Anytime Touchdown Scorer Predictions

How these touchdown probabilities are built

Each number on this board is the output of an XGBoost classifier that estimates the chance a player scores at least one touchdown in his next game. The model reads recent usage rather than reputation: carries and targets inside the twenty, share of his team's red-zone work, snap share, role changes caused by injuries ahead of him on the depth chart, and the defence he faces — how often it concedes rushing scores versus receiving scores.

Touchdowns are a rare, lumpy event. A featured running back on a good offence lands somewhere around 50-60%, a primary receiver in the thirties, and a rotational player in single digits. A 25% player who does not score is the expected outcome, not a miss — which is why this page ranks probabilities instead of publishing picks.

Return touchdowns are excluded. The board covers rushing and receiving scores only, matching how sportsbooks settle the standard anytime touchdown market.

Reading anytime touchdown odds

Sportsbooks quote this market in American odds. To compare a price with a probability on this page, convert it: a positive price becomes 100 / (odds + 100), so +150 is 40%. A negative price becomes odds / (odds + 100), so -140 is 58.3%. If the model gives a player a higher chance than the converted price implies, that is where the value would sit.

One caveat that matters more here than in any other market: the vig on anytime touchdown is enormous. Add up the implied probabilities of every player in a single game and the total regularly reaches 400-500% rather than 100%, because a book prices twenty-five players who can each score independently. A raw price therefore overstates every player's real chance, and a naive comparison against it will look like an edge on almost everybody. Any honest comparison has to strip that overround out first.

That is how we check ourselves. Anytime touchdown is a reference projection, not a graded play — in the NFL only the moneyline is graded. We still compare it head-to-head against one book's de-vigged price from a single pre-kickoff snapshot — a row counts for us only when our number was closer to what actually happened than theirs was — and that comparison is shown on the performance page, wins and losses alike.

Why the board changes during the week

Availability is applied when you load the page, not when the projections were generated. A player who lands on injured reserve, or who is ruled out on the final injury report, disappears from this board even though the underlying projection still exists — a projection for someone who will not play is worse than no projection at all.

That makes the late-week board meaningfully different from the Tuesday one. When a starting running back is ruled out, his backup's touchdown chance is the number that moves most, and it is usually the last thing the market adjusts. The NFL injury report page lists who is out game by game, with starters flagged.