Touchdown Regression: Why Last Year's TD Leaders Disappoint

The Most Predictable Disappointment in Fantasy

Every August, drafters pay premium picks for last season's touchdown leaders — and every season, most of those players score less. This isn't cynicism; it's arithmetic. Fantasy Points' expected-touchdown study tracked players who dramatically outscored their touchdown expectation and found that in 137 of 151 cases — 90.7% — the player scored fewer touchdowns the following season. Of the decade's top-25 seasons by positive touchdown differential, exactly one player improved on his total the next year, and on average the group's touchdowns fell by 52%. A stat that halves on schedule is not a skill you're buying — it's a coin flip you're paying retail for.

Why Touchdowns Lie

Touchdowns are enormous in fantasy scoring and tiny in sample size. A wide receiver sees maybe 120 targets a season, of which perhaps a dozen are genuine scoring opportunities — so three bounces (a tipped ball, a goal-line vulture, a blown coverage) can swing his fantasy finish by twenty ranks. As FanDuel Research's regression primer explains, a high or low touchdown total simply isn't predictive of the next one; players drift back toward the average implied by their opportunity. The yardage stats around touchdowns — targets, carries, snap and target share — are stable, decision-driven volume. The touchdowns sprinkled on top are weather.

Expected Touchdowns: Measuring the Luck

The tool that separates skill from weather is expected touchdowns (xTD). The method, used in FantasyPros' touchdown regression reports, weighs every carry and target by where it happened — a target at the 2-yard line is worth a large fraction of a touchdown, a catch at midfield almost none — and sums a league-average conversion rate over each player's actual opportunities. The result: how many touchdowns an average player would have scored with this exact usage. A player who scored 13 on 8.1 expected is a regression candidate no matter how good he looked doing it; a player who scored 4 on 8 expected is the discount aisle. The beauty of xTD is that it doesn't ask you to doubt talent — it asks whether the opportunity supports the output.

Regression Does Not Mean He Is "Due"

This is the single most misunderstood idea in fantasy analysis, and getting it backwards costs people money in both directions. Regression to the mean does not say that a player who scored below his expected rate will now score above it to balance the books. Football has no memory, and no ledger is being settled. It says something narrower and more useful: the next stretch of games will look more like his underlying rate than like the unusual stretch you just watched.

The practical difference matters. "He is due" leads you to start a slumping player in a bad matchup because the touchdowns have to come sometime. What regression actually licenses is a valuation claim — that the market has priced a temporary result as though it were a permanent skill, and you can buy or sell against that. It is an argument about price, not about this Sunday.

Which is also why regression is a trade tool far more than a lineup tool. It tells you almost nothing about who to start this week, and quite a lot about who to acquire this month.

Some Touchdown Rates Are Real

The correction has an exception worth respecting, or you will sell good players for the wrong reason. A rate that comes from a ROLE is much more durable than one that comes from variance. A back who gets the ball at the one-yard line because that is his job will keep converting at an elevated rate for as long as the job lasts. A big receiver his team targets in the corner of the end zone is being handed a structural advantage, not a lucky one.

So the question is never simply whether a rate is above average. It is whether anything explains why. If you can point at the deployment — goal-line carries, red-zone target share, a quarterback who looks his way inside the ten — treat the rate as partly earned and discount your regression call accordingly. If the only explanation is that the ball kept finding him from forty yards out, regress it hard.

Drafting Against the Crowd

Regression cuts both ways, and both edges are tradable. The overperformers get drafted at their ceiling: analysts flag them every summer — ESPN's 2026 projections piece named a dozen players, headlined by Jonathan Taylor, projected to score fewer touchdowns this season — yet ADP barely budges, because the market anchors on last year's point totals. The underperformers are the actual buys: players whose yardage and usage held elite while the touchdowns went missing. Positional analyses like 4for4's touchdown trend series walk through both lists each preseason. The discipline is mechanical: fade the player whose ranking requires last year's touchdown rate to repeat, and buy the one whose ranking assumes his bad luck was skill.

• •Check the gap: actual touchdowns minus expected touchdowns is the luck component — big positive gaps regress ~90% of the time
• •Re-rank TD outliers on yardage and usage alone, then let touchdowns be the tiebreaker, not the thesis
• •Buy the unlucky: elite volume with a low TD total is the cheapest way to acquire this year's leap
• •Remember it's regression, not collapse: overperformers usually stay good — they just stop being historic, and their price doesn't know that

The In-Season Version of the Same Edge

Regression thinking isn't just a draft-day tool — it's a weekly waiver filter. When a running back scores three times on eleven touches, your league's FAAB market prices the touchdowns; the xTD lens prices the eleven touches, which is the number that will still be there next Sunday. The same logic protects you from panic-dropping: a starter mired in a "touchdown drought" while his snap share and red-zone usage hold steady isn't broken, he's owed. Volume tells you who to hold and who to chase; touchdowns mostly tell you who just got expensive. In both directions, the manager who prices the opportunity instead of the outcome is trading against a market that keeps making the same mistake.