Research notes
The 2026 touchdown regression file.
July 12, 2026 · 7 min read
Our stickiness study showed that touchdown rates barely carry over from one season to the next. This note applies that finding forward and names names: the 2025 players whose touchdown totals ran far ahead of the chances they actually got, and the ones who got robbed. Then, because this is a research page and not a hot-take page, we show the historical evidence that these gaps really do close, and we’ll grade this exact list in January.
Expected touchdowns, in one paragraph
Every carry and target can be priced by how often it becomes a touchdown given the situation: field position, down, distance, depth of target. Sum those probabilities over a player’s season and you get his expected touchdowns, the scoring his opportunities deserved (we use the nflverse expected-points models, 2019 to 2025). A player who beat his expectation converted better than the league does from the same spots. The whole question is whether that conversion is a skill he keeps.
Mostly, he doesn’t keep it
For receivers, this season’s EXPECTED touchdown rate predicts next season’s actual rate better than this season’s actual rate does (correlation .30 vs .25 among players with 60+ targets in back-to-back years). Read that again: if you want to know who scores next year, the scoreboard is a worse guide than the situations the player was put in. Quarterbacks are the partial exception, as they were in the stickiness study; their actual rate holds slightly more information (.36 vs .32), because elite QB conversion is closer to a real skill. Running back TD rates predict almost nothing either way.
The mean reversion is brutal and symmetric. Receivers who beat their expected touchdowns by 3+ lost an average of 4.2 touchdowns the following season; receivers who ran a touchdown or more behind expectation gained 1.5. Overperformance is borrowed, and the loan comes due in about a year:
| Receivers, year one | n | Avg gap | Avg TD change, year two |
|---|---|---|---|
| 3+ TDs over expectation | 34 | +3.8 | -4.2 |
| 1 to 3 over | 79 | +1.9 | -1.6 |
| About as expected | 169 | 0.0 | +0.2 |
| 1+ under expectation | 105 | -2.1 | +1.5 |
WR/TE with 60+ targets in consecutive seasons, 2019-2024 pairs. The rushing version of this table looks the same (the 3+ over group gave back 2.9).
Right of the line: scored more than the chances deserved. Left: fewer.
The fade list
These players scored well beyond what their 2025 opportunities deserved. To be precise about what this claims: nothing here says these are bad players, and their volume is mostly excellent. It says their touchdown TOTAL had luck in it, and history prices that luck at roughly minus three or four scores next season. Jahmyr Gibbs nearly doubled his expected rushing touchdowns; Dallas Goedert scored eleven times on chances worth five and a half.
| Player | Lane | 2025 TD | Expected | Gap |
|---|---|---|---|---|
| Jalen Hurts PHI | Pass | 25 | 19.2 | +5.8 |
| Dallas Goedert PHI | Rec | 11 | 5.5 | +5.5 |
| Jahmyr Gibbs DET | Rush | 13 | 7.6 | +5.4 |
| Tee Higgins CIN | Rec | 11 | 6.4 | +4.6 |
| Jared Goff DET | Pass | 34 | 29.6 | +4.4 |
| De'Von Achane MIA | Rush | 8 | 4.1 | +3.9 |
| Jonathan Taylor IND | Rush | 18 | 14.6 | +3.4 |
| Jameson Williams DET | Rec | 7 | 4.2 | +2.8 |
| James Cook BUF | Rush | 12 | 9.3 | +2.7 |
| Puka Nacua LA | Rec | 10 | 7.8 | +2.2 |
The bounce-back list
The other side of the ledger, and the more actionable one, because these players come with a built-in discount. Justin Jefferson is the headliner: two touchdowns on 141 targets against an expectation near nine. That is one of the most touchdown-starved seasons a target load like his has produced in this dataset, and everything durable about his profile (the target share) is intact. Dak Prescott threw 30 touchdowns on chances worth 41; Patrick Mahomes ran eight behind his expectation.
| Player | Lane | 2025 TD | Expected | Gap |
|---|---|---|---|---|
| Dak Prescott DAL | Pass | 30 | 40.7 | -10.7 |
| Patrick Mahomes KC | Pass | 22 | 30.1 | -8.1 |
| Justin Jefferson MIN | Rec | 2 | 8.6 | -6.6 |
| CeeDee Lamb DAL | Rec | 3 | 7.8 | -4.8 |
| Bo Nix DEN | Pass | 25 | 29.3 | -4.3 |
| Christian McCaffrey SF | Rush | 10 | 14.1 | -4.1 |
| Tyler Warren IND | Rec | 4 | 7.2 | -3.2 |
| Brian Thomas Jr. JAX | Rec | 2 | 5.0 | -3.0 |
| Jaylen Warren PIT | Rush | 6 | 8.7 | -2.7 |
| Xavier Worthy KC | Rec | 1 | 3.2 | -2.2 |
What the board already prices
Our projections never see this list, and they don’t need to: the model regresses every touchdown rate hard toward position norms (that’s the stickiness finding, built into the architecture), so it reprices this luck automatically. The 2026 board projects Gibbs at about 10 rushing touchdowns instead of 13, Goedert at about 4 instead of 11, Tee Higgins at about 4 instead of 11, and it hands Jefferson a rebound to roughly double his 2025 touchdown count. If you draft off our board you are already fading this list without lifting a finger; if you draft off last season’s points, this note is the correction to apply by hand.
The January receipt
A regression call that never gets graded is just a vibe with a citation. This note goes in the ledger: in January we’ll re-run the file and score both lists, fade by fade, bounce by bounce, in a follow-up note. If the lists don’t beat a coin flip, you’ll read that here too.
Fine print: 2025 gaps computed on regular-season weeks with volume floors of 300 attempts (QB), 120 carries (RB), and 60 targets (WR/TE); lists curated to fantasy-relevant names from the full file. Touchdown totals next season also move with volume, health, and team changes, which is exactly why the validation compares rates under matched floors. 2025 rookies on these lists carry extra uncertainty in both directions since their roles are still settling. Expected-touchdown data: nflverse ff_opportunity.