Two Weeks In: Who the Market Got Wrong, and Why It Won't Stay Wrong

Through two weeks of the 2026 season, the Los Angeles Chargers have lost twice at home, by twelve points each time, while favored in both games. They entered the season carrying the largest average point spread in football — the market's single strongest expression of confidence in any team — and they have now missed that expectation by a combined 39 points.

That is the 14th-worst two-week start in 796 team-seasons since 2002. It is a genuine, measurable surprise.

It is also worth almost nothing to you in Week 3, and this article is about why.

What "surprising" actually means

A 2-0 record against the spread is not, by itself, interesting. A team can beat the number by half a point twice and post the same record as a team that buried it twice.

The honest measure is cover margin: how far a team's results landed from what the market expected, summed across both games. That figure has a direction and a size, and it is what separates a mild overperformance from a genuine market miss.

By that measure, here is where 2026 actually stands after two weeks.

Beat the market's expectation by the most

Negative spread = favored. "ATS record" is against the spread; "straight-up record" is wins and losses.

Seven teams are 2-0 against the spread. Two of them — the Saints and the Jets — are only 1-1 straight up, which is the distinction that gets lost when people list "unbeaten" teams.

The Saints are the purest case. They faced an average spread of +7.75, the second-largest underdog role in the league, and beat it twice: a one-point road loss in Detroit as a 7-point underdog, then an outright road win in Baltimore as an 8.5-point underdog. They have one win. They are also one of the two most underrated teams in football by this measure.

The Raiders are the other. Average spread faced: +1.75 — the market saw a roughly neutral team. They won both games, including by twelve in Los Angeles, and sit at +29.5.

Missed the market's expectation by the most

The Chargers lead by a distance, and the gap between their expectation and their results is the story of the season so far. The Falcons are second and got there strangely: they missed by half a point in Pittsburgh in Week 1, then lost 34-3 at home to Carolina.

And then there are the Eagles, who are 2-0 straight up and 0-2 against the spread. They have won every game they have played and disappointed the market in both. Both readings are correct; they are just answering different questions.

What 24 seasons say happens next

This is where the useful part starts. We took every team-season from 2002 through 2025, sorted each team by how its first two regular-season games went, and measured what that team did from Week 3 onward.

That is 764 team-seasons and roughly 11,000 subsequent games.

Sorted by how the first two weeks went against the spread

A 2-0 start against the spread is followed by 49.4% accuracy the rest of the way. A 0-2 start is followed by 50.6%. Both are coin flips, and the tiny difference between them points the opposite direction to intuition.

Sorted by how the first two weeks went straight up

This is the clearest version of the finding, and it is two results at once:

• A 2-0 start tells you a great deal about how good a team is. Those teams go on to win 55.8% of their remaining games straight up. The 0-2 teams win 40.8%. That is a fifteen-point gap in real quality, and it is not noise.
• It tells you nothing about how that team will perform against the number — except that 2-0 teams are the worst of the three groups at 48.5%.

The market reads those first two weeks at least as fast as you do, and then charges for what it learned. By Week 3 the good news is already in the price.

The two most extreme starts in 24 years

The cohort tables are the evidence. These two seasons are the illustration.

The best two-week start since 2002 belongs to the 2024 Saints, who beat expectations by a combined +64.5 points and were 2-0 both ways. From Week 3 on, they went 4-10-1 against the spread and 3-12 straight up.

The worst belongs to the 2011 Chiefs, at −75.0. From Week 3 on, they went 9-5 against the spread.

Widen it slightly and the pattern holds. The four hottest starts on record went a combined 23-34 (40.4%) against the spread afterward. The three coldest went 26-17 (60.5%).

Those are small samples and we are not presenting them as proof — 57 games and 43 games decide nothing on their own. They are here because they show the shape of what the 764-team-season table measures: the correction is not gentle, and it frequently overshoots.

So what is a two-week start good for?

Something, but a narrower something than it looks.

Use it as evidence about quality. The 55.8%-to-40.8% straight-up gap is real and large. The teams sitting at 2-0 are, on average, genuinely better than the teams at 0-2, and they will keep winning more games. That is a legitimate thing to learn in September.

Do not use it as evidence about future market misses. That is the part that has already been priced. A team that beat expectations twice will face a larger number next week, and across 24 seasons that adjustment has been almost exactly right — 49.4% and 50.6%, on either side of a coin flip.

This is the same mechanism we found in our look at five-game form, measured a different way: hot teams keep winning at 65.7% and keep covering at 50.2%, because the spread moves ten points to meet them. Two weeks into a season is simply the earliest, loudest version of that story.

It also sits alongside what we found about our own model's accuracy by week — that September is the model's sharpest stretch, not its foggiest, because efficiency data carries over from last season while records and narratives do not. Two games of won-lost evidence is the least durable thing on the table in Week 3. The efficiency numbers underneath those games are worth considerably more.

The honest caveats

• Two games is two games. Every 2026 figure here rests on a sample of two per team. The Chargers' −39.0 is a fact about what happened, not a forecast, and nothing in this article claims to know which way it resolves.
• The historical cohorts exclude four team-seasons where a team did not play exactly two games in Weeks 1-2 (hurricane postponements in 2008 and 2017).
• Closing numbers only. All spreads here are closing numbers. This says nothing about where lines opened or how they moved during the week.
• Pushes and ties are excluded from every percentage — each rate is wins over decisions. Across the rest-of-season sample that is 281 ATS pushes and 11 tied games.
• The rest-of-season windows are not equal length. Teams eliminated from contention still play; teams resting starters in Week 18 are in these samples too. That is a known contaminant in the late-season portion and we have not removed it.
• Regular season only. Playoff games are excluded from both the start and the rest-of-season measurements.

See it yourself

Every number above comes from pages you can open:

• Standings — 2026 records against the spread and straight up, for every team
• Spreads — how results break down by the size of the number
• Breakdown — any team's game-by-game log with the spread it faced
• Predictions — our graded record, published in full, wins and losses alike

Week 3 opens Thursday night, and the numbers have already moved toward the 2-0 teams and away from the 0-2 teams. That is the market doing exactly what the last 24 seasons say it does well.