Power Rankings
The 2026-2027 draft has not happened yet, so there are no rosters to rank. This page fills in as soon as the last pick is in — it reads the draft straight from Sleeper, the same way the live board does.
What it will show, and how it gets there, is below.
Method — what this will measure, and how well it works
Method
Three things go into a player's projection, and ADP is not one of them.
- Usage, not points. A ridge regression per position maps last season's per-game volume — targets, carries, air yards, target share, WOPR — onto this season's points per game. Volume survives the offseason; points carry touchdown luck that does not. Held out on the two most recent seasons, that beats projecting last year's points forward at receiver and tight end and ties at running back.
- Draft capital for rookies. A rookie has no usage to read, so the prior is where the NFL drafted him, fit against what drafted rookies have actually scored since 2021. That is 32 front offices spending picks, not a fantasy consensus.
- Nothing for kickers and defenses. The same model fit to K and DST has negative held-out skill: last season's kicker points do not predict this season's at any amount of regularization. Both positions get the positional mean, so they cancel out of the rankings entirely.
Each projection carries its own error bar, and every simulated season deals each player a true rate drawn from it. Without that step the page would quote playoff odds far more confident than a projection this uncertain can support.
How well does it work? Player by player: well.
| 2023 | 2024 | 2025 | Average | |
|---|---|---|---|---|
| Position | ||||
| QB | +0.42 | +0.41 | +0.31 | +0.38 |
| RB | +0.56 | +0.61 | +0.70 | +0.62 |
| WR | +0.67 | +0.57 | +0.58 | +0.61 |
| TE | +0.43 | +0.41 | +0.58 | +0.47 |
Correlation between a player's projected points per game and what he actually scored, for every season the league has played, with the model given only the years before each one. Running backs and receivers land around +0.6 every year; quarterbacks and tight ends around +0.4. That is the part of this page that demonstrably works.
Roster by roster: unproven.
Run the whole page on a past season's draft-day rosters, with only the seasons before it to learn from, and compare its ranking to what actually happened over the fourteen weeks that followed.
Rank correlation between this page's projected finish and points actually scored: +0.42.
| Manager | Projected | Proj. Points | Actual | Actual Points | Wins |
|---|---|---|---|---|---|
| Trevor | 1 | 1,782 | 4 | 1,886 | 17 |
| George | 2 | 1,768 | 3 | 1,891 | 15 |
| Everett | 3 | 1,750 | 10 | 1,658 | 5 |
| Max | 4 | 1,748 | 1 | 1,983 | 19 |
| Jackson | 5 | 1,744 | 2 | 1,943 | 19 |
| Tyler | 6 | 1,731 | 5 | 1,872 | 14 |
| Austin | 7 | 1,730 | 8 | 1,776 | 16 |
| Colin | 8 | 1,723 | 7 | 1,793 | 10 |
| Padgett | 9 | 1,722 | 9 | 1,752 | 8 |
| Mark | 10 | 1,646 | 6 | 1,834 | 17 |
Across 3 seasons those correlations average +0.12, on ten teams a year. In other words: at the level of a whole roster, over the sample this league has actually played, this page has not demonstrated that it can pick the season's best team from the draft. Thirty team-seasons cannot tell a good model from a coin flip, and a roster's points are mostly decided after the draft — by waivers, by injuries, and by who each manager benched on the wrong week. Read the rankings as a description of what was drafted, which they measure well, rather than as a forecast of what will happen.