Anytime touchdown
A player scoring at least one touchdown in the game, regardless of when. This model estimates rushing and receiving TDs only. Sportsbook anytime-TD markets can also settle on returns, fumble recoveries, and other scoring types.
How to read this product
Experimental projections
Model output for research and comparison. Not a proven profitable betting recommendation, and not a claim that a price is a lock or a steal.
Unavailable
The value is missing in the cached data. It is not a zero. Examples: no Yes price at the selected books, no snap share on file, or no completed games yet this season.
Pregame
Information that would have been knowable before the game’s kickoff. Current-game stats are never used to project that same game.
Poisson baseline
The scoring rule P(at least one rushing or receiving TD) = 1 − exp(−λ). λ is the player’s expected rushing TD count plus expected receiving TD count. This is a baseline assumption, not a guarantee of how scoring is distributed.
Allocation
Stage B of the model: after team expected rushing and passing TDs are estimated, those TDs are split among the team’s skill players using pregame usage. Passing TDs are given to the receiver, not the passer. A residual share is left for unmodeled scorers.
Projection table
Player
The offensive skill player being projected. Click a row to expand model components, book prices, and recent usage.
Pos / position group
Roster position, grouped as QB, RB (including FB/HB), WR, or TE. Position group is used for priors when a player’s own sample is small.
Team
The player’s team for that game.
Opp
The opponent. “vs” means the player’s team is at home; “@” means on the road.
Kickoff
Scheduled game start, shown in your local timezone.
Model TD %
The model’s probability that the player scores at least one rushing or receiving touchdown. It is not the sportsbook implied probability and is not vig-free market consensus.
Expected rush TDs
The player’s allocated slice of the team’s expected rushing touchdowns. Return TDs are excluded.
Expected rec TDs
The player’s allocated slice of the team’s expected receiving (passing) touchdowns. Quarterbacks do not receive this for throws they make; the target does.
λ / expected TD count
Expected rush TDs plus expected rec TDs. The Model TD % is 1 − exp(−λ). A player can have λ below 1.0 and still have a substantial chance of scoring once.
Team implied points
How many points the market (or nflverse lines, if sportsbook totals are missing) implies this team will score. Used to anchor team TD volume. Not points divided by 7.
Input summary
A plain-language description of stored model inputs for the row. It is not a claim that any one statistic caused the projection.
Odds and estimated return
Fair odds
American odds that would be break-even if the Model TD % were exactly true: decimal odds = 1 / model probability, then converted to American. This is not a sportsbook price.
American odds
US-style price. Positive (for example +150) means a 100 stake wins 150. Negative (for example −150) means you must stake 150 to win 100.
Decimal odds
Total return per 1 unit staked, including stake. +100 American is 2.00 decimal. Used internally to compute expected return.
Yes price
The sportsbook price on the player to score a touchdown (“Yes” in the anytime-TD market). Comparison only; not a model input.
Best book
Among the checked sportsbooks, the longest (best for the bettor) Yes price and the book offering it. Unchecked books are ignored. If several books match, the longest decimal odds wins.
Est. return
Estimated expected return per 1 unit staked at the displayed best Yes price: model probability × decimal odds − 1. Shown in cents on the dollar (for example +12.0¢). An estimate at that price, not a guaranteed edge. Recalculated when you change sportsbooks or player history.
Raw implied / sportsbook implied probability
The Yes price converted to a probability, including the book’s margin (vig). It is not a vig-free estimate and is not comparable 1:1 with Model TD %.
Vig
The sportsbook hold built into prices. Because of vig, implied probabilities on a market usually sum to more than 100%.
Suppressed / comparison suppressed
A Yes price is on file but is not used for best-price or expected-return comparison, usually because matching or settlement support is incomplete. It is not guessed into a number.
Sportsbooks filter
Checkboxes choose which books’ Yes prices appear. Unchecked books disappear from the Best book column, estimated return, and the expanded price table. This does not refetch odds.
Usage and opportunity
Carry
An official-style rushing attempt: designed runs credited to the rusher. Nullified plays and two-point attempts are excluded. Kneels and spikes are not treated as carries.
Target
A pass intended for the receiver (complete or incomplete). Being on a pass play does not prove the player ran a route or was targeted. Two-point passes and nullified plays are excluded.
Carry share / pregame carry share
The player’s share of team carries, then smoothed toward the position average only if the player actually recorded carries. Observed zeros stay zeros. This is an allocation input.
Target share / pregame target share
The player’s share of team targets (or red-zone targets). After the player has games, a 0 target count is not blended with a typical WR/TE share. Share is not the same as targets per game.
Snap share
Estimated share of offensive snaps from PFR snap counts, lagged to pregame and smoothed. Participation charts from 2023+ are not used as in-season features.
Smoothing / position prior
Small samples are blended with a position-group average only for usage the player actually recorded. Weight on the player’s own rate is n / (n + 4), where n is prior games. If they have games and 0 carries, 0 targets, or 0 red-zone opportunities, those shares stay 0. Position priors are still used in week 1 / no history. We do not have red-zone snap counts; red-zone usage is carries and targets from play-by-play.
Usage tier
Lagged, pregame ranking of how the player has been used, then used to allocate team TDs. Order: red-zone carries or targets; then any carries or targets; then snaps with no touches; then no snaps. Historical anytime-TD rates follow that order (about 12%, 3%, 0.9%, and 0.2% on 2018–2024 skill-position games). Allocation weights are the ratio of each tier’s next-game TD rate to the red-zone rate. This is not a red-zone snap share; nflverse in-season participation is not used.
RZ20
Plays at or inside the opponent’s 20-yard line (yardline_100 ≤ 20). Includes scoring chances outside the 10 and 5. A missing yardline is not treated as a red-zone opportunity.
Inside-10
Plays at or inside the opponent’s 10-yard line.
Inside-5
Plays at or inside the opponent’s 5-yard line.
Carries / game and targets / game
Raw lagged averages of counts (not smoothed shares). Two-decimal 0.00 means the average is zero or rounds to zero. These can disagree with the corresponding share % because share is smoothed.
Recent weekly usage
Completed games before this kickoff. Black bars are carries; light-blue bars are targets. The steel dashed line is season-to-date carries per game when available. Under All seasons this can include last year; under Current season only it does not.
Season-to-date
Averages from completed games in the same NFL season as the projected game, still before kickoff.
Filters and views
Player history
Which games feed player usage, shares, and the allocation. Team implied points and sportsbook prices stay this week’s market either way. The model is not retrained when you switch views.
All seasons
Uses the player’s full stored history (lagged, through prior seasons). Default view. A veteran’s carry share can still look like last year’s role.
Current season only
Rebuilds usage, shares, Model TD %, fair odds, and estimated return from this season’s completed games before kickoff. Unused roles stay at 0 after games have been played; week 1 still uses the position prior.
Week, game, team, position
Display filters only. They do not retrain the model or fetch odds. Search matches player name.
Stats page
Observed lagged usage and scoring from completed games for the same upcoming slate as Projections: carries, targets, red-zone carries/targets, rushing and receiving TDs per game, snap share, and usage tier. These are raw per-game averages, not position-prior smoothing and not Model TD %. Current season only uses this year’s completed games. Players with no games yet in that window show Unavailable.
Warnings
Warning chips are data-quality notes, not injury reports. Click a chip on the projections table to open its definition.
Role change flag
The player’s last-4-game carry share differs from season-to-date carry share by more than 15 percentage points. It does not mean injured, demoted, or a new starter. Early in the season it can fire noisily because season-to-date is one or two games. Under Current season only, both sides of the comparison are within this year.
No player history
No prior games are in the history window used for this view. Rookies, new call-ups, and week 1 under Current season only often have this. Position priors may still be used when there are no games yet to observe. After the player has had games and still has no snaps, carries, or targets, see no play time.
No play time
The player has prior games in this view’s window but no recorded snaps, carries, or targets. Position-average shares are not applied. Pregame carry, target, and snap shares are set to 0 so the allocation does not invent a typical RB/WR/TE role. This is not an official inactive or injury tag.
Snaps only
The player has been on the field (snaps) but has no lagged carries or targets. Carry, target, and red-zone shares stay 0. Allocation is reduced to the historical snaps-only TD rate, which is much lower than for players who get the ball.
No red zone usage
The player has lagged carries or targets, but none inside the 20. Red-zone shares stay 0. Rushing/receiving allocation is reduced toward the historical open-field TD rate rather than treated like a red-zone role. This uses red-zone carries and targets, not red-zone snaps.
Snap share missing
No lagged offensive snap share is on file for this player-game. The model does not fill that with zero silently.
Implied points missing
No market or schedule total/spread was available to form team implied points. Team TD volume is weaker without it.
Small current season sample
Shown in the Current season only view when the player has fewer than four completed games this season before kickoff. Shares are still only smoothed for usage that actually happened; zeros stay zeros.
No comparable anytime TD price
No usable Yes price was matched for this player among the cached books (or among the books you still have checked). Best book and estimated return show Unavailable.
Roster week carried forward
The weekly roster used for this game was copied from an earlier week when the same week’s roster was not on file. It is not an injury designation.
Status and freshness
Projection update
When the cached projection file was written. Opening this page does not retrain.
Odds snapshot
When the cached Odds API file was fetched. Opening, sorting, or filtering does not spend API credits.
Stale
The projection file is older than 24 hours, or the odds snapshot is older than the configured stale window. The numbers are still shown, with this warning.
Demo fixture
Labeled demo rows, not live model output. The dashboard will not invent demo players if that file is missing.