🏈 The 1921 NFL season

Every team’s rating, week by week, on one model that runs from 1920 to today. A rating is what the season looked like as it happened, not what the final table says.

21 teams · APFA · weeks 0–13 · built 2026-09-06
Jump to any season19202026

The season, week by week

One line per team. 1500 is the league average by construction.

How this is measured
Elo moves after every game by the margin and by how surprising the result was, so a rating is a running answer to “how good is this team right now” rather than a summary written at the end. Every league that ran in a season is rated in ONE pool, so the 1946-49 AAFC and the 1960-69 AFL are rated against the NFL of their day; only the standings below are split. A bye or a week after elimination inherits the previous rating and is drawn dashed, because it is a held value and not a fresh measurement.
full season
All-AmericansStaleysProsBulldogs21 of 21 shown · tap or drag any point for that week · tap a line to lock it

Tap or drag on the chart: the nearest team's rating that week.

Standings

Record, points, rating and what the season came to, grouped however you want to read it.

How this is measured
The record is the final regular-season record. The strip on each row fills in from the left as a team went further: playoffs, division, best record in its conference, conference final, championship game, championship. An asterisk on a record is the best record in the league, and a seed is the number a team carried into the playoffs. Teams level on record are ordered by the NFL tiebreaking procedure, applied to the results as they stood after each week: head-to-head, division and conference records, common games, strength of victory and schedule, points rankings and net points, in the order the league used that season. Scrub to any week and the seed column shows where each team would have been seeded had the season ended there. The playoff and title columns are odds, not records: after each week the games so far are as played and the rest of the season is drawn 2,000 times from the Elo ratings of that week (65 points to a home side), each drawn season seeded by the tiebreaking procedure and played through the bracket of the era; the title is the league championship to 1965 and the Super Bowl from 1966. "clinched" and "out" appear only where the records alone settle it, allowing for who still plays whom; a clinch that rests on a tiebreaker shows as a number, and a small number is not an elimination.
Standings21 teams
full season

Order inside a division and a conference by the NFL tiebreaking procedure; the conference view is seeding order, division winners first.

APFA

TeamW-L-TElo
ChicagoStaleysdiv128-53Staleys won its division, best record in conference, won the championship9-1-1*1632
BuffaloAll-Americans211-29All-Americans won no year-end honours9-1-21639
AkronPros148-31Pros won no year-end honours8-3-11626
CantonBulldogs106-55Bulldogs won no year-end honours5-2-31604
Rock IslandIndependents65-30Independents won no year-end honours4-2-11556
Green BayPackers70-55Packers won no year-end honours3-2-11433
EvansvilleCrimson Giants89-46Crimson Giants won no year-end honours3-21381
DaytonTriangles96-67Triangles won no year-end honours4-4-11512
ChicagoCardinals54-53Cardinals won no year-end honours3-3-21485
RochesterJeffersons85-76Jeffersons won no year-end honours2-31481
ClevelandIndians95-58Indians won no year-end honours3-51449
WashingtonSenators21-43Senators won no year-end honours1-21308
MinneapolisMarines37-41Marines won no year-end honours1-31475
HammondPros17-45Pros won no year-end honours1-3-11394
CincinnatiCelts14-117Celts won no year-end honours1-31432
DetroitHeralds19-109Heralds won no year-end honours1-5-11283
ColumbusPanhandles47-222Panhandles won no year-end honours1-81355
LouisvilleBrecks0-27Brecks won no year-end honours0-21351
MuncieFlyers0-28Flyers won no year-end honours0-21408
TonawandaKardex0-45Kardex won no year-end honours0-11442
New YorkBrickley's Giants0-72Brickley's Giants won no year-end honours0-21272
playoffsits divisionbest record in conferenceconference finalchampionship gamechampionship

The season as a shape

One box per week, week 1 at the bottom, the playoffs above the line: green a win, red a loss; bright and outlined means nobody saw it coming.

How this is measured
One row per regular-season week, stacked bottom to top from week 1 so every team's column is the same height and lines up on the same week axis; above the seam, one row per playoff round the season had, so the champion's column runs all the way up and a team that went home in the wild-card round stops there. A box's colour is win or loss; its shade is how surprising the result was, from the probability the model gave the side that actually won, banded into expected (60% or better), toss-up, and shock (under 40%, which also gets an outline and a corner dot so it still reads without colour). A bye, in the season or as a seeded rest, is an empty dashed slot; a tie is a thin grey box; a blank slot above the line is a season already over. The label under each column is the team's abbreviation in that season, so the 1999 Rams read STL. Teams are ordered by regular-season wins.
1
5
10
BUF
CHI
AKR
CAN
RCK
DAY
CRD
CLE
GB
EVA
ROC
MIN
CIN
HAM
COL
WAS
DET
TON
MUN
LOU
NEW

Tap or drag across a team's boxes to read the game: opponent, score, and what the model had given it.

Expected win (given 60% or more)Toss-up win (given 40 to 60%)Shock win (given under 40%)Expected loss (given under 40%)Toss-up loss (given 40 to 60%)Shock loss (given 60% or more)a dashed box is a bye · a grey box is a tie

The race for the title, week by week

The standings leader after each week, in a league that crowned its champion from the table, ties excluded from the percentage as the league counted it.

How this is measured
Computed from the results as they stood after each week with the tiebreaking procedure of that era (head-to-head, division and conference records, net points; common games from 1980; strength of victory and schedule and the points rankings from 2002), so the number is a position, not a forecast. E and W are East and West; CAP, CEN, CST and CTR are the 1967-69 Capitol, Century, Coastal and Central divisions. Until 1966 in the NFL and to the end in the AFL a tied division title was played off, so clubs level on the record after a week are all shown, starred, and the PO column carries the playoff that settled it: the winner filled, the loser struck through. The NFL used tiebreakers from 1967. Where a league took two clubs from each division (the 1969 AFL) the cell shows the division rank, which fixed the pairing; a single group (the 1949 AAFC) keeps its rank. Rows are the final order; the line is the cut. The scrubber above highlights its week here. In 8 weeks two clubs were level on every step the ledger can score and the club first by name was placed first; the raw file lists them.

APFA

Team151013weeks in
CHI1*1*1*1*1*1*1*119/13
BUF1*1*1*1*1*1*1*1*1110/13
AKR11*1*1*1*1*1*1*8/13
CAN0/13
RCK0/13
GB1*1/13
EVA1*1*2/13
DAY1*1/13
CRD1*1/13
ROC0/13
CLE1*1*2/13
WAS0/13
MIN0/13
HAM0/13
CIN0/13
DET1*1/13
COL0/13
LOU0/13
MUN0/13
TON0/13
NEW0/13

The best games of 1921

Ranked by Game Score, which nobody voted on.

How this is measured
Game Score is built from the two ratings going in, how close the result was, what was at stake and how far the winner was from being favourite, so a tight upset in a Super Bowl scores above a blowout in week 3 without anyone deciding it should. The same measure ranks the all-time board on the NFL hub.
1
14-0 · Reg. Season · 1921-11-24
1.31
score
2
10-7 · Reg. Season · 1921-12-04
1.17
score
3
14-0 · Reg. Season · 1921-12-03
1.15
score
4
7-6 · Reg. Season · 1921-11-24
1.08
score
5
3-0 · Reg. Season · 1921-11-20
1.03
score
6
0-0 · Reg. Season · 1921-11-13
0.97
score
7
10-0 · Reg. Season · 1921-12-11
0.95
score
8
3-0 · Reg. Season · 1921-10-23
0.84
score
9
3-0 · Reg. Season · 1921-11-13
0.79
score
10
23-0 · Reg. Season · 1921-10-09
0.78
score

Who the season changed its mind about

Rating at the end, against the rating it started with.

How this is measured
A team can gain a lot and still miss the playoffs, and a team can win its division while the model thinks less of it in January than it did in August. That gap is the season's real story and no final table contains it.
1
1550 to 1639 · peak 1652
+88
Elo
2
1546 to 1632 · peak 1641
+86
Elo
3
1520 to 1604 · peak 1610
+84
Elo
4
1300 to 1381 · peak 1381
+81
Elo
5
1507 to 1556 · peak 1566
+48
Elo
·
1478 to 1442 · peak 1478
-36
Elo
·
1470 to 1408 · peak 1470
-62
Elo
·
1414 to 1351 · peak 1414
-63
Elo
·
1503 to 1432 · peak 1513
-71
Elo
·
1444 to 1355 · peak 1444
-90
Elo

What a model that had to say so beforehand made of it

Every game of the season, priced before kick-off and scored afterwards.

How this is measured
The rating on this page is a description of what happened. Turning it into a probability BEFORE a game and then grading that probability is a different and much harder claim, and it is the one the game log makes. The comparison is against the closing betting line, which is the number to beat, and mostly it wins.

The 1921 season, priced before it happened

Open the 1921game log →

Every one of the 66 games that season carries the probability the model gave the home side before kick-off, the result, and what the betting market said. The model scored 0.1553 on Brier. No usable closing line survives from 1921, so there is nothing to score it against.

The three results it got most wrong
  • wk 10Canton Bulldogs beat Akron Pros14-0given 26%
  • wk 3Detroit Heralds beat Dayton Triangles10-7given 27%
  • wk 9Dayton Triangles beat Akron Pros3-0given 33%

Where these numbers come from

Enough to disbelieve this page on purpose rather than by accident.

Ratings are Neil Paine’s NFL Elo, carried in this site’s own NFL workbook and read here one season at a time: 48,636 team-weeks covering 1920 to 2026, with a rating and a league rank on every one. Home advantage is worth 65 rating points, which is the constant the win probabilities on this page use.

Nothing here is a forecast. The model was not held out from the seasons it rates, so read it as a description of what happened rather than a prediction of what will. Against Expectation is where the same ledger is scored against the betting market, and it loses.