DIAMOND_IN_THE_ROUGH // 2003_WEEK_17

THE 28–3 NOBODY REMEMBERS: MINNESOTA'S 4TH-AND-25 SEASON

Everyone can draw the Falcons' collapse from memory. Here's one with the same shape and none of the fame: a 9–6 Vikings team, a 3-win Arizona club, 96.9% win probability, and a season that ended on the last snap of the year.

BY PROFESSOR CHEDDARHEAD · 2026-07-05 · GAME: 2003-12-28 · DATA: NFLVERSE / CHEDDARHEAD_ANALYTICS

MIN

Minnesota Vikings

LOSS

—

ARI

Arizona Cardinals

WIN

MIN PEAK WIN PROB

96.9%

4th qtr, leading

FINAL-PLAY WP SWING

-59.1%

against Minnesota

MIN DRIVE EPA

+0.5

on offense

ARI DRIVE EPA

-5.5

on offense

Minnesota's win probability across all 152 plays of the 2003 Week 17 game at Arizona. It hovers around 50% for three quarters, climbs steadily to a peak of 96.9% late in the fourth quarter, then drops almost vertically on the final play — a 59.1-point swing — ending in a loss. — **MINNESOTA'S WIN PROBABILITY, EVERY PLAY.** Three quarters of coin flip, one slow climb to 96.9%, and then a single vertical line down. That last drop is one play: a fourth-and-25 touchdown as the clock hit zero.
[ OPEN THIS GAME IN DRIVE_TRACKER → ]

A GAME THAT DECIDED A SEASON

On paper this was garbage time for a whole franchise. Arizona came in at 3–12, eliminated, playing out the string in a mostly empty stadium. Minnesota came in at 9–6, needing a win to stay in front of Green Bay in the NFC North. The Vikings had led the division for most of the season. Win this and they are 10–6 with the tiebreaker edge — a better conference record than the Packers, 8–4 to 8–6 — and they are hosting a playoff game. Lose, and the math turns cruel: 9–7, and Green Bay's 10–6 takes the North. There was no wild-card safety net. This was the whole season, against a team with nothing to play for.

For fifty-eight minutes, Minnesota did enough. They led 17–6. The win-probability line spent the fourth quarter climbing toward certainty. Then a 3–12 team playing out a lost season put together the only two drives it needed all year.

Pull up the Drive Tracker line and the shape is almost comic in its calm. For three quarters this is a coin flip between two teams that can't get out of their own way — the win-probability line wanders around 50% like it can't find the exit. Then Minnesota scores twice in the middle of the fourth, the line climbs into the 90s, and everything looks decided. What the chart can't show you is the thing that made it matter: a scoreboard in one city and a standings page in another, both waiting on this game between a contender and a corpse.

THE BETTER TEAM LOST (AND WASN'T EVEN GOOD)

Expected points added per drive for both teams. Most bars are small and negative for both offenses. Minnesota has two clear positive touchdown drives worth about +6.2 and +5.6 EPA. Arizona's two tallest bars are its final two touchdown drives, including the game-winner. — **EXPECTED POINTS ADDED, PER DRIVE.** Look how short most of these bars are. Neither offense was any good — Minnesota finished at **+0.5** total drive EPA, Arizona at **−5.5**. The Vikings were the better team by the numbers. "Better," here, means "less bad."

This is the honest part the highlight would skip: nobody played well. Minnesota's offense generated barely positive expected points for the day. Arizona's was flatly negative — a genuinely bad offense having a genuinely bad day, right up until it wasn't. The Vikings didn't blow a lead because they collapsed statistically. They blew it because in a game between two mediocre teams, the mediocre team with nothing to lose made the two plays that a season turned on, and the mediocre team with everything to lose made none.

A strip of all 21 drive outcomes in sequence. Most cells are gray punts and red turnovers. Minnesota's two green touchdowns sit in the middle. Arizona's two green touchdowns are the last two cells on the right — back-to-back scores to end the game. — **DRIVE OUTCOMES, LEFT TO RIGHT.** A field of punts and turnovers — then Arizona's two green cells at the very end, back to back. The Cardinals scored on their last two possessions of a lost season. The second one won it.

FOURTH AND TWENTY-FIVE

Down 17–12 in the final minute, Arizona faced fourth-and-25 from the Minnesota 28. Our model gave the Vikings a 96.9% chance to win at their peak, and they were still comfortably in the 90s here. A stop — any stop, a batted ball, a checkdown, an incompletion — ends the season the right way. Instead Josh McCown escaped pressure and found Nathan Poole in the corner of the end zone as time expired. Touchdown. Final: 18–17.

THE SINGLE BIGGEST PLAY OF THE GAME: that fourth-and-25 touchdown swung Minnesota's win probability by −59.1 points on one snap — from a game they were overwhelmingly winning to a game they had lost, with no time to answer. There is no bigger single-play swing anywhere in this game. There rarely is one bigger in any game. It is the entire vertical drop on the chart above, compressed into four seconds.

Minnesota finished 9–7 and stayed home. Green Bay finished 10–6 and won the division that had been Minnesota's to keep. Arizona finished 4–12, and this was one of their four wins — a throwaway result for them, a season-ending guillotine for someone else.

WHAT THE NUMBERS SAY vs WHAT THE STORY SAYS

// THE NUMBERS

Minnesota was the better team: +0.5 to −5.5 in drive EPA, and 96.9% to win at the peak. By every measure the model has, the Vikings should have been in the playoffs. The data doesn't see "must-win." It just saw two below-average offenses and gave the lead to the one that had one.

// THE STORY

Football doesn't award the season to the better EPA. It awards it to whoever makes the last play. A 3-win team with nothing to lose played loose; a contender with everything to lose played tight. Fourth-and-25 is a 3% play. It hit. Seasons hang on 3% plays, and nobody hangs a banner for expected points.

THE DRIVE LOG

Every drive, in order, with the possessing team's expected points added and its net win-probability shift. Watch drive 21 — Arizona's game-winner — carry a +71.6% swing. Every figure here is computed from the play-by-play, not remembered.

2003 Week 17 drive log — ARI 18, MIN 17. EPA and WP shift are from the possessing team's perspective.
#	TEAM	RESULT	DRIVE EPA	WP SHIFT
1	ARI	PUNT	-1.83	-4.7%
2	MIN	DOWNS	+0.04	-1.3%
3	ARI	FG	+3.47	+8.6%
4	MIN	MISS	-1.57	-5.9%
5	ARI	PUNT	-1.03	-3.9%
6	MIN	PUNT	-1.46	-6.8%
7	ARI	PUNT	-1.86	-5.2%
8	MIN	TO	-4.13	-12.2%
9	ARI	FG	-1.28	-5.7%
10	MIN	TO	-0.17	-3.1%
11	MIN	PUNT	-1.77	-3.5%
12	ARI	PUNT	-2.25	-3.9%
13	MIN	TD	+6.18	+16.9%
14	ARI	PUNT	-1.35	-9.3%
15	MIN	PUNT	-1.55	-10.5%
16	ARI	PUNT	-3.40	-16.8%
17	MIN	TD	+5.57	+25.8%
18	ARI	TO	-4.97	-10.9%
19	MIN	FG	-0.65	+0.2%
20	ARI	TD	+4.37	-1.5%
21	ARI	TD	+4.67	+71.6%

THE PROFESSOR'S VERDICT

PROFESSOR'S VERDICT

Super Bowl LI gets a documentary; this gets a footnote. Same shape, same 90-plus percent certainty erased on the final drive — the only difference is the marquee. If you want a lesson from the Film Room, it's this: win probability is real, and it is not a receipt. A 96.9% lead is a very good place to be and a terrible place to relax. Minnesota had the better team, the better EPA, and the division in hand, and lost all three to a fourth-and-25 nobody outside of Arizona has thought about since. The data remembered. That's the job.

[ DRIVE_TRACKER ] [ MIN_HUB ] [ ARI_HUB ] [ WHAT_IS_WIN_PROBABILITY? ] [ THE_FAMOUS_28-3 → ]