Model accuracy

Does the number hold up?

Every 2025 game replayed with the rating each team carried before that game — the same calculation the site shows for Week 1, with no hindsight. Then the 2026 season, tracked game by game as results land.

Tuning

Fitted to 2025, not to a textbook

The model started with the usual Elo defaults. Each constant was swept against the 272 stored 2025 games, fit on Weeks 1-13 and validated on Weeks 14-18 — a setting only survives if it also helps on games it was not fit to. Log loss is the objective; anything worth less than 0.002 was left at its round-number default.

Straight-up

61.8% → 64.0%

held-out weeks 64.1% → 69.2%

Brier score

0.226 → 0.222

held-out weeks 0.225 → 0.216

Log loss

0.644 → 0.634

held-out weeks 0.637 → 0.618

Constant	Was	Now	Held-out LL
K-factor With the weekly decay in place a slightly faster K grades best; 24-32 are all inside a point of each other.	26	28	-0.0004
Home field Home teams went 142-123 (.536) in 2025 — worth roughly 25 rating points, not the 55 the textbook default assumes.	35 pts	25 pts	-0.0006
In-season decay Ratings bleed 10% back toward 1500 every week, so September results stop carrying December games.	none	10% per week	-0.0028
Scoring blend Mixing season point differential into the pre-kickoff rating was the biggest single gain — Elo grades wins, this remembers how they were won.	0%	25% after 4 games	-0.0015
MOV dampener Flat from 1.6 to 3.0 — the blowout dampener barely moves the grade. Left alone.	2.2	2.2	kept
Spread divisor 22 cut margin error from 10.265 to 10.255 points per game. Noise, so the round number stays.	25	25	kept
Offseason regression 30-35% graded a hair better on the in-season proxy, inside the noise guard.	25%	25%	kept
Blend weights Elo already prices margin of victory, so the scoring terms were double-counting it.	55 / 20 / 15 / 10	80 / 5 / 15 / 0	-0.0059

Honest read: the change that mattered was giving the pre-kickoff rating a memory of scoring margin, not just wins, and letting old results fade 10% a week. Trimming home field to 25 points — home teams went 142-123 (.536) in 2025 — helped at the edges. Together they lifted the 2025 backtest from 62.5% to 64.0% straight up and cut the Brier score from 0.224 to 0.222. That is real but modest; NFL Elo does not have much more room than this, and every number here is fit on the same season it is graded on.

Straight-up

64.0%

174 of 272 winners called in 2025

Brier score

0.222

Lower is better. 0.25 = always guessing 50/50

Avg spread error

10.3 pts

44% of games within a touchdown

Log loss

0.635

Penalises confident wrong calls hardest

Accuracy by confidence band

The bands on every matchup card, judged against reality. A "Strong" call should win far more often than a "Lean".

Coin flip

57%

54-40 · claimed 53%

Lean

62%

58-35 · claimed 59%

Confident

66%

37-19 · claimed 70%

Strong

86%

25-4 · claimed 79%

Calibration

When the model says 70%, does it win 70% of the time?

50–55% favorites94 games · said 53% · won 57%

55–60% favorites64 games · said 57% · won 58%

60–65% favorites29 games · said 62% · won 72%

65–70% favorites31 games · said 68% · won 58%

70–75% favorites25 games · said 73% · won 76%

75–80% favorites21 games · said 78% · won 81%

80–100% favorites8 games · said 84% · won 100%

Grey bar: what the model claimed. Coloured bar: what actually happened. A well-calibrated model has them roughly equal in every band.

2026 season tracker

66 of 107 correct · 62% · 14 still to play

Straight-up

62%

66 of 107 final games

Brier score

0.233

2025 backtest ran 0.222

Avg spread error

9.5 pts

2025 backtest ran 10.3 pts

Week	Game	Final	Model pick	Spread	Result
Wk 7	DAL at PHI	38–21	PHI 53%	PHI −0.9	Miss
Wk 7	KC at SEA	3–20	SEA 53%	SEA −0.9	Hit
Wk 7	GB at DET	17–20	DET 55%	DET −1.3	Hit
Wk 7	LAR at LV	35–21	LAR 50%	LAR −0.1	Hit
Wk 7	DEN at ARI	30–16	DEN 67%	DEN −4.9	Hit
Wk 7	CLE at TEN	27–24	CLE 60%	CLE −2.7	Hit
Wk 7	NYG at HOU	27–21	HOU 61%	HOU −3.1	Miss
Wk 7	MIA at NYJ	28–25	NYJ 56%	NYJ −1.6	Miss
Wk 7	TB at CAR	43–13	CAR 66%	CAR −4.7	Miss
Wk 7	CIN at BAL	35–13	BAL 51%	BAL −0.4	Miss
Wk 7	IND at MIN	10–16	MIN 78%	MIN −8.7	Hit
Wk 7	SF at ATL	20–29	SF 67%	SF −4.9	Miss
Wk 7	PIT at NO	16–10	PIT 61%	PIT −3.1	Hit
Wk 7	NE at CHI	24–30	NE 54%	NE −1.1	Miss
Wk 6	WSH at SF	17–23	SF 77%	SF −8.2	Hit
Wk 6	DAL at GB	13–23	GB 59%	GB −2.5	Hit
Wk 6	LAC at KC	6–22	KC 75%	KC −7.6	Hit
Wk 6	BUF at LV	33–30	BUF 62%	BUF −3.4	Hit
Wk 6	ARI at LAR	18–16	LAR 70%	LAR −6.0	Miss
Wk 6	TEN at IND	26–20	IND 66%	IND −4.7	Miss
Wk 6	BAL at CLE	10–13	BAL 58%	BAL −2.2	Miss
Wk 6	NO at NYG	34–27	NYG 64%	NYG −3.9	Miss
Wk 6	CAR at PHI	25–14	CAR 57%	CAR −1.9	Hit
Wk 6	PIT at TB	26–19	PIT 54%	PIT −1.2	Hit
Wk 6	NYJ at NE	19–26	NE 71%	NE −6.1	Hit
Wk 6	CHI at ATL	31–23	ATL 53%	ATL −0.9	Miss
Wk 6	HOU at JAX	34–31	JAX 65%	JAX −4.4	Miss
Wk 6	SEA at DEN	14–17	DEN 50%	DEN −0.0	Hit
Wk 5	BUF at LAR	37–34	LAR 60%	LAR −2.8	Miss
Wk 5	BAL at ATL	30–23	ATL 68%	ATL −5.3	Miss
Wk 5	SF at SEA	32–29	SEA 57%	SEA −1.9	Miss
Wk 5	DET at ARI	45–21	DET 55%	DET −1.5	Hit
Wk 5	DEN at LAC	23–17	DEN 65%	DEN −4.4	Hit
Wk 5	MIN at NO	23–13	MIN 68%	MIN −5.2	Hit
Wk 5	CLE at NYJ	12–17	CLE 65%	CLE −4.3	Miss
Wk 5	NYG at WSH	27–27	NYG 61%	NYG −3.0	Miss
Wk 5	CIN at MIA	31–24	CIN 64%	CIN −4.1	Hit
Wk 5	IND at PIT	22–34	PIT 55%	PIT −1.5	Hit
Wk 5	HOU at TEN	26–3	HOU 54%	HOU −1.2	Hit
Wk 5	LV at NE	17–45	NE 53%	NE −1.0	Hit
Wk 5	CHI at GB	7–31	GB 55%	GB −1.3	Hit
Wk 5	PHI at JAX	17–19	JAX 75%	JAX −7.6	Hit
Wk 5	TB at DAL	24–16	DAL 77%	DAL −8.4	Miss
Wk 4	ATL at NO	45–24	ATL 53%	ATL −0.9	Hit
Wk 4	DET at CAR	26–32	DET 52%	DET −0.5	Miss
Wk 4	DEN at SF	14–24	SF 60%	SF −2.9	Hit
Wk 4	KC at LV	30–27	LV 53%	LV −0.7	Miss
Wk 4	LAC at SEA	23–30	SEA 81%	SEA −10.3	Hit
Wk 4	MIA at MIN	10–15	MIN 75%	MIN −7.6	Hit
Wk 4	GB at TB	17–14	GB 64%	GB −4.0	Hit
Wk 4	DAL at HOU	34–30	HOU 58%	HOU −2.3	Miss
Wk 4	NE at BUF	29–26	BUF 58%	BUF −2.1	Miss
Wk 4	TEN at BAL	18–24	BAL 64%	BAL −4.1	Hit
Wk 4	LAR at PHI	24–20	LAR 57%	LAR −2.1	Hit
Wk 4	NYJ at CHI	12–23	CHI 57%	CHI −2.0	Hit
Wk 4	JAX at CIN	22–17	JAX 60%	JAX −2.8	Hit
Wk 4	ARI at NYG	24–36	NYG 61%	NYG −3.3	Hit
Wk 4	IND at WSH	30–13	IND 69%	IND −5.7	Hit
Wk 4	PIT at CLE	24–27	CLE 51%	CLE −0.4	Hit
Wk 3	PHI at CHI	7–27	PHI 61%	PHI −3.2	Miss
Wk 3	LAR at DEN	26–30	DEN 62%	DEN −3.4	Hit
Wk 3	BAL at DAL	34–31	BAL 59%	BAL −2.7	Hit
Wk 3	LV at NO	35–27	NO 58%	NO −2.4	Miss
Wk 3	MIN at TB	23–16	MIN 65%	MIN −4.3	Hit
Wk 3	ARI at SF	30–36	SF 65%	SF −4.3	Hit
Wk 3	NE at JAX	6–35	JAX 58%	JAX −2.2	Hit
Wk 3	CAR at CLE	18–21	CLE 54%	CLE −1.0	Hit
Wk 3	CIN at PIT	27–30	CIN 55%	CIN −1.3	Miss
Wk 3	KC at MIA	24–10	KC 65%	KC −4.2	Hit
Wk 3	SEA at WSH	31–33	SEA 85%	SEA −12.1	Miss
Wk 3	LAC at BUF	16–24	BUF 79%	BUF −9.1	Hit
Wk 3	NYJ at DET	24–31	DET 65%	DET −4.2	Hit
Wk 3	TEN at NYG	7–12	NYG 60%	NYG −2.9	Hit
Wk 3	HOU at IND	17–19	HOU 55%	HOU −1.3	Miss
Wk 3	ATL at GB	35–14	GB 62%	GB −3.4	Miss
Wk 2	NYG at LAR	6–28	LAR 51%	LAR −0.2	Hit
Wk 2	IND at KC	30–33	KC 69%	KC −5.5	Hit
Wk 2	SEA at ARI	31–7	SEA 55%	SEA −1.3	Hit
Wk 2	WSH at DAL	20–37	DAL 66%	DAL −4.6	Hit
Wk 2	MIA at SF	13–35	SF 76%	SF −8.1	Hit
Wk 2	JAX at DEN	13–20	JAX 61%	JAX −3.2	Miss
Wk 2	LV at LAC	26–14	LAC 55%	LAC −1.3	Miss
Wk 2	PHI at TEN	24–20	PHI 63%	PHI −3.8	Hit
Wk 2	NO at BAL	24–17	BAL 53%	BAL −0.9	Miss
Wk 2	CIN at HOU	20–6	HOU 52%	HOU −0.5	Miss
Wk 2	CLE at TB	23–19	TB 52%	TB −0.5	Miss
Wk 2	PIT at NE	3–20	NE 58%	NE −2.2	Hit
Wk 2	GB at NYJ	20–17	NYJ 55%	NYJ −1.5	Miss
Wk 2	MIN at CHI	9–3	CHI 67%	CHI −4.8	Miss
Wk 2	CAR at ATL	34–3	CAR 55%	CAR −1.4	Hit
Wk 2	DET at BUF	31–41	BUF 66%	BUF −4.5	Hit
Wk 1	DEN at KC	10–31	DEN 66%	DEN −4.5	Miss
Wk 1	DAL at NYG	20–28	NYG 60%	NYG −2.8	Hit
Wk 1	WSH at PHI	22–24	PHI 58%	PHI −2.3	Hit
Wk 1	GB at MIN	22–39	MIN 65%	MIN −4.3	Hit
Wk 1	ARI at LAC	26–14	LAC 69%	LAC −5.6	Miss
Wk 1	MIA at LV	13–27	MIA 55%	MIA −1.4	Miss
Wk 1	BAL at IND	41–23	BAL 56%	BAL −1.6	Hit
Wk 1	NO at DET	30–31	DET 53%	DET −0.8	Hit
Wk 1	CHI at CAR	59–37	CHI 54%	CHI −1.2	Hit
Wk 1	BUF at HOU	36–31	HOU 57%	HOU −1.8	Miss
Wk 1	NYJ at TEN	23–10	TEN 58%	TEN −2.3	Miss
Wk 1	TB at CIN	27–33	CIN 59%	CIN −2.5	Hit
Wk 1	CLE at JAX	10–34	JAX 72%	JAX −6.5	Hit
Wk 1	ATL at PIT	13–20	PIT 54%	PIT −1.2	Hit
Wk 1	SF at LAR	27–7	SF 51%	SF −0.1	Hit
Wk 1	NE at SEA	10–13	SEA 59%	SEA −2.5	Hit

Worst calls of 2025

The games the model was most sure about and still got wrong.

Week	Game	Final	Model pick	Result
Wk 5	TEN at ARI	22–21	ARI 79%	Miss
Wk 13	CIN at BAL	32–14	BAL 78%	Miss
Wk 18	WSH at PHI	24–17	PHI 76%	Miss
Wk 13	LAR at CAR	28–31	LAR 76%	Miss
Wk 10	BUF at MIA	13–30	BUF 74%	Miss

Sharpest calls of 2025

Right winner, and the model spread landed nearly on the final margin.

Week	Game	Final	Model pick	Result
Wk 1	BAL at BUF	40–41	BUF 54%	Hit
Wk 3	ARI at SF	15–16	SF 53%	Hit
Wk 6	LAC at MIA	29–27	LAC 57%	Hit
Wk 16	LAR at SEA	37–38	SEA 54%	Hit
Wk 9	LAC at TEN	27–20	LAC 74%	Hit

Week by week, 2025

Wk 1

9/16

Wk 2

11/16

Wk 3

9/16

Wk 4

7/16

Wk 5

7/14

Wk 6

9/15

Wk 7

10/15

Wk 8

9/13

Wk 9

8/14

Wk 10

10/14

Wk 11

11/15

Wk 12

11/14

Wk 13

9/16

Wk 14

11/14

Wk 15

9/16

Wk 16

12/16

Wk 17

12/16

Wk 18

10/16

Loading diagnostics…

Honest caveat: the 2025 backtest is in-sample in one sense — ratings were built from the same season — but each prediction only uses ratings as they stood before that kickoff, so no game informs its own forecast. Early-season weeks are the weakest, because every team starts the year near the league average.