JD
LOWFI LOWFI

History

Every team, every week, since 2004. Pick a rating, hover a line to read it, and click one to pin it.

Every Team, Every Week

Leave the team list empty to show all 32, or hold Cmd or Ctrl to pick a few.

League Scoring

Best and Worst

LOWFI/1 NET

BEST WORST
End of season
NE 2007 (+12.85)
LA 2009 (−9.66)
Season average
NE 2007 (+11.16)
LA 2009 (−7.98)
Single week
NE 2007, week 11 (+13.41)
LA 2009, week 16 (−9.66)

LOWFI/1 OFFENSE

BEST WORST
End of season
NE 2007 (+9.51)
LA 2011 (−6.97)
Season average
DEN 2013 (+8.89)
LA 2009 (−5.81)
Single week
DEN 2013, week 8 (+10.63)
BUF 2018, week 9 (−7.12)

LOWFI/1 DEFENSE

BEST WORST
End of season
SEA 2013 (+4.37)
DET 2008 (−5.32)
Season average
NE 2019 (+3.12)
SF 2005 (−3.59)
Single week
SEA 2013, week 15 (+4.44)
DET 2008, week 16 (−5.32)

LOWFI/0 TRUE STRENGTH

BEST WORST
End of season
NE 2007 (84.7%)
CLE 2017 (13.5%)
Season average
NE 2007 (75.5%)
CLE 2017 (21.5%)
Single week
NE 2007, week 17 (84.7%)
CLE 2017, week 17 (13.5%)

Take the Data

The full database - 12,960 rows, one per team per week since 2004 - lives in the LOWFI data repo, with a data dictionary and the retrodiction caveat attached to it.

Notes

Retrodiction, not a track record

Most of this page is retrodiction, not a track record. LOWFI didn't exist in 2011. Every completed season here is today's model read backwards over a season that's already been played, revealed to it a week at a time - so it answers "how does LOWFI read 2011," not "what did LOWFI say during 2011." For 2023 through 2025 it's reading its own training data.

The shaded stretch at the right is 2026, and that part is live: every point is what that Tuesday's run produced, before the games.

League scoring

Average points per team per game, straight from the schedule. No model here - just what happened, for context behind the ratings above.

Best and worst

Each rating gets three readings. End of season and season average use completed seasons only - where a team finished, and how it looked across the whole year, which can be very different for a team that collapsed in December. Single week is the highest and lowest weekly estimate anywhere in the database, live weeks included.