Read The Report

QB = Quarterback, the player who throws (and sometimes runs) the ball. RB = Running Back, who mostly runs the ball and catches short passes. WR = Wide Receiver, who mainly catches passes, often downfield. TE = Tight End, a hybrid role that both catches passes and blocks. Every table on this site is organized around these four.

Every table on the site is split into two labelled bands, and they answer different questions.

Projected is the model's forecast for this game — fantasy points and the counting stats underneath them. These are shaded against the player's own recent pace. Blue means we expect more from him than he has been producing; red means less. Nobody else's numbers enter into it.

Context is everything that explains the forecast — snap share, matchup, touchdown probability, red-zone work, target depth, depth-chart spot. These are shaded against everyone else playing this week at his position. Blue means he ranks well among that group; red means poorly.

That distinction matters more than it sounds. A backup can be blue right across the Projected band — he is projected well above his own quiet recent pace — while being red right across Context, because compared with the rest of the league he is still a backup. Both readings are true, and the split is there so you can see both at once instead of one of them quietly overwriting the other.

Inside the Projected band, shading is per stat, not per player — a QB can show blue for passing yards and red for interceptions in the same row. For interceptions specifically, blue means fewer projected than normal, since fewer turnovers is the better outlook.

Inside the Context band, the middle of the field is neutral and the colour builds out towards the edges: only players genuinely near the top or bottom of their position get strong colour.

FPTS carries the projection on the first line and the likely range under it, so the number and its uncertainty are read together rather than three columns apart.

The player's name is coloured too, and it means something different from every other colour on the page: it is how much history the trailing numbers rest on. White is a full season or more behind them. Grey is about half a season. Gold means only three to seven games — thin. A muted red means fewer than three, which is barely a sample at all. It is not a judgement of the player; it is a judgement of how much the numbers beside him are worth.

Best plays sits in each team's header card: the three players on that side most worth a look, ranked by projected points relative to their own position blended with Matchup Score, 60/40. Relative to position is what makes a mixed list possible — 17.8 points is an ordinary week for a quarterback and an enormous one for a tight end, so ranking raw points would just nominate three quarterbacks every week. A tight end at 5.9 can outrank a receiver at 11.7 for the same reason.

Two things worth being upfront about. The 60/40 split is a reasoned choice, not a backtested one; and Matchup Score already contains the player's own recent volume, so volume counts twice, softly, which tilts the list toward high-usage players. That is why the card shows the projection and the matchup score beside each name rather than one blended number — you can disagree with the order on the evidence.

Each position box shows the top twelve of whatever is sorted, not a fixed twelve reshuffled — change the column and the set of players changes with it.

Tap any row to open it. The detail panel carries the floor and ceiling, the sample size, his team's Vegas-implied total and pace, and the opponent's ranks — the numbers that matter sometimes, without spending permanent width on a table you already have to scroll sideways.

A real 0-100 score for how favorable this specific player's opportunity is this week — not just “is the defense bad,” but “is this player actually positioned to exploit it.” Two players on the same team facing the same defense can have very different Matchup Scores if one of them barely plays.

It combines the defense's real difficulty (what they actually give up to this specific position — yards and touchdowns, not one overall rank) with the game's Vegas-implied scoring environment, then scales that by the player's own recent volume.

The weights aren't guesses. Tested across 17,348 real player-weeks, this version correlates 0.539 with actual fantasy points, up from 0.445 for the original formula — about 21% better. The biggest single finding was that volume matters roughly twice as much as originally assumed: a low-volume player in a great matchup deserves to be pushed well down, not held near the middle.

The real probability a player scores at least one touchdown this week — not a raw projected count. Converted from the model's projected TD mean via the Poisson distribution (the standard way to make this conversion): 1 minus e to the power of negative the projected mean. For running backs and wide receivers, rushing and receiving TD chances are combined into one number, since an anytime-TD bet doesn't care which way they got in. For quarterbacks, only rushing TDs count toward their own odds — a passing touchdown is thrown, not scored by the QB. The projected mean comes from his projected workload — carries and targets — times how often players at his position score per touch, which ranked players better than projecting touchdowns directly. Checked on 54 weeks of real results (2023–2025): take one player who scored and one who didn't, and it had the scorer higher about 71% of the time. From 20% up the numbers match how often players actually scored. Below 20% they run about 3 points high: players given 10–20% scored about 12% of the time, not 15%.

Since 10 October (Week 5, from the Sunday games on) two more things go into it. Goal-line work right now:a quarter of the goal-line part comes from his share of his team's runs inside the 5 over its last eight games, so the back getting those carries today counts for more than career history alone. The team's touchdown budget: each team's chances are nudged part of the way toward the touchdowns its Vegas total points to, up or down. On the same 54 weeks this improved the score against the base rate in every season (9.4% to 9.8% in 2025, the season it wasn't fitted on) and ranked scorers slightly better. The accuracy record grades it as its own line, never mixed with the earlier version.

Every stat comes with a projection plus a Low and High. That range isn't a guess — it's built from how wrong this exact model has actually been, tested on real weeks it never saw during training.

In backtesting it was an 80% range — the result landed inside it about four times in five. This season it has caught between 61% and 72% of results a week through Week 4 (66%, 68%, 61%, 72%), so for now read it as closer to a two-in-three range. Until enough of the season has been scored to say, the range is labelled simply as the likely range rather than carrying a number it hasn't earned yet. The real figure goes back on the label once it is measured rather than assumed.

A tight range means the model has been reliably accurate for this type of player. A wide range is an honest “this could go either way” — often because the player is in a role with more inherent uncertainty (a backup, a rookie, a committee back).

The 1st Quarter tab on each matchup is deliberately the odd one out: it contains no projections at all. It is a record of how each offense has actually opened games and who has been on the field for it — plays run, pass rate, and each player's first-quarter touches.

The row that does the most work is Script gap: first-quarter pass rate minus full-game pass rate. A team well above zero opens through the air and settles down later; one well below opens on the ground. That is the scripted portion of the game plan, and it is not visible anywhere else on the site.

Q1 TD % is built from first-quarter volume and the league touchdown-per-touch rate, exactly the way Anytime TD % is built — not from how often a player has actually scored in a first quarter. That version looks more authoritative and is mostly noise: five first-quarter touchdowns in seventeen games reads as 29%, but the honest range around five-of-seventeen runs from roughly 10% to 56%.

The player columns on that tab are shaded the same way the Context band is — against everyone else at the same position, not against the other players in this game. So a receiver with 1.9 first-quarter touches can be a deeper blue than a running back with 2.0, because those are very different numbers for the two roles. All Touches is left unshaded on purpose: it is there as the denominator that makes Early Share readable, not as a verdict on anybody.

We track this openly. Every week, we compare projections against what actually happened and publish the real results — not cherry-picked good weeks.

This site uses PPR scoring by default — short for “Points Per Reception,” meaning a player earns a real point just for catching a pass, not only for the yards or touchdown that follow. The full breakdown: 1 point per 25 passing yards, 4 per passing TD, -2 per interception; 1 point per 10 rushing/receiving yards, 6 per rushing/receiving TD; 1 point per reception.

Play a different scoring format? League Settings lets you match your real league's rules, and every FPTS number on the site updates automatically.

Status — this week's real injury designation. Out and Doubtful players are already zeroed out.

NEW badge — this player has zero career games logged (a genuine first-time call-up: practice-squad promotion, waiver signing, never-played rookie). Their row uses a real draft-capital debut-game baseline instead of the main model — treat with extra caution.

On bye — named beside the matchups at the top of the page. Teams with no game this week are left off the board entirely, so their players don't appear in any table, ranking or color, and they aren't counted in the accuracy record for that week.

Every stat and column you'll see on the site, grouped by what it's telling you.