Every offseason, the running back market sets itself up the same way. A back posts a monster year — 1,400 yards, 14 touchdowns, RB3 finish — and his ADP rockets into the first round. Think of the way a Raheem Mostert-style TD spike or a De'Von Achane-style explosive efficiency run can pull a whole draft room toward last year's ceiling. A year later, half of those backs have drifted back into the RB2 range. The other half got hurt. This is not bad luck. It is regression to the mean, and the Shark Snip projection model is built around expecting it.
What regression actually means
Regression to the mean is not a prediction that a good player will become bad. It is the observation that extreme outcomes contain noise, and noise does not repeat. A running back who scored 14 touchdowns is partly a great player and partly the lucky recipient of a season's worth of coin flips inside the 5-yard line. Strip the luck out, and his "true" rate was probably closer to 9 or 10 — which means his fantasy ceiling next year is lower than his last finish suggests.
Our model breaks every RB stat line into two components: opportunity (carries, targets, red-zone touches, snap share) and efficiency (yards per carry, yards per route, TD rate per touch). Opportunity is sticky. Efficiency, especially TD rate, is not. You can see this in any backtest going back to the snap-share era: backs who keep their volume keep most of their finish. Backs who lose their volume do not, no matter how efficient they were the year before.
The mental model that gets this wrong is "talent." Talent does exist, but it shows up in median outcomes, route participation, and ability to stay on the field — not in the tail outliers that drive a top-3 RB finish. When you draft a back at his last finish, you are buying both the talent and the tail. Only one of those repeats.
Why TD rate is the biggest regression flag
Touchdown rate per carry has the lowest year-over-year correlation of any major RB stat. League average is roughly one TD per 60 carries. A back who scored one TD per 30 carries last year almost certainly will not repeat that. The reason is structural — touchdowns are scarce events in a small window of the field, so a few extra coin-flip outcomes inflate or deflate them dramatically. Bijan Robinson's 2025 TD rate sat well above the RB baseline, and Jahmyr Gibbs' similar spike both fall in this zone, even though both are genuinely excellent runners. TD regression points Bijan's 2026 total back toward the league baseline rather than another league-leading year, and the ADP has not moved enough to reflect it.
Yards per carry behaves similarly. Elite single-season YPC is one of the least sticky RB stats year over year — backs who post a big YPC season usually drift back toward their career norm the following year, which is enough to drop a top-5 fantasy finish well down the rankings without anything "going wrong" in a narrative sense. Saquon Barkley posted 5.6 YPC behind Philadelphia's line in 2024; his YPC regressed toward his career norm in 2025 with similar volume, and his finish slipped accordingly. That is not a collapse. That is the math working exactly as advertised.
The three regression archetypes
Our model flags RBs as regression candidates using a feature called opportunity_minus_efficiency_share, which compares a back's volume percentile to his per-touch percentile. When the per-touch number is much higher than the volume number, regression is likely. We see three flavors of this every year, and they are the easiest fades to find on a draft board.
1. The TD-vulture beneficiary
A back who got 6 of his 12 TDs from the 1-yard line behind a great offensive line. Volume was fine; TDs did the heavy lifting. Jamaal Williams' 2022 Lions season is the clean historical example: the role near the goal line was real, but the TD total was the fantasy engine. Strip out half those TDs and you are looking at an RB2, not an RB1. If the offensive line lost a starter or the team's red-zone trips are projected to fall, you have a clean fade.
The 2026 version of this profile is Kyren Williams. He posted 12 TDs on 230 carries with a goal-line role that has been worth 4-5 fantasy points per game in scoring. The 2026 Rams shift their OC and almost certainly cede some red-zone share to Blake Corum. If you back out 3 of those TDs, Kyren projects RB14-16, not RB6, and his current ADP sits at RB7.
2. The big-play-driven YPC outlier
A back whose 5.6 YPC was juiced by three 60+ yard runs. The model strips explosives out and looks at median rush yards. Median rushing yards is much stickier than mean, and it is the better predictor of next-year efficiency. De'Von Achane's 2024 explosive-play rate was a clear outlier — five runs of 40+ yards is roughly double the per-touch rate of the next-best back. Achane is real, but the 2026 projection bakes in mean reversion to roughly a 4.9 YPC line, not the 6.1 you saw last year.
3. The high-leverage receiver
A back who caught 70 passes but averaged 12+ yards per reception on a tiny target diet of designed downfield concepts. PPR scoring is sticky on volume; it is not sticky on yards-per-target. When the target depth was driven by 4 or 5 broken plays, the next-year median catch radius collapses, and the PPR floor with it. Christian McCaffrey is the rare back who maintains this profile because his receiving role is structural, not opportunistic — the model treats him differently because of his route participation rate above 65%.
What does NOT regress
Just as important as the regression flags is knowing what holds. The model treats these features as sticky year-over-year:
- Snap share. Backs over 70% snap share retain that workload roughly 75% of the time when healthy. Christian McCaffrey, Bijan Robinson, Saquon Barkley, and Jonathan Taylor are the current 70%+ club, and they tend to stay there.
- Route participation rate. If the back is on the field on third-and-7, he was last year too, and that is one of the strongest signals for PPR floor. Bucky Irving's route participation is the reason his ADP is climbing despite a moderate TD total.
- Target share among RBs on his team. Coaching staffs treat backs the same way two years in a row much more often than not, even across OC changes.
- Red-zone carries (not red-zone TDs). Volume in the red zone is sticky; conversion rate is not. A back with 35 red-zone carries last year will get 30-40 this year if healthy; the TD count attached to those carries is where the noise lives.
The cleanest fade is a back whose opportunity metrics all dropped slightly while his efficiency metrics all spiked. Those backs hit RB1 finishes in their follow-up year less than 25% of the time, even at top-12 ADPs.
How to use this in a draft
Pull up our picks page and sort RBs by the regression-score column. Anything in the red — meaning efficiency outpaced opportunity by more than one standard deviation — is a candidate to fade relative to ADP. Anything in green is a candidate to draft early. The model is not asking you to never draft a hot back; it is asking you to discount the ADP based on how much of last year's finish came from sticky inputs. Then sanity-check that discount against the preseason RB1 hit-rate backtest, because fragile RB1 profiles are where the market most often overpays.
Pair this with the ADP value tiers piece for a structural view of where consensus systematically overpays at the position, and the target share guide for the WR-side mirror of the same idea.
The veteran-back trap
One specific case worth calling out: 30+ year old RBs coming off a top-10 finish. Historically, that profile has hit a top-15 finish again about 30% of the time. Consensus drafts them in the top-10. That is a 65-cent dollar at every pick, year after year, and it is the single most consistent regression edge in fantasy football. Derrick Henry's 2024 Ravens line had everyone reaching in 2025 drafts; his 2025 production regressed off his 2024 peak as the age curve caught up, well short of where his ADP priced him. Aaron Jones in 2025 is the same pattern with a different jersey.
The reason is workload memory. Backs over 28 have already absorbed a career's worth of hits, and the next 250-carry season usually comes with a soft-tissue dip mid-year. The model penalizes age-adjusted touches above 1,800 career carries by about 8% on the projected workload, and the 35+ tier (Henry, Jones, occasionally an Alvin Kamara) takes a 15% penalty. ADP rarely moves that much.
Building this into your own model
If you want to test these features yourself, open the model builder in tinker and start with a regression target of "next-season fantasy PPR finish rank." Use these inputs in order of historical importance: prior snap share (highest), route participation rate, age, age-squared, team total over/under for the upcoming season, OC continuity flag, red-zone carries, and TD-per-touch (lowest, because it should regress to a baseline anyway). The model gives the highest weight to snap share and the lowest to TD rate, which inverts how most draft boards are built.
If you would rather build a full reproducible pipeline, the workshop exposes the same player feature store and projection layer you can fork into your own backtest. The builder lets you ship the result as a daily refresh that runs on your machine.
What the 2026 board looks like through this lens
Three concrete fades and three concrete buys, based on the current model output as of mid-May:
- Fades: Bijan Robinson at RB1 (TD regression), Kyren Williams at RB7 (TD regression plus OC turnover), Aaron Jones at RB18 (age plus team passing-game tilt).
- Buys: Bucky Irving at RB10 (snap share trending up, route participation elite), Jonathan Taylor at RB6 (workload stable, schedule softer than 2025), James Cook at RB11 (target share growing, no obvious TD regression flag).
These shift weekly as ADP moves. The point is not to memorize the names — it is to apply the framework. Open the desk any week, sort by the regression score, and the same shape of edge appears across every round of the draft.
Build the regression score yourself
The regression framework is more useful when you can recreate it on your own roster, not just read about it. The Shark Snip workshop ships with the running-back feature panel pre-wired: opportunity-share, snap-share trend, route-participation rate, red-zone carries, target-share-among-RBs, age-curve adjustment. Drop them onto a regression head, fit on the last five seasons of weekly data, and the model produces the same opportunity-minus-efficiency score we expose in our rankings. The advantage of building it yourself is league-specific tuning — half-PPR, full-PPR, TE-premium, and superflex all change which usage features matter most, and a custom build catches that where a generic projection set cannot.
Once you have a model worth trusting, click open this projection in the lab on any RB to see his exact feature contributions sorted by magnitude. The lab also lets you stress-test a fade by holding opportunity flat and pushing efficiency back to league average; the resulting projection is usually the floor a regression candidate hits inside three weeks of the season tipping. Public RB models are also browseable in the Shark Snip marketplace if you would rather fork a finished model and tweak rather than starting from a blank canvas.
One final cross-check: regression scores by themselves do not tell you where in the draft to act on them. Pair this framework with our sibling ADP value tiers piece, which spells out the four draft zones and where consensus is most consistently wrong about RBs in each. Together, the regression score gives you the who and the value-tiers framework gives you the when.
Bottom line
RB regression is the most predictable phenomenon in fantasy. Touchdowns and yards-per-carry are noisy; snap share and route participation are not. When a back's last season was driven by the noisy stats, his ADP is a trap, even if the name on the jersey is great. When it was driven by usage, you can pay up. The math is the same every year — drafters keep paying for last season's TD luck and getting bitten by it.
Use the Shark Snip draft kit to see every RB's regression score next to their ADP and pick your spots accordingly. Cross-check the public leaderboards to confirm which regression flags have actually hit this season before you commit a top-100 pick to the fade.
Verified stat anchors and 2026 price checks
Use names as evidence, not decoration. The useful SEO win is that Jonathan Taylor, Kyren Williams, Bijan Robinson, Jahmyr Gibbs and De'Von Achane and Ravens, Lions, Rams, Chiefs and Bills appear inside decisions, thresholds, and internal links instead of being dumped into a keyword list.
Calibrate the fantasy take with real 2025 production before moving to 2026 price. StatMuse season pages list Jonathan Taylor at 1,559 rushing yards, 18 rushing TDs, and 44 receptions; Bijan Robinson at 1,478 rushing yards with 79 catches for 820 receiving yards; Jahmyr Gibbs at 1,223 rushing yards, 77 catches, and 616 receiving yards; Puka Nacua at 166 targets, 129 catches, and 1,715 receiving yards; and Amon-Ra St. Brown at 172 targets, 117 catches, 1,401 yards, and 11 receiving TDs.
- ADP rule: pay full freight only when role, team total, and contingency value all support the ceiling.
- FAAB rule: 45-70% for a real lead-RB takeover, 25-45% for a target-share breakout, 10-25% for a stable flex, 1-8% for streamers, and 0-3% for bench stashes.
- PPR tiebreaker: a Kyren Williams-style rushing profile and a Gibbs or Bijan receiving profile should not be priced the same if catches are worth a full point.
- QB rushing rule: Josh Allen and Jalen Hurts archetypes deserve separate math from pocket passers because goal-line rushing can change weekly ceiling and late-round replacement value.
Turn those names into decisions: draft, fade, trade, stash, or bid only when the 2026 price leaves room after role risk. Related workflows: fantasy ADP value tiers, target share vs air yards, FAAB strategy.
Research note board
Use this draft-room board before moving a player up or down. It keeps projection, price, and format separate.
| Decision | Check first | Example application | Do not act if |
|---|---|---|---|
| Draft | ADP, scoring format, role certainty | Jonathan Taylor at sticker price versus Kyren Williams at a discount | The room is charging for ceiling while role risk is still unresolved |
| Trade | Rest-of-season role, playoff schedule, roster need | Bijan Robinson as a need-based target instead of a generic upgrade | Both sides depend on the same fragile team environment |
| Waiver or stash | Injury-away upside, first-team reps, FAAB reserve | Jahmyr Gibbs profile compared with a short-term streamer | The move costs flexibility without adding a clear starting path |
Educational analysis only, not a bet recommendation. Check current lines, injuries, rules, contest terms, and local regulations before acting.
DFS projected ROI vs ownership %
Projected GPP ROI multiplier vs projected ownership across simulated lineups. Sub-10% leverage plays compound when they hit; chalk plays cap your upside even when the projection is dead-on.
Prop OVER hit rate vs line distance from median
Empirical hit rate of OVER bets as the prop line moves away from the player projection median, measured in standard deviations. A line set 1sd below the median hits ~84% of the time — but books price the juice to match.
