Comebacks, collapses and upsets

Every NFL game since 1999, ranked three ways: the deepest holes winners climbed out of, the biggest upsets by the betting line, and the wildest win-probability rides. 2026 is through week 9.

Season Team Games

Comebacks and collapses

How low the winner’s chances fell at any snap. Every comeback is also a collapse: at that moment the loser’s chance was 100% minus the winner’s. Pick a team to see both sides of its games.

Rank by

#	Game	Winner’s lowest WP (our model)	Winner’s lowest WP (nflfastR)	Most points behind	Low point	Came back	Collapsed	Date
1	IND 38 at TB 35	0.12%	0.13%	21	Q4 3:43, 14–35	IND	TB	2003-10-06Wk 5
2	CLE 21 at CHI 27	0.16%	0.24%	14	Q4 3:09, 7–21	CHI	CLE	2001-11-04Wk 8
3	NYJ 31 at CLE 30	0.26%	0.24%	13	Q4 1:55, 17–30	NYJ	CLE	2022-09-18Wk 2
4	PIT 23 at CIN 26	0.26%	0.41%	14	Q4 4:01, 10–23	CIN	PIT	2001-12-30Wk 16
5	PHI 38 at NYG 31	0.27%	0.17%	21	Q4 7:43, 10–31	PHI	NYG	2010-12-19Wk 15
6	NYG 21 at TEN 24	0.28%	0.36%	21	Q4 13:49, 0–21	TEN	NYG	2006-11-26Wk 12
7	IND 36 at MIN 39	0.34%	0.11%	33	Q3 12:57, 0–33	MIN	IND	2022-12-17Wk 15
8	DAL 21 at PHI 20	0.37%	0.46%	13	Q4 3:51, 7–20	DAL	PHI	2005-11-14Wk 10
9	TB 27 at CAR 21	0.48%	0.96%	11	Q4 1:09, 13–21	TB	CAR	2012-11-18Wk 11
10	DEN 18 at MIA 15	0.53%	0.59%	15	Q4 5:50, 0–15	DEN	MIA	2011-10-23Wk 7
11	IND 35 at DET 33	0.56%	0.65%	12	Q4 4:24, 21–33	IND	DET	2012-12-02Wk 13
12	LA 33 at SEA 27	0.61%	0.93%	17	Q4 7:47, 10–27	LA	SEA	2004-10-10Wk 5
13	SEA 41 at BAL 44	0.62%	1.0%	17	Q4 6:56, 24–41	BAL	SEA	2003-11-23Wk 12
14	TEN 28 at MIA 27	0.66%	0.60%	14	Q4 3:08, 13–27	TEN	MIA	2023-12-11Wk 14
15	CLE 31 at TEN 28	0.70%	1.2%	14	Q4 3:08, 14–28	CLE	TEN	2002-09-22Wk 3
16	BAL 38 at DEN 35	0.73%	1.4%	7	Q4 1:15, 28–35	BAL	DEN	2013-01-12Playoffs
17	KC 44 at IND 45	0.81%	0.29%	28	Q3 13:20, 10–38	IND	KC	2014-01-04Playoffs
18	DET 34 at DAL 30	0.89%	0.20%	24	Q3 10:30, 3–27	DET	DAL	2011-10-02Wk 4
19	LAC 30 at JAX 31	0.91%	0.13%	27	Q2 2:26, 0–27	JAX	LAC	2023-01-14Playoffs
20	IND 31 at HOU 27	0.93%	0.69%	17	Q4 4:10, 10–27	IND	HOU	2008-10-05Wk 5
21	TB 20 at NO 17	0.99%	1.1%	17	Q2 8:13, 0–14	TB	NO	2009-12-27Wk 16
22	MIA 37 at NYJ 40	1.1%	0.40%	23	Q4 14:46, 7–30	NYJ	MIA	2000-10-23Wk 8
23	LA 33 at HOU 27	1.1%	2.1%	21	Q4 0:34, 17–27	LA	HOU	2005-11-27Wk 12
24	BAL 40 at BUF 41	1.1%	0.73%	15	Q4 7:27, 25–40	BUF	BAL	2025-09-07Wk 1
25	SF 24 at PHI 23	1.1%	1.1%	20	Q3 11:01, 3–20	SF	PHI	2011-10-02Wk 4

Biggest upsets

Underdogs that won, ranked by their chance before kickoff from the closing point spread (an underdog by 7 points wins about 27% of the time).

#	Game	Underdog by (points)	Winner’s chance before kickoff	Won by	Winner	Loser	Date
1	MIA 27 at NE 24	17.5	6.5%	3	MIA	NE	2019-12-29Wk 17
2	NYJ 23 at LA 20	17.5	6.5%	3	NYJ	LA	2020-12-20Wk 15
3	BUF 27 at MIN 6	16.5	7.7%	21	BUF	MIN	2018-09-23Wk 3
4	PIT 17 at CIN 27	14.5	10.5%	10	CIN	PIT	2020-12-21Wk 15
5	LV 27 at PIT 24	14.5	10.5%	3	LV	PIT	2009-12-06Wk 13
6	BUF 6 at JAX 9	14.5	10.5%	3	JAX	BUF	2021-11-07Wk 9
7	CLE 16 at PIT 15	14.5	10.5%	1	CLE	PIT	1999-11-14Wk 10
8	ARI 21 at PHI 20	14.5	10.5%	1	ARI	PHI	2001-10-07Wk 4
9	HOU 24 at PIT 6	14.0	11.3%	18	HOU	PIT	2002-12-08Wk 14
10	ATL 26 at NO 9	14.0	11.3%	17	ATL	NO	2019-11-10Wk 10
11	IND 11 at JAX 26	14.0	11.3%	15	JAX	IND	2022-01-09Wk 18
12	NYG 23 at DEN 10	14.0	11.3%	13	NYG	DEN	2017-10-15Wk 6
13	MIN 24 at PHI 14	14.0	11.3%	10	MIN	PHI	2010-12-28Wk 16
14	PHI 9 at LV 13	14.0	11.3%	4	LV	PHI	2009-10-18Wk 6
15	LA 17 at NE 20	14.0	11.3%	3	NE	LA	2002-02-03Playoffs
16	TB 20 at NO 17	14.0	11.3%	3	TB	NO	2009-12-27Wk 16
17	MIA 20 at ATL 17	14.0	11.3%	3	MIA	ATL	2017-10-15Wk 6
18	HOU 21 at MIA 20	14.0	11.3%	1	HOU	MIA	2003-09-07Wk 1
19	TEN 28 at MIA 27	14.0	11.3%	1	TEN	MIA	2023-12-11Wk 14
20	LAC 29 at HOU 41	13.5	12.2%	12	HOU	LAC	2021-12-26Wk 16
21	NO 21 at LA 31	13.5	12.2%	10	LA	NO	2011-10-30Wk 8
22	DET 39 at DAL 31	13.5	12.2%	8	DET	DAL	2006-12-31Wk 17
23	NO 31 at LA 24	13.5	12.2%	7	NO	LA	2000-11-26Wk 13
24	PHI 30 at LA 23	13.5	12.2%	7	PHI	LA	2018-12-16Wk 15
25	ARI 21 at PHI 17	13.5	12.2%	4	ARI	PHI	2011-11-13Wk 10

Thrillers

How far the win-probability line travelled: the sum of every snap-to-snap change, plus the final result. 100 points is one full swing from 0% to 100%.

Rank by

#	Game	WP points moved	Lead changes	Overtime	Winner’s lowest WP (our model)	Winner	Date
1	HOU 34 at DET 31	919	1	OT	11.7%	HOU	2012-11-22Wk 12
2	BUF 45 at MIN 39	884	9	OT	9.6%	BUF	2002-09-15Wk 2
3	NYG 37 at DAL 40	873	7	OT	9.3%	DAL	2025-09-14Wk 2
4	PIT 23 at CIN 20	866	0	OT	6.9%	PIT	2022-09-11Wk 1
5	LV 30 at TB 24	860	3	OT	12.6%	LV	2016-10-30Wk 8
6	BAL 9 at CLE 12	857	1	OT	22.8%	CLE	2018-10-07Wk 5
7	NYG 21 at SEA 24	853	2	OT	11.1%	SEA	2005-11-27Wk 12
8	JAX 13 at IND 16	851	2	OT	13.0%	IND	2015-10-04Wk 4
9	CHI 20 at MIN 23	850	3	OT	10.3%	MIN	2013-12-01Wk 13
10	NE 13 at BUF 10	826	2	OT	9.5%	NE	2000-12-17Wk 16
11	GB 25 at CIN 22	819	1	OT	19.3%	GB	2021-10-10Wk 5
12	CLE 42 at LV 45	815	4	OT	1.5%	LV	2018-09-30Wk 4
13	KC 30 at DEN 27	798	4	OT	3.4%	KC	2016-11-27Wk 12
14	DEN 27 at LAC 30	793	1	OT	17.5%	LAC	2002-12-01Wk 13
15	LA 16 at SF 19	786	3	OT	13.3%	SF	2016-01-03Wk 17
16	PHI 19 at LAC 22	785	2	OT	20.3%	LAC	2025-12-08Wk 14
17	DEN 23 at KC 26	785	0	OT	27.2%	KC	2001-12-16Wk 14
18	SF 14 at ATL 16	782	1		29.0%	ATL	2010-10-03Wk 4
19	MIN 17 at DET 20	780	0	OT	19.4%	DET	2007-09-16Wk 2
20	SEA 27 at SF 24	779	1	OT	8.6%	SEA	2019-11-11Wk 10
21	CHI 34 at DET 37	770	4	OT	16.4%	DET	2015-10-18Wk 6
22	LV 34 at SF 28	767	2	OT	14.8%	LV	2000-10-08Wk 6
23	NYJ 23 at MIA 20	765	1	OT	19.1%	NYJ	2012-09-23Wk 3
24	ATL 24 at TB 27	763	3	OT	16.6%	TB	2005-12-24Wk 16
25	SF 23 at IND 26	760	0	OT	26.0%	IND	2017-10-08Wk 5

How these are computed

Win probability is our model’s (the registered win-probability model behind the 4th-down tool, point estimate), read at every snap. nflfastR’s (vegas_wp) is read at the same snaps. For 1999–2025 both are in-sample values.
Only snaps count, not kickoffs or extra points: there our model treats the next kickoff as a normal kick, so a trailing team about to try an onside kick late in a game would look almost hopeless.
Most points behind: the largest deficit the winner faced at any point. Low point: the quarter, clock and score (winner first) at the snap where the winner’s chance was lowest.
Upsets: the closing point spread from nflverse’s schedules. Chance before kickoff = Φ(spread ÷ 11.6), a probit fitted on 7,238 decided games from 1999–2025. We use the betting line here rather than our model’s kickoff win probability because our model understates very big underdogs (15+ points), which would push those games up the list.
Lead changes: the times the lead passed from one team to the other; a tie in between doesn’t count as a change.
Left out: tied games (no winner) and the 20 games whose source play-by-play failed our data-quality checks.
Teams appear under their current abbreviation, as in nflfastR: the St. Louis Rams as LA, the Oakland Raiders as LV, the San Diego Chargers as LAC.