NFL football analytics and data science, measured weekly
Seattle Seahawks lead the Elo board at 1681 after week 3. Atlanta Falcons gained 49 points in one week, Green Bay Packers gave back 49, and 6 teams are still unbeaten.
By Miles OkaforRatings updated Sep 28, 2026
since the 2026 preseason+102
NFL power ratings, week 3
Three rating systems, one table. Every number is recomputed from the game results after each week of the 2026 season.
| Rk | Team | Elo | Markov | Massey | Wk chg | Next · win prob |
|---|---|---|---|---|---|---|
| 01 | Seattle Seahawks2-1 · vs LAC | 1681 | 1.32 | +7.0 | −18 | |
| 02 | Buffalo Bills3-0 · vs NE | 1653 | 1.06 | +4.8 | +7 | |
| 03 | San Francisco 49ers3-0 · vs DEN | 1617 | 1.28 | +8.9 | +7 | |
| 04 | Minnesota Vikings3-0 · vs MIA | 1611 | 1.48 | +5.9 | +15 | |
| 05 | Denver Broncos2-1 · at SF | 1602 | 1.43 | +3.4 | +13 | |
| 06 | Jacksonville Jaguars2-1 · at CIN | 1599 | 1.16 | +10.4 | +31 | |
| 07 | Los Angeles Rams1-2 · at PHI | 1578 | 0.90 | +2.9 | −13 | |
| 08 | New England Patriots1-2 · at BUF | 1577 | 1.00 | +1.9 | −31 | |
| 09 | Philadelphia Eagles2-0 · at CHI | 1574 | 0.57 | -0.2 | 0 | |
| 10 | Kansas City Chiefs3-0 · at LV | 1558 | 1.35 | +5.9 | +23 | |
| 11 | Detroit Lions2-1 · at CAR | 1556 | 1.09 | +0.3 | +8 | |
| 12 | Baltimore Ravens2-1 · vs TEN | 1551 | 1.35 | +2.7 | +11 | |
| 13 | Chicago Bears1-1 · vs PHI | 1537 | 1.00 | +5.8 | 0 | |
| 14 | Houston Texans0-3 · vs DAL | 1534 | 0.88 | -1.9 | −13 | |
| 15 | Cincinnati Bengals2-1 · vs JAX | 1527 | 0.93 | +1.8 | −12 | |
| 16 | Pittsburgh Steelers2-1 · at CLE | 1521 | 0.77 | -1.6 | +12 | |
| 17 | Green Bay Packers1-2 · at TB | 1471 | 0.52 | -7.3 | −49 | |
| 18 | Atlanta Falcons1-2 · at NO | 1461 | 0.47 | -5.2 | +49 | |
| 19 | Dallas Cowboys1-2 · at HOU | 1456 | 1.14 | -0.4 | −11 | |
| 20 | Las Vegas Raiders3-0 · vs KC | 1453 | 1.19 | +1.8 | +31 | |
| 21 | Washington Commanders1-2 · vs IND | 1453 | 1.10 | -2.7 | +18 | |
| 22 | Cleveland Browns2-1 · vs PIT | 1448 | 0.59 | -3.3 | +12 | |
| 23 | Indianapolis Colts1-2 · vs WAS | 1447 | 1.25 | -3.1 | +13 | |
| 24 | Carolina Panthers1-2 · vs DET | 1443 | 0.51 | -1.0 | −12 | |
| 25 | New Orleans Saints1-2 · vs ATL | 1442 | 1.54 | -0.4 | −31 | |
| 26 | New York Giants2-1 · vs ARI | 1441 | 0.80 | -4.3 | +9 | |
| 27 | Los Angeles Chargers0-3 · at SEA | 1440 | 0.90 | -5.4 | −7 | |
| 28 | Tampa Bay Buccaneers0-3 · vs GB | 1430 | 1.05 | -2.9 | −15 | |
| 29 | Arizona Cardinals1-2 · at NYG | 1416 | 1.43 | -3.2 | −7 | |
| 30 | Miami Dolphins0-3 · at MIN | 1396 | 0.71 | -7.3 | −22 | |
| 31 | New York Jets1-2 · at CHI | 1376 | 0.64 | -4.7 | −8 | |
| 32 | Tennessee Titans0-3 · at BAL | 1311 | 0.60 | -8.7 | −9 |
01Seahawks1681
- Record
- 2-1
- Wk chg
- −18
- Markov
- 1.32
- Massey
- +7.0
- Next
- vs LAC · 85%
02Bills1653
- Record
- 3-0
- Wk chg
- +7
- Markov
- 1.06
- Massey
- +4.8
- Next
- vs NE · 69%
0349ers1617
- Record
- 3-0
- Wk chg
- +7
- Markov
- 1.28
- Massey
- +8.9
- Next
- vs DEN · 61%
04Vikings1611
- Record
- 3-0
- Wk chg
- +15
- Markov
- 1.48
- Massey
- +5.9
- Next
- vs MIA · 83%
05Broncos1602
- Record
- 2-1
- Wk chg
- +13
- Markov
- 1.43
- Massey
- +3.4
- Next
- at SF · 39%
06Jaguars1599
- Record
- 2-1
- Wk chg
- +31
- Markov
- 1.16
- Massey
- +10.4
- Next
- at CIN · 51%
07Rams1578
- Record
- 1-2
- Wk chg
- −13
- Markov
- 0.90
- Massey
- +2.9
- Next
- at PHI · 41%
08Patriots1577
- Record
- 1-2
- Wk chg
- −31
- Markov
- 1.00
- Massey
- +1.9
- Next
- at BUF · 31%
09Eagles1574
- Record
- 2-0
- Wk chg
- 0
- Markov
- 0.57
- Massey
- -0.2
- Next
- at CHI · 46%
10Chiefs1558
- Record
- 3-0
- Wk chg
- +23
- Markov
- 1.35
- Massey
- +5.9
- Next
- at LV · 56%
Elo: K 20, home field 65 points, margin multiplier, one third regression to 1505 between seasons. Markov: share of a damped random walk from losers to winners, 1.00 = league average. Massey: points better than an average team on a neutral field, season to date.
Five football analytics models, taken apart
Elo ratings
One number per team, nudged after every game by how surprising the result was.
Read the model SHOT 02Markov chain ratings
A random walk that drifts toward whoever beats the teams that beat everyone else.
Read the model SHOT 03Win probability
Score, clock, down, distance and field position turned into one live percentage.
Read the model SHOT 04Expected points
What a first down at the opponent's 35 is worth before the snap, and what EPA measures after it.
Read the model SHOT 05Clustering tiers
k-means groups players whose projections overlap, so tiers come from the data, not from rank order.
Read the modelRun the numbers yourself
Win probability calculator
Set the game state. The estimate comes from a logistic model fitted on 324,950 NFL plays (2016–2024).
Watch the math move
How an Elo rating moves after one game
Expected score, K-factor and the margin multiplier, one step at a time.
How a Markov chain ranks 32 teams
A voter wanders from loser to winner until the league settles into a ranking.
How k-means sorts running backs into tiers
Centroids move, players switch groups, and the tiers stop changing.
From the data science notebook
NFL data science, done in the open
The spread is a football analytics lab for NFL data science, built on open data. Every rating on this page is recomputed from the game results after each week, and every one of the football analytics models behind it is written up from the formula outward: Elo, Massey least squares, Markov chains, win probability and expected points.
NFL data science has a reproducibility problem: most public numbers arrive without the code or the constants that made them. Here the constants sit on the methodology page, the backtest shows how the ratings would have done in every season since 2000, and the calculators run the same football analytics models in your browser. Game and play-by-play data come from nflverse under a CC-BY 4.0 license.