Accuracy
How accurate are our projections? Every Audible projection is saved before kickoff, then graded against what the player actually scored. Sleeper's projection for the same player is graded the same way, so you can see whether we add anything. Results are published as they come in, good or bad.
How we did · Week 9
Week 9: We beat Sleeper's projections: 5.15 vs 5.32 points off.
In Week 9, closer at QB, RB and WR; behind at TE. 77% of players landed inside our floor-to-ceiling range (the target is 70%).
Biggest win over Sleeper
Drew Lock QB
We projected 13.3, Sleeper 17.3. He scored 12.6.
Biggest miss
Jordan Love QB
We projected 17.6 (Sleeper 17.1). He scored 36.2.
Season so far
Players graded
717
2026 season, projected for 5+ points and played
Audible avg. miss
5.43 pts
Sleeper: 5.49 pts on the same 717 players. Lower is better.
Range hit rate
73%
Share of outcomes between floor and ceiling. Target: 70%.
Lean
+0.37
Average projected minus actual. Positive means we run high.
Audible has been 0.05 points closer per player than Sleeper so far. Small samples swing a lot, so read this over many weeks.
By position
| position | Players | Audible error | Sleeper error | In range |
|---|---|---|---|---|
| QB | 117 | 5.99 | 6.15 | 74% |
| RB | 191 | 5.59 | 5.62 | 71% |
| TE | 133 | 4.95 | 4.89 | 74% |
| WR | 276 | 5.32 | 5.40 | 74% |
By week
| week | Players | Audible error | Sleeper error | In range |
|---|---|---|---|---|
| Week 6 | 186 | 5.41 | 5.33 | 71% |
| Week 7 | 175 | 5.47 | 5.58 | 72% |
| Week 8 | 169 | 5.74 | 5.74 | 72% |
| Week 9 | 187 | 5.15 | 5.32 | 77% |
Before launch: back-test on past seasons
These are not live predictions. They show how the model's method performed on real NFL games from 2023 to 2025 (QB, RB, WR, TE averaging 5+ points), using only information available before each game.
5.68
avg. miss using recency-weighted form (our base)
5.88 / 6.06
avg. miss using the season average / last 3 games
70%
of outcomes landed inside the floor-to-ceiling range (target 70%)
Fantasy scoring is noisy: even good projections miss by several points per player in a typical week. The back-test had no Sleeper history to compare against, which is why the live comparison above matters.
How grading works
- We grade the last projection saved before the player's own game kicked off. Later saves are never used.
- Scoring is PPR. Your league's scoring changes the numbers you see in the app, not what is graded here.
- Only players projected for 5 or more points are graded.
- Players who didn't appear in a game aren't graded, for Audible and Sleeper alike, so errors are slightly better than a full-roster view.
- "Avg. miss" is the average absolute difference between projected and actual points.