the shares, then scores the stat line with DraftKings rules. Passing yards are 0.04 points each and a passing touchdown is 4. Rushing and receiving yards are 0.1 each, and those touchdowns are 6. A reception is 1 point, which is the PPR scoring DraftKings uses for NFL. An interception costs a point, and so does a lost fumble.

Yardage bonuses are included. A player gets 3 extra points at 300 passing yards, at 100 rushing yards, and at 100 receiving yards. Those bonuses are lumpy. A receiver at 99 yards and the same receiver at 100 yards are not a tenth of a point apart. Running the week many times is what turns that jump into a rate, instead of forcing every player onto a single stat line that either includes the bonus or does not.

You do not have to read every simulated game. The board keeps the distribution and the expected volume.

• p10 is a down game. About 10% of simulations score that or less.
• p50 is the median, the middle of the simulations.
• p90 is the ceiling on this board. About 10% of simulations score that or more.
• mean is the expected score, the average of the simulations, with bonuses included.

Beside the score you get expected rushes, targets, receptions, yards, and touchdowns. When a mean looks aggressive, those columns are the check. A high score with a thin target share is a different bet from a high score the volume already explains.

There is also a Base column. It is off by default. Turn it on from Columns when you want it. Base is the linear DraftKings score from the expected stats, with the yardage bonuses left out. It is the volume story without the spikes. mean is the same week with the spikes averaged in. Most of the gap between them is how often a 100-yard or 300-yard bonus is expected to land.

What this is not

This is not a play-by-play forecast of one script, and it is not a pick. It will not tell you the exact stat line on Sunday. It tells you the range that follows from the plays, the snaps, and the shares, scored the way DraftKings scores.

Team defense is not built from snap share and target share. Defense scores are a simpler estimate tied to the opponent, not the same opportunity path used for quarterbacks, running backs, receivers, and tight ends.

Classic slates and captain Showdown slates use the same player projections. The slate picker switches draft groups, including Thursday, Sunday, and Monday captain slates. Showdown captain scoring, at 1.5 times, applies when that roster is built. The player projection underneath is the same opportunity number.

Projections, start/sit, snaps, and stacks are open without an account. Google sign-in is only required when you want to build lineups.

Read the range next

mean, p50, and p90 are three different questions about the same simulations. Cash games and large tournaments do not want the same answer. The next note is what mean, median, and ceiling mean in DFS.