Glossary

A proposition where all three votes said Over, both quantile artifacts landed on the same side of the book price, and the Scheme Score grade is A+, A, or B where a grade exists. A row the Scheme Score never graded can still reach Tier A and carries the NO_SCHEME_GRADE flag, so a missing grade is never read as a good one.

See also Tier B, Votes, Scheme Score grade

Anything else that cleared every gate: at least two of the three votes, short of the Tier A bar. Tier B is ranked on the same score as Tier A, so the two are directly comparable.

See also Tier A, Gates

Three independent signals decide whether a proposition is in play: the model, RotoWire, and FantasyPoints. The model votes Over when its own modelled Over probability beats the de-vigged book price by more than 0.03. Each vendor votes Over when its projection clears the posted line by at least 0.25 spread units. RotoWire and FantasyPoints are two different companies and are read as two separate votes, never as one vendor with a fallback.

See also Spread unit, De-vigged price

The checks a proposition passes before it is ranked at all: the game has not kicked off, the player is not OUT or QUESTIONABLE, at least one quantile artifact priced the market, at least two of the three votes say Over, and on receiving markets the role is confirmed and the Scheme Score grade is not D or F where a grade exists.

See also Availability, Role confirmation

The one-sigma-equivalent width of a row’s own quantile curve, (q80 - q20) / 1.6832. A vendor gap is scored in spread units rather than as a percentage of the line, because six yards over a 3.5-yard line is close to noise while the same six yards over a 90-yard line is a real disagreement. A spread unit is never a stake: nothing on this site sizes a bet.

See also Vendor gap

How far the blended vendor projection sits above the posted line, reported in spread units first and as a percentage of the line second.

See also Spread unit, Vendor override

The book price with the operator’s margin removed, so the Over side reads as a probability that can be compared with a modelled one. The model’s vote is measured against this number, not against the raw price.

See also American odds, Votes

A letter from A+ down to F for how a receiver’s role fits the defense in front of him. It weighs alignment, target funnels, man and zone rates, likely coverage responsibility, and whether a staff change makes last year’s evidence less trustworthy. Grade bands are percentile-based and recomputed from each week’s graded population, so the cutoffs move with the week rather than sitting on a fixed 0-100 scale.

See also Ungraded, Scheme term

A row the grader could not score, left ungraded rather than guessed. The Scheme Score grades one population, the receiving markets, so a rushing or passing proposition reads Ungraded rather than carrying a grade the product does not compute.

See also Scheme Score grade

The bounded contribution the Scheme Score makes to a board ranking. It is deliberately small: the grade’s forward predictiveness is unvalidated for the opening weeks, so it sits next to an edge term worth several times more on a live row.

See also Scheme Score grade, Tier A

Up to three published 2025 headline metrics for a player, shown beside a projection as historical evidence. It measures last season. It is never a forecast and never model output, and a player with no matched profile says so rather than showing a blank.

How a published projection was assembled. Blended: the model projection was shifted toward the vendor mean by the row’s blend weight, with the quantile spacing unchanged. Model only: the model’s own projection, unchanged, because no vendor supplied anything that met the blend rules or neither vendor covered the player.

See also Vendor override, Vendor only

Both vendors carried a positive projection, agreed with each other, and the model sat outside the wide band around them, so the vendor mean was taken outright at full weight.

See also Blend kind

No model projection stands behind the number. The row was filled from the vendors, so it is scheme and fantasy context and is not priced against a market. Every surface that shows such a number labels it.

See also Blend kind

A published model export carrying a row’s 20th, 50th, and 80th percentile outcome for one player market. Two artifacts landing on the same side of the book price is one of the Tier A conditions.

See also Tier A, Projection range

The 20th to 80th percentile band around a projection, shown as floor, projection, and ceiling. It is the model’s own uncertainty, not a range of opinions.

See also Quantile artifact, Spread unit

One read of a market at one moment, recorded with its operator, its source kind, its as-of time, and the market status at the time it was read. Every line on this site traces to a capture.

See also Capture set, American odds

Every capture behind one board, newest first. A board built from a single pull states one; a merged board lists each pull with its own time, source kind, market status, and scope, because no single scope describes a merged board and a reader comparing a row against a book needs to know which pull it could have come from.

See also Capture, As-of stamp

Three separate times, stated together at the top of every board: when the projection was generated, when the lines were captured, and when availability was last reviewed. They are different clocks and are never collapsed into one. All three are shown in Eastern time.

See also Capture set, Availability

The price format used everywhere on this site, always signed: -106 risks 106 to win 100, +120 risks 100 to win 120. Prices shown are for the Over side, read off the capture named on the row.

See also Capture, Price withheld

A line published with no price beside it. A price is only shown where the capture actually recorded one; where it did not, the row says so rather than borrowing a price from another operator or another moment. A line without price or payout terms cannot yield a break-even probability, so a withheld price is a stated limit on what the row can support.

See also American odds, Line move

A change in the posted number measured within one operator. A different operator quoting a different number is disagreement, not movement, and is withheld as such.

See also Capture

Whether a player is expected to play. It comes from a licensed depth chart plus hand-reviewed news, never from an official team injury report, and it carries the basis it rests on. Every row still needs a live inactives check before kickoff.

See also As-of stamp, Role confirmation

Whether the player’s expected role for this week is confirmed, thin, or absent. On receiving markets a confirmed role is a gate, and the confirmation weight is part of the ranking rather than a note beside it.

See also Gates, Availability