Scores View all
FRI 6:00 PM ET
ISU
UCF
—
—
FRI 6:30 PM ET
BROC
MS
—
—
FRI 7:30 PM ET
CLEM
DUKE
—
—
FRI 7:30 PM ET
BGSU
TOL
—
—
FRI 8:00 PM ET
UTEP
AF
—
—
FRI 8:00 PM ET
ORE
MSU
—
—
FRI 9:00 PM ET
UNM
WYO
—
—
SAT 12:00 PM ET
LIU
RM
—
—
Football players in action under stadium lights, illustrating FSN's Week 8 DFS forecast review.
NFL 2026 NFL Season

FSN DFS Week 8 Report Card: Injuries and Breakout Misses

FSN reviews 264 DraftKings and 262 FanDuel matched Week 8 forecasts, separating injury-driven role changes, breakout misses and close calls.

FSN newsroom article

Produced through an FSN-managed editorial workflow using sourced reporting, research and publication tools; subject to FSN editorial and corrections standards. How FSN uses AI & automation Corrections

FSN Football Sports News1852 words · 9 min read
Share X f ✉
In this story9 sections
  1. Week 8 DFS accuracy: the matched-sample results
  2. Quarterbacks: Baltimore’s injuries belong in the explanation, not outside the sample
  3. Running backs: Allgeier’s expanded role and Johnson’s explosive receiving day
  4. Receivers: Mitchell and Tucker show why a game environment is not a player forecast
  5. Tight ends: separate useful forecasts from touchdown-driven departures
  6. The six-point return that could not be treated as zero
  7. Range coverage: useful evidence, not a guarantee
  8. Methodology and coverage limits
  9. FSN takeaway

Tyler Allgeier’s saved FSN forecast was 6.47 DraftKings points. His Week 8 result was 39.10. The size of that miss is easy to measure. Explaining it requires more than calling the player a surprise.

Arizona’s official account of Allgeier’s performance describes a backfield reshaped by injuries to Jeremiyah Love and Bam Knight. Allgeier finished with 147 rushing yards, 44 receiving yards and two touchdowns. In Buffalo, Baltimore lost two quarterbacks during the game, turning several substantial projections into very small results.

The central lesson of FSN’s Week 8 DFS report card is that the same point error can come from very different failures or events. A role that changes after kickoff, an explosive play and an overstated expectation should remain in the accuracy calculation. They should not automatically receive the same explanation.

Scope: This retrospective evaluates older archived pre-kickoff forecasts, not today’s model. It covers 264 matched DraftKings records and 262 FanDuel records, including one independently reconciled return-touchdown score per platform. Exact equivalence between these legacy records and every forecast displayed by the public builder was not recorded.

Week 8 DFS accuracy: the matched-sample results

2026 Week 8 offensive forecasts, including the documented return-touchdown reconciliation
Platform Records Mean absolute error Bias: actual minus forecast Root mean square error
DraftKings 264 5.36 -0.27 7.5
FanDuel 262 4.89 -0.23 6.9

Errors are fantasy points. Mean absolute error measures the average size of a miss; signed bias measures whether forecasts generally ran high or low. Root mean square error gives large misses more influence. Figures are rounded; the reconciliation and calculation limits are explained below.

The small negative biases mean forecasts ran slightly high on average. That is not the same as being consistently close. Underestimates and overestimates offset one another in the bias calculation, while absolute error preserves both. A lineup does not recover from one bad selection simply because the model missed another player in the opposite direction.

This is an equally weighted comparison of matched player records, including low-usage players. It is not a salary-weighted evaluation, a main-slate-only report or evidence of contest profitability. The platforms have different scoring rules and slightly different samples, so the smaller FanDuel error is not proof of a better model.

Quarterbacks: Baltimore’s injuries belong in the explanation, not outside the sample

The Ravens’ official recap documents Lamar Jackson’s thumb injury on Baltimore’s first offensive play and Tyler Huntley’s toe injury in the second quarter. Their exits forced the offense to its third quarterback.

The archive had Jackson at 17.86 points on both platforms. His result was zero. Huntley’s 16.65-point forecasts became minus 1.00 on DraftKings and minus 2.00 on FanDuel after a lost fumble. Those results stay in the audit. Removing injury-affected misses afterward would create a more flattering but less representative sample.

They do warrant a different interpretation from a healthy quarterback playing a full game and throwing poorly. The legacy records also contain substantial forecasts for both Baltimore quarterbacks; the records alone do not establish the participation assumptions behind them. That deserves review without pretending we have already diagnosed the old model’s cause.

Josh Allen supplies a useful contrast from the same game. The official box score records 289 passing yards, two passing touchdowns, one interception, 30 rushing yards and a rushing touchdown. His 27.56-point result on either platform was close to the saved forecasts: 26.97 on DraftKings and 26.94 on FanDuel.

Quarterbacks nevertheless had the largest raw-point mean absolute error: about 8.1 DraftKings points and 8.2 FanDuel points across 29 records each. Those samples include backups, not just 29 starting-quarterback performances.

Running backs: Allgeier’s expanded role and Johnson’s explosive receiving day

Allgeier produced the largest underestimate in the matched sample: 32.63 points above his DraftKings forecast and 31.23 above his FanDuel forecast. His saved FanDuel projection was 5.37; his calculated result was 36.60.

Arizona’s report puts the performance in context: 11 carries, five receptions and 48 offensive snaps after the other backs left injured. His DraftKings ceiling was 10.49 and FanDuel ceiling 9.10. The outcome exceeded both by a wide margin.

That does not justify rewriting his pregame forecast to assume injuries that had not happened. It does make contingency handling a useful research question: how much probability should a backup’s distribution place on a sudden workload expansion? One breakout cannot answer that, but a documented archive can help test it over time.

Emmett Johnson’s miss came through a different route. The Chiefs-Broncos box score records only one carry for 15 yards, but five catches for 128 yards and a touchdown. His six offensive touches generated 28.30 DraftKings points and 25.80 FanDuel points against forecasts of 3.72 and 2.90.

This was not a 25-carry takeover. It was a highly productive receiving performance, including a 73-yard touchdown. The modeling question is about receiving opportunity and explosive outcomes, rather than simply raising expected rushing volume after a big fantasy score.

Jahmyr Gibbs was much closer to expectation. His official game log shows 23 carries for 153 yards and a touchdown, plus two catches for 15 yards. He scored 27.80 DraftKings points against 28.83 projected, and 26.80 FanDuel points against 26.32 projected. These are close final-score forecasts; they do not independently prove that every workload component was predicted accurately.

Receivers: Mitchell and Tucker show why a game environment is not a player forecast

Adonai Mitchell and Tre Tucker played in the same Raiders-Jets game, but their comparisons went in opposite directions. The official NFL gamebook credits Mitchell with seven receptions for 142 yards and a touchdown. Tucker finished without a catch on one target.

Mitchell’s 6.00-point DraftKings forecast became 30.20; his 4.71-point FanDuel forecast became 26.70. Tucker’s forecasts were 20.82 and 18.27, respectively, against zero on each platform.

That contrast puts player-level opportunity under scrutiny. Projecting a productive game does not establish which receiver will collect its points. The archive does not supply enough evidence to attribute Tucker’s miss to a particular coverage choice or coaching decision, so the result should identify a review target rather than manufacture an explanation.

Zay Flowers was another major overprediction: 22.36 DraftKings points and 18.71 FanDuel points against zero. Baltimore’s recap confirms his catchless game. The quarterback injuries provide important context, but they do not make the receiver error disappear. They also illustrate why several player misses from one offense are not independent pieces of evidence.

Tight ends: separate useful forecasts from touchdown-driven departures

Brock Bowers caught seven passes for 92 yards and two touchdowns in the Raiders-Jets gamebook. That produced 28.20 DraftKings points against a 17.64 forecast. His productive result was still an underestimate of 10.56 points, not an automatic model success because he was a good fantasy selection.

Travis Kelce offered a closer comparison. His five catches for 67 yards in the Chiefs-Broncos box score produced 11.70 DraftKings points against 9.52 projected, and 9.20 FanDuel points against 7.49 projected. The final errors were 2.18 and 1.71 points.

The position-level table helps keep those examples in proportion. Tight ends had the smallest absolute errors in raw points, but lower-scoring positions can produce smaller point errors without being more accurate relative to their typical output.

Mean absolute error by position, rounded to one decimal point
Position DraftKings records DraftKings error FanDuel records FanDuel error
QB 29 8.1 29 8.2
RB 73 5.4 71 5.2
WR 107 5.1 107 4.3
TE 55 4.4 55 3.8

The six-point return that could not be treated as zero

Dylan Laube’s offensive stat line did not tell his scoring story. The Raiders confirmed his 95-yard kickoff-return touchdown, and the official gamebook establishes the return score without additional offensive production or a lost fumble. Applying the DraftKings Classic rules and FanDuel’s NFL scoring rules gives him 6.00 points on each platform.

We included that independently calculated total against forecasts of 2.48 DraftKings points and 2.02 FanDuel points. The errors are plus 3.52 and plus 3.98. Return yardage itself was not added as rushing or receiving yardage.

This is an editorial reconciliation, not a claim that a platform certified the total or that the database’s review flag was cleared. The automated reader excluded this exceptional scoring record from its full-score metrics; this article adds it separately and documents the difference.

Range coverage: useful evidence, not a guarantee

After that reconciliation, 109 of 263 DraftKings results with saved ranges landed inside them, or 41.44%. Another 106 were below the floor and 48 above the ceiling. FanDuel had 102 inside out of 261, or 39.08%, with 106 below and 53 above.

The saved boundaries are labeled the 25th and 90th percentiles. Their nominal interval spans 65 percentage points, not a promise that a player cannot score outside it. The observed coverage merits investigation, but one incomplete, correlated weekly sample cannot establish long-run calibration. Old ranges must remain unchanged while additional weeks are evaluated.

Methodology and coverage limits

We inspected all five archive pages for each platform. Each contained 425 timestamped pre-kickoff records, including 28 defensive entries outside this offensive audit. Of 397 offensive forecasts per platform, 264 DraftKings records matched results, or 66.50%; 262 FanDuel records matched, or 65.99%.

The DraftKings exclusions were 132 unavailable statistical matches and one unresolved identity. FanDuel excluded 129 unavailable matches and six unresolved identities. Missing rows were never converted to zero. Explicit zero or negative scores in matched rows remained in the sample.

Ordinary actuals use FSN’s imported statistics and full-component scoring calculations. Official team sources and the NFL gamebook were checked for the examples above; every ordinary row was not independently reconciled against settled contest totals. Later statistical corrections can change results.

For transparency, the reader’s source-complete subsets before adding Laube were 263 DraftKings records and 261 FanDuel records. We combined their reported three-decimal aggregates with Laube’s verified scoring components. The displayed precision is deliberately limited so rounding uncertainty in those input aggregates does not change the published figures.

Forecasts are saved means, not medians. Internal capture times are not independent external attestations. Current values on the Week 8 DFS builder were not substituted for archived forecasts. This is a retrospective with its real publication date, not a backdated prediction article.

FSN takeaway

Week 8 points toward three separate review priorities: participation assumptions, distributions that represent abrupt role changes, and player-level receiving opportunity. Those are hypotheses to test, not fixes proven by one week’s outcomes.

Gibbs and Allen show that close forecasts belonged in the same archive as Allgeier’s breakout and Baltimore’s injury-affected misses. Keeping both is the point. The goal is a record that helps improve the process, not a highlight reel that makes the model look infallible.

Continue with the Week 9 DFS report card, or use the FSN NFL DFS Lineup Builder for current research. Different weekly player pools and exclusions mean these reports should not be mistaken for a controlled before-and-after model test.

NFL data via nflverse (CC BY 4.0) · College data via CollegeFootballData.com