How accurate are our NFL fantasy projections? 2026 results
Across 4 graded weeks of the 2026 season, our projections ordered two players at the same position correctly 65.1% of the time, over 11,608 comparisons. The average projection missed a player’s actual half-PPR score by 5.73 points.
Start/sit calls
65.1%
share of same-position pairs we put in the right order, over 11,608 comparisons
Average miss
5.73
points between a projection and the player's actual half-PPR score, averaged over every graded projection
Projections graded
596
player-weeks measured against a real score, across 4 graded weeks
By position
| Position | Graded | Average miss | Start/sit |
|---|---|---|---|
| Quarterbacks | 118 | 6.48 | 60.4% |
| Running Backs | 171 | 5.83 | 68.3% |
| Wide Receivers | 204 | 5.68 | 64.3% |
| Tight Ends | 103 | 4.78 | 65.9% |
A dash in the start/sit column means that position is still under 200 graded comparisons, the floor below which the figure is noise.
Week by week
| Week | Graded | Average miss |
|---|---|---|
| Week 1 | 159 | 6.33 |
| Week 2 | 150 | 5.62 |
| Week 3 | 147 | 5.37 |
| Week 4 | 140 | 5.53 |
What these numbers are
Start/sit accuracy is the question a league-mate is actually arguing about: given two players at the same position, did we put the one who scored more on top? Every same-position pair in every graded week is counted. Pairs that finished level are dropped rather than scored, because a real tie is not a call we got wrong, and a pair we projected level counts as half.
Average miss is the mean absolute error: the gap between the number we published and the number the player scored, in half-PPR points, averaged over every graded projection. It is the honest measure of how close a single number lands, and it is larger than most people expect for every projection system, including this one. Fantasy scoring carries a lot of touchdown variance that nobody forecasts.
Both are computed from the same published rows the board shows, and both move every Tuesday when the week grades. A player who did not play is excluded rather than scored as a zero, so an absence never flatters or punishes the average. The line is one offensive snap, taken from the league’s snap counts rather than from whether a player turned up in a stat line: a receiver who played most of the game and was never thrown to scored zero, and we count it against ourselves. Setting that bar any higher would mean choosing a number that only ever deletes our own misses.