How to Predict Anytime Touchdown Scorers
Anytime touchdown is the most popular player prop in football and the one most people bet on feel. It is also a ranking problem with a known base rate, which means it can be done properly. Here is what actually moves a player's chance of scoring, how to turn that chance into a price, and what our model does and does not claim.
Published September 2026 · 9 min read
1. Start With the Base Rate
Before any model, know the floor. Among skill-position players who are actually in the rotation, roughly one in five scores a touchdown in a given game. It is not evenly spread:
Share of player-games with at least one touchdown — running backs, receivers and tight ends averaging three or more carries-plus-targets over their prior six games, 11 seasons from 2015.
That table is the whole reason naive touchdown picks look good. Write down five running backs every Sunday and about one of them scores, without any analysis at all. A prediction only means something if it does better than the position it is picking from — and the books already price it that way, which is why a lead back sits at -130 and a third receiver at +400.
2. It Is a Ranking Problem, Not a Yes/No Question
The wrong way to build this is a classifier that answers “will he score?” With 80% of the answers being no, that model learns to say no and looks 80% accurate while telling you nothing. The same trap exists in baseball home run props, where 89% of batter-games produce no homer.
The right question is which players on this week's slate are most likely to score, relative to each other. That is a ranking. You score a ranker by taking the top of the list and measuring how often those players actually scored against the base rate — and, more importantly, against the simplest ranking anyone could make without a model.
That last part is the bar we hold ourselves to. Ranking players by their recent carries and targets beats the base rate on its own, because volume predicts scoring. So our model is scored against that one-line ranker on the identical slate every week, walk-forward by season. If it cannot beat “sort by touches,” the other seventy features are not earning their place and it is not a product.
3. What Actually Moves the Probability
Opportunity, first and by a distance
Carries plus targets over the last six games. You cannot score without the ball, and how often a player gets it is the most stable thing about him from week to week. Talent shows up in the model, but it shows up through usage — coaches give the ball to the players who produce.
Where the opportunities happen
Ten carries between the 20s and ten carries inside the 10-yard line are not the same ten carries. A goal-line role converts a small share of touches into a large share of a team's touchdowns, which is why a 12-carry back can out-score a 20-carry back on the same team. Red-zone targets do the same job for receivers and tight ends; a tight end's 14% base rate is dragged up sharply by red-zone usage.
Recent touchdown rate — useful, but noisy
A player's touchdown rate over his last ten games carries real information, but touchdowns are rare events and ten games is a small sample. Two scores in a game moves it a lot; one bad week moves it a lot. The model uses it, weighted below opportunity, and so should you.
Team scoring environment
Touchdowns are shared out of a team's total. An offense expected to score 28 has more touchdowns to distribute than one expected to score 17, and the market's team total is a decent proxy. Game script matters at the margin — a team likely to be trailing throws more, which favours receivers over the lead back late.
What matters less than people think
Opponent defensive rankings. “Worst run defense in the league” is a real effect but a small one next to usage, and it is already in the team total. A player's name, his contract and his fantasy ranking predict nothing on their own.
4. Turning a Probability Into a Price
A probability is only useful next to the odds. The conversion:
| Model probability | Fair odds | Worth a look at |
|---|---|---|
| 60% | -150 | -125 or better |
| 45% | +122 | +150 or better |
| 30% | +233 | +275 or better |
| 20% | +400 | +450 or better |
The gap between the fair price and what you need is the book's margin, which on anytime touchdown runs 8–15% — higher than on spreads, because this market is popular and popular markets are priced tighter against the bettor. A 60% player at -175 is a bad bet even though he will probably score. A 30% player at +300 is a reasonable one even though he probably will not. That is the whole discipline: bet the price, not the player.
5. Injuries Move This Market More Than Matchups Do
When a lead back is ruled out on Friday, his fifteen carries do not vanish. They move, usually to one player, and that player's touchdown probability can double overnight — a bigger move than any defensive matchup produces. Watching the injury report for props is mostly watching where the opportunities go.
Two separate lists decide whether a player can even be on the board. Roster status — injured reserve, PUP, the exempt list, practice squad — makes a player ineligible, and he never appears on the weekly report at all. The weekly designation — Out, Doubtful, Questionable — covers players who are eligible but hurt. Our board removes both ineligible players and anyone Out or Doubtful automatically; Questionable players stay, flagged, because most of them play. There is a fuller guide to reading the report for props here.
6. What We Publish, and What We Don't Claim
Every week the model ranks every eligible running back, receiver and tight end on the slate by his probability of scoring. The full list is free, no account required, on the anytime touchdown scorer board, with position filters and players ruled out removed.
What you will not see on it is a play badge. On our MLB boards, a badge means the pick has a graded record that clears the price. Anytime touchdown does not yet: graded against the closing price, the record is close to a coin flip, and the market's margin is wide. The ranking beats the naive volume ranker it is measured against — that is the claim we can support. “This price is wrong” is a claim we cannot support yet, so we do not make it. We would rather publish a number and be honest about its limits than publish a badge and be wrong.
Where the model earns its keep is the same place it does in every sport we grade: giving you a probability to hold against the price, so the decision is about the number and not the name.
This week's anytime touchdown board
Every eligible player on the slate, ranked by the model's probability of scoring. Free, updated weekly, ruled-out players removed. Rushing and receiving yardage projections are on the NFL props board.
Frequently Asked Questions
What percentage of NFL players score a touchdown in a given game?
About one in five skill-position players in a regular rotation score in any given game. By position the base rate is roughly 24% for running backs, 19% for wide receivers and 14% for tight ends. Those numbers are the floor any anytime-TD prediction has to beat: a list of running backs will "hit" a quarter of the time on its own.
What is the best predictor of an anytime touchdown?
Opportunity. Carries plus targets over the last few games predicts scoring better than any single talent stat, because you cannot score without the ball. The second-best predictor is where those opportunities happen: a player who gets the ball inside the 10-yard line converts a small share of his touches into a large share of his team's touchdowns. Recent touchdown rate matters too, but it is noisier than volume because touchdowns are rare events.
How do I convert a touchdown probability into fair odds?
Fair American odds for a probability p above 50% are -100 × p ÷ (1 − p); below 50% they are +100 × (1 − p) ÷ p. A 40% player is fair at +150, a 25% player at +300, a 60% player at -150. If the book's price implies a higher probability than the model gives — say +110 (47.6%) on a 40% player — there is no value regardless of how good the player is.
Why does Prediction Engine not mark anytime-TD picks as plays?
Because the graded record against the closing price is close to a coin flip, and we do not attach a play badge to anything we cannot show a graded edge for. The ranking itself is real — it beats ranking by recent volume, which is the bar it is scored against — but "the model ranks this player highly" and "this price is wrong" are different claims. The board publishes the first and lets you judge the second.
Do injuries change anytime touchdown probabilities?
Mostly through redistribution. When a lead running back is ruled out, the touches do not disappear — they move, usually to one player, and that player's probability jumps by more than any matchup factor could move it. Players ruled Out or Doubtful, and anyone on injured reserve or the practice squad, are removed from our board automatically; Questionable players stay on it because most of them play.
The week ahead, once a week
Last week's graded record — wins and losses — the plays for the week ahead, and a subscriber discount code. One email, no daily noise, unsubscribe in a click.
Responsible Gambling
Educational content only. Past performance does not guarantee future results. Sports betting carries financial risk and may be illegal in your jurisdiction. Must be 21+ in legal US states.
If you or someone you know has a gambling problem, call 1-800-GAMBLER or visit ncpgambling.org.