With only two teams on bye in Week 9 of the 2026 NFL season, fantasy managers face a wealth of options but also tough start-sit decisions. New quarterbacks such as Shedeur Sanders, Gardner Minshew, Drew Lock, and Joe Milton add intrigue, while backup running backs Kenny Gainwell (140 total yards), Tyler Allgeier (191 yards, two touchdowns), Braelon Allen (153 yards, one touchdown), and Mike Washington Jr. (83 yards) are coming off strong performances.
A proven computer model that simulates every NFL game 10,000 times has released its Week 9 rankings. The model has outperformed human experts over several seasons and correctly identified breakout players like A.J. Brown, Jonathan Taylor, Derrick Henry, Christian McCaffrey, Alvin Kamara, and Jahmyr Gibbs.
One surprise: the model is low on Carolina Panthers wide receiver Tetairoa McMillan. Despite ranking among the top 15 wide receivers in fantasy points and being started in over 80% of leagues, McMillan does not crack the top 35 at his position this week. He has been held under 60 yards in each of his last two games and has found the end zone only once this season. A difficult matchup against the Denver Broncos, who have allowed the fourth-fewest fantasy points to wide receivers, contributes to the downgrade.
Conversely, the model is high on Miami Dolphins quarterback Malik Willis, projecting him as a top-12 fantasy quarterback despite a 19% start rate last week. Willis has accounted for seven total touchdowns over his past three games after managing just two in his first four starts. He averaged the seventh-most fantasy points during that stretch and now faces a Detroit Lions defense that has allowed the most passing yards, passing touchdowns, and fantasy points to quarterbacks this season. The model ranks Willis above even Josh Allen.
The model also highlights a surprising tight end projected to finish in the top five at the position, despite barely ranking inside the top 15 in positional fantasy points this season. Full rankings for PPR and standard leagues are available from the model.
